I came across a Japanese YouTube about screw compressors.

The commenter had worked for a company that manufactured them. The unusual part was the rotor machining. The tooth profile required such specialized techniques that the machining area was physically enclosed inside the factory. Even other employees were generally not allowed to see it, partly to prevent visitors from learning how it was done.

At first glance, this sounds like an obscure piece of industrial trivia.

It is actually a useful way to understand why advanced manufacturing still matters to wealthy countries—and perhaps why it matters to democracy.

A screw compressor is not technologically impressive in the way an AI model or smartphone appears impressive. Two precisely shaped rotors turn together and compress gas.

But making those rotors extremely well is difficult.

Mayekawa, one of Japan's major compressor manufacturers, describes newer rotor profiles as requiring “high-level production techniques”; improving the profile reduces internal gas leakage and increases efficiency.

Japan did not invent the modern screw compressor. Kobe Steel entered the business in 1955 through a technical alliance with Sweden's SRM and produced Japan's first domestically manufactured oil-free screw compressor the following year.

But then something important happened.

Japan kept making them.

And making them.

And making them.

Kobe eventually developed new proprietary rotor geometries, expanded the technology into refrigeration and process gases, and accumulated decades of experience in machining, metallurgy, seals, tolerances, testing and failure analysis.

That is a very different kind of technological advantage from writing a piece of software.

Some technologies take decades to become good at

A sophisticated industrial product is rarely just a blueprint.

The real capability is distributed across:

machine tools,
cutting tools,
metallurgy,
heat treatment,
metrology,
control systems,
suppliers,
production engineers,
technicians,
maintenance workers,
quality systems,
and thousands of small lessons accumulated after things go wrong.

A competitor can buy your compressor and measure it. It can obtain patents. It can hire engineers.

It still does not instantly inherit 20 or 50 years of production experience.

That matters economically because this capability is sticky.

A software founder can move from Toronto to San Francisco. Intellectual property can be transferred between corporate entities. A hedge fund can change jurisdictions. A technology company can shift headquarters surprisingly quickly.

An industrial ecosystem is harder to move.

You cannot instantaneously relocate a cluster containing machine shops, toolmakers, experienced production engineers, specialized chemical suppliers, vocational schools, testing facilities and thousands of technicians who have spent their careers solving related problems.

Germany and Japan still possess millions of workers embedded in these systems. Germany had about 7.25 million people working in manufacturing in 2025, roughly 16% of total employment. Japan had about 10.33 million manufacturing workers in 2025.

Canada, by comparison, had only 8.8% of employment in manufacturing in 2025, down from 19.1% in 1976.

That difference may matter for more than GDP.

The forgotten political virtue of advanced manufacturing

There is an old democratic bargain hidden inside an industrial economy.

Consider what it takes to operate a globally competitive machinery, automotive, electronics or chemical industry.

A country needs:

engineers,
machinists,
electricians,
technicians,
toolmakers,
production workers,
maintenance workers,
logistics workers,
supplier companies,
and managers.

Not everyone needs to attend an elite university.

But a very large fraction of the population needs to be healthy, numerate, technically competent and reliable.

That gives both firms and governments a direct economic interest in the quality of the broad population.

The feedback loop looks something like:

good mass education
→ skilled workers
→ sophisticated production
→ valuable exports
→ higher wages and tax revenue
→ better infrastructure and training
→ still more sophisticated production.

The important feature is not simply that factory workers have jobs.

It is that ordinary citizens are economically necessary to producing the national surplus.

That can create a different political economy from one in which most internationally competitive income is generated by a relatively small group of financiers, AI researchers, software founders, resource companies or owners of intellectual property.

In the latter economy, society can still be rich.

But the relationship changes.

A relatively small productive layer generates enormous surplus. The state taxes some of it. That income circulates through healthcare, construction, education, retail, restaurants and other domestic services.

The median citizen may still have a comfortable job.

But the country's ability to earn income from the rest of the world depends much less directly on that citizen's productive capability.

This does not prove that manufacturing produces democracy. It plainly does not: authoritarian states can have formidable manufacturing industries, while democratic countries can prosper with service-heavy economies.

But there is evidence that economic agency matters politically.

Research on unions finds that they historically reduced economic disparities partly by increasing the political representation of ordinary workers. The OECD's 2026 trust survey found an enormous 47-percentage-point gap in trust in national government between people who felt they had political voice and those who did not. Financial insecurity was also strongly associated with lower trust.

And research on the American “China shock” found that communities facing greater import competition suffered persistent employment and income losses, along with significant political realignment and polarization.

The relevant issue may therefore be less “factory jobs versus service jobs” than productive agency.

Why construction and healthcare are different

Construction and healthcare can provide excellent jobs.

An electrician may earn more than a machinist. A nurse may earn more than either.

Both industries are socially indispensable.

But they occupy a different place in the national economic system because they are overwhelmingly nontradable.

A Toronto electrician cannot wire an apartment in Seoul and bring foreign income back to Canada.

A nurse treating a Canadian patient usually does not create export revenue.

A Canadian machine-tool company selling equipment to Korea does.

This distinction matters because prosperous domestic services ultimately require purchasing power generated somewhere.

Construction converts income and capital into domestic assets. Healthcare converts income into health and care. Neither is automatically a source of the foreign purchasing power required to import Taiwanese semiconductors, Japanese machinery, German chemicals or Korean cars.

Construction is also unusually sensitive to credit.

Canada illustrates the scale. Residential investment supported more than 1.2 million Canadian jobs and C$152.1 billion of GDP in 2025. Canada's housing stock was worth about C$4.4 trillion, roughly one-quarter of national wealth.

Meanwhile residential mortgage debt reached approximately C$2.44 trillion by June 2026.

That does not make construction artificial. Houses are real capital.

But there is an obvious difference between:

selling a C$5 million industrial machine abroad

and

borrowing against future Canadian household income to construct a C$1 million house today.

The first earns an external claim.

The second creates a domestic asset matched partly by domestic debt.

Healthcare has a different financing problem. It is not primarily mortgage-driven, but in wealthy countries a large fraction of healthcare demand is collectively financed through taxes and social insurance. In the United States, Medicare and Medicaid alone paid for about 39% of total healthcare expenditure in 2024. Aging societies make that demand increasingly difficult to reduce.

So neither construction nor healthcare is “fake.”

The problem arises when a country increasingly relies on these sectors as substitutes for a weakening tradable productive base.

Eventually the question becomes: what generates the income that supports them?

Singapore and Saudi Arabia show another solution

Small rich economies have discovered a different way of resolving this tension: import the workers.

Singapore intentionally segments its labor market. Its current rules allow foreign Work Permit and S Pass workers to constitute as much as 83.3% of employment in construction, 60% in manufacturing and 35% in services, subject to quotas and levies.

This allows Singapore to specialize a relatively small citizen population more heavily toward finance, technology, management, engineering and other high-productivity activities while importing workers for large parts of construction and other labor-intensive sectors.

The Gulf states represent a more extreme version. The ILO describes the Arab Gulf as having the world's highest migrant-worker shares, with migrant workers heavily concentrated in construction, hospitality, domestic work and other labor-intensive activities.

Economically, this creates an adjustment mechanism.

When export income or construction demand falls, a country with a large temporary foreign workforce can reduce new permits, allow contracts to expire or slow recruitment.

The resulting labor-market adjustment does not have to remain entirely inside the citizen population.

Japan historically had much less access to this mechanism.

When Japan lost industrial pricing power during the rise of Korea, Taiwan, China and Southeast Asia, the people standing at the bottom of the labor market were overwhelmingly Japanese residents.

Japan could not export its unemployed population.

Part of the adjustment therefore appeared internally as fewer good jobs for young workers, non-regular employment, weaker wage growth, supplier pressure and eventually the large 氷河期世代 underclass Japan is still dealing with today.

Canada chose yet another model

Canada after approximately 2014 is particularly interesting.

Manufacturing continued declining as a share of employment. The commodity investment boom ended after the oil-price crash. Instead of accepting something resembling Japan's demographic contraction, Canada dramatically increased population growth.

Population growth itself created demand:

more people
→ more housing
→ more construction
→ more mortgages
→ more banking and real estate activity
→ more retail
→ more healthcare and education
→ more government services.

For a time, this was an extremely effective aggregate-growth machine.

But it did not solve Canada's underlying productivity problem.

And Canada's industrial position was already weak by German or Japanese standards.

Even before the current tariff conflict, the federal government's Canadian Occupational Projection System projected manufacturing's employment share falling from 9.0% in 2023 to 8.5% in 2033, while construction and services continued gaining employment.

The tariff conflict has made that projection look increasingly optimistic.

Canadian manufacturing output fell 2.6% in 2025, its third consecutive annual decline, while manufacturing payroll employment dropped by about 40,600 jobs during the year.

The Bank of Canada now explicitly expects tariffs to damage potential output partly by reducing investment in affected industries and reallocating workers and capital toward less productive sectors.

That is almost the textbook mechanism of gradual deindustrialization.

That would leave Canada increasingly dependent on some combination of:

resources,
technology and professional services,
finance,
foreign investment income,
construction,
healthcare,
and government-supported domestic services.

Canada can remain a rich country under such a structure.

The harder question is what happens to broad productive employment.

Immigration may become the adjustment valve

This is where I expect Canada's adjustment to become politically much more difficult than current immigration policy suggests.

The first stage has already begun.

Canada's 2026–28 immigration plan cuts new temporary-resident arrivals to 385,000 in 2026 and 370,000 thereafter, while holding permanent-resident admissions at 380,000. The government also intends to reduce temporary residents to below 5% of Canada's population.

That is the easy adjustment.

Temporary workers and students are explicitly temporary. Governments can issue fewer permits, tighten eligibility and allow existing permits to expire.

The more difficult question comes later.

What happens if, during the 2030s, Canada has a substantially larger permanent population but fewer internationally competitive jobs per resident?

Suppose manufacturing continues declining, resource industries remain extraordinarily productive but employ relatively few people, AI and finance create enormous income for a narrow professional class, and the housing sector can no longer expand through ever-rising household leverage.

At that point Canada is no longer deciding merely how many additional workers to admit.

It is deciding how to distribute a limited stock of high-productivity employment, housing, infrastructure and fiscal capacity among the population already here.

My expectation is that political pressure would then move progressively through the immigration hierarchy:

temporary workers and students first
→ fewer new permanent residents
→ much more selective permanent residence
→ eventually, pressure to reconsider the rights and obligations attached to permanent residence itself.

This last step would be politically and legally much harder.

Under current law, an expired PR card does not end permanent-resident status. Permanent residents generally lose status only through defined legal processes such as failure of the residency obligation followed by a formal determination, a removal order, voluntary renunciation or citizenship. Current law requires 730 qualifying days during each five-year period.

But permanent residence is not constitutionally identical to citizenship.

The Canadian Charter gives citizens the explicit constitutional right to enter, remain in and leave Canada. Permanent residents receive important mobility and livelihood rights, but their right to enter and remain in Canada is established by the Immigration and Refugee Protection Act and is subject to that Act.

That distinction could become politically important in a prolonged period of economic compression.

I do not expect Canada suddenly to announce that hundreds of thousands of economically unsuccessful permanent residents must leave.

A more plausible path would be incremental:

tighter physical-presence requirements,
more aggressive enforcement of residency obligations,
fewer exemptions,
greater selectivity before citizenship,
changes to eligibility or waiting periods for particular publicly funded benefits where legally permissible,
and potentially new debates about how unconditional permanent residence should be before citizenship.

Some of these measures would require legislative changes and could face serious Charter and court challenges. There is no current government proposal to carry out this program.

The prediction is instead about where the political pressure goes.

Canada's immigration system was constructed during an era when policymakers generally assumed that adding working-age residents increased Canada's productive capacity.

If the economy instead enters a period in which workers become abundant relative to internationally competitive jobs, that assumption can reverse.

Then immigration status itself becomes a labor-market buffer.

This resembles Singapore's model, but with an important difference.

Singapore intentionally maintains a large class of workers whose continued presence is conditional on employment permits. When labor demand falls, it can reduce foreign-worker quotas or simply issue fewer renewals.

Canada converted a much larger portion of its migrant population into permanent residents.

That means Canada has progressively transformed what could have been a flexible foreign-labor buffer into part of the permanent domestic population.

If economic conditions deteriorate, those people cannot simply be treated like temporary workers under current Canadian law.

That makes adjustment much harder.

But it also creates political incentives to draw a progressively sharper distinction between:

citizen
permanent resident
temporary resident.

In a rapidly expanding economy, those distinctions can become less economically important.

In a stagnant economy facing competition for good jobs, housing and public expenditure, they may become considerably more important.

That is why I expect the current restrictions on students and temporary workers to be the beginning of Canada's immigration adjustment rather than necessarily its endpoint.

If Canada's internationally productive economy continues weakening relative to its population through the 2030s, the political debate may eventually cease to be merely:

How many new immigrants should Canada admit?

and become:

What obligations does Canada owe to non-citizens already granted permanent status, and what obligations should continued permanent status impose in return?

That would mark a profound reversal from the political economy of Canada between roughly 2015 and 2024.

The thing that cannot be recreated quickly

This brings us back to the screw compressor.

The attraction of advanced manufacturing is not nostalgia for factories.

Low-value assembly is not inherently desirable. A wealthy country should not try to preserve every factory job merely because it involves physically making something.

The strategically interesting industries are the ones in which decades of accumulated knowledge allow ordinary skilled workers to participate in production valuable enough to support rich-country wages.

A screw-compressor rotor.

A precision reducer.

A semiconductor-production subsystem.

An aircraft engine component.

An industrial robot.

A specialized chemical.

A high-end machine tool.

These industries produce something unusual:

high productivity + international tradability + broad occupational participation + geographically sticky accumulated knowledge.

AI and finance can produce extraordinary wealth, but their direct employment bases are relatively narrow and their highest-value people and capital are unusually mobile.

Construction and healthcare can employ millions, but their prosperity ultimately depends heavily on income and fiscal capacity generated elsewhere.

Resources can produce enormous export rents, but require surprisingly few workers.

Advanced manufacturing is one of the few economic systems capable of putting a large middle of society directly inside internationally competitive production.

That may turn out to matter politically as much as economically.

The strongest version of the argument is not:

Factories create democracy.

They don't.

It is:

A democracy may be more stable when national prosperity depends on the competence, skills and productivity of a broad part of its own population.

If millions of ordinary citizens are necessary to produce what the rest of the world wants, then educating them, training them, keeping infrastructure functional and allowing them to participate in rising productivity is not merely social policy.

It is economic necessity.

The alternative can still be prosperous.

A narrow layer of AI, finance, resource and intellectual-property winners can generate enormous wealth. That wealth can support millions of people working in healthcare, construction and local services.

Singapore and the Gulf demonstrate versions of this model at small scale, supplemented by large imported workforces.

But for countries with tens or hundreds of millions of permanent citizens, the political economy is different.

The question eventually becomes not merely:

How rich is the country?

but:

How many of its citizens are genuinely necessary to producing that wealth?

Germany, Japan, Korea and Taiwan still have unusually strong answers to that question.

Canada increasingly does not.

And if China's industrial rise begins compressing the remaining rents of Japan and Germany as well, the coming decades may provide one of the largest tests yet of whether rich democracies can preserve broad economic agency after the industrial system that historically produced it begins to shrink.

For most of the 20th century, advanced societies became increasingly good at finding intelligent people and giving them important things to do.

Schools sorted students by grades. Universities selected for academic performance. Professional schools filtered again. Corporations added interviews, aptitude tests, credentials and technical assessments. Science rewarded publication and citation. Engineering rewarded the ability to solve increasingly abstract problems.

It wasn't literally an IQ test. But the machinery pointed in roughly the same direction.

IQ itself is approximately normally distributed, conventionally centered around 100 with a standard deviation of 15. That means roughly 68% of people fall between 85 and 115. Around 16% score above 115. Around 2% score above 130.

Modern institutions created increasingly lucrative environments for that upper tail.

If you were unusually good at abstraction, symbolic manipulation, learning from text, holding complicated systems in your head and producing correct answers, the 20th century gave you more and more places to convert that ability into status and money.

You could become a physician, lawyer, engineer, scientist, programmer, analyst or executive.

This changed culture too.

We began to associate being useful with being cognitively impressive.

The smart kid was told to study harder. The successful adult was encouraged to keep learning, read widely, develop expertise, remain curious and reinvent themselves. Silicon Valley pushed this worldview even further: learn quickly, reason from first principles, teach yourself new fields, write code, absorb enormous quantities of information.

For people near the upper end of the cognitive distribution, this advice often works remarkably well.

There is only one problem.

We may now be automating a substantial fraction of the thing our institutions spent a century learning to reward.

AI has an unusually intellectual view of humanity

Talk to an advanced AI system for long enough and something becomes noticeable.

Its model of a competent person looks suspiciously like the worldview of a highly educated, verbally sophisticated knowledge worker.

Understand the evidence.

Clarify your assumptions.

Break the problem into pieces.

Read more.

Compare alternatives.

Write clearly.

Learn continuously.

Update your beliefs.

Think probabilistically.

These are good habits. But they are not a neutral sample of human competence.

They partly reflect the material from which AI learned.

The most useful text available for training is disproportionately produced by people who successfully convert thought into written symbols: textbooks, academic papers, technical documentation, journalism, legal documents, software discussions, manuals, essays, books and expert explanations.

A brilliant mechanic may possess extraordinary tacit knowledge but leave behind little text.

A politically gifted executive may know exactly which three people need to be convinced before a project can move, yet never document how they knew.

A great salesperson can walk into a room, identify the real decision-maker, sense that the stated objection isn't the real objection, and restructure the conversation within thirty seconds.

A nightclub promoter may understand human status dynamics better than an organizational-behavior professor while being unable to explain any of it formally.

The written record disproportionately captures what can be articulated.

AI therefore inherits an unusually articulate civilization.

And the people building advanced AI systems are themselves heavily selected from technical and intellectual environments. Researchers, engineers and product builders naturally construct tools around problems they understand.

The result is subtle.

AI doesn't merely automate some intellectual labor.

It also tends to describe life from inside the intellectual worldview.

The missing curriculum

School teaches us surprisingly little about another form of intelligence that governs much of adult life.

Call it social strategy.

Not simply "social skills."

Social strategy includes understanding incentives, building coalitions, maintaining relationships, exchanging favors, managing reputation, identifying power, negotiating status, generating trust and knowing when to spend political capital.

A socially strategic person walks into an organization and asks questions that rarely appear on an exam:

Who actually makes this decision?

Who can quietly block it?

Who needs to feel ownership?

Who trusts whom?

Who feels threatened?

What does this person want that costs me relatively little to provide?

Who owes me a favor?

Whose status increases if this succeeds?

Who must be consulted even though they have no formal authority?

What relationship should I invest in today because it may matter three years from now?

Politicians are often exceptionally good at this.

So are certain salespeople, founders, executives, fundraisers, Hollywood producers, diplomats and organizational operators.

Their intelligence isn't necessarily expressed through producing the best written answer.

It is expressed through causing a network of humans to move.

Formal education has difficulty teaching this because it is difficult to standardize.

Consider two exams.

The first asks:

"What caused the First World War?"

The second asks:

"Alice formally controls the budget. Bob doesn't control the budget but has the CEO's confidence. Carol controls implementation and believes your project threatens her team. David owes you a favor. You have six weeks to get the project approved without creating an enemy. What do you do?"

We have spent more than a century getting extraordinarily good at evaluating the first kind of problem.

The second may matter more once you are actually running something.

Yet there is no SAT for it.

We teach it accidentally

People do learn social strategy.

We just usually call the places where they learn it something else.

Sports teach hierarchy, coalition, competition, morale and leadership.

Student government teaches constituency management.

Sales teaches rejection, persuasion and incentive detection.

Fundraising teaches credibility and relationship maintenance.

Theater teaches presence and audience perception.

Organizing events teaches one of adulthood's most important lessons: having a logically correct plan does not cause anybody to help you.

Running a company teaches that organizations are not spreadsheets populated by rational agents.

They are repeated games between humans.

Even traditional elite institutions have historically understood this better than their formal curricula suggest. The importance of fraternities, dining clubs, alumni networks, sports teams, family connections and social rituals wasn't merely recreational.

Those environments trained and transmitted social capital.

A person could learn how powerful people behave around one another long before having much power themselves.

Modern meritocracy publicly emphasized grades while quietly continuing to reward networks.

The AI U-turn

AI could produce an interesting reversal.

Imagine human economic value as a collection of capabilities.

For much of the last century, one especially scarce capability was the ability to manipulate complicated information.

Take a difficult problem.

Research it.

Understand it.

Write the analysis.

Produce the spreadsheet.

Write the software.

Draft the contract.

Create the presentation.

Remember the relevant facts.

A highly capable knowledge worker could do things an average person simply could not do.

Now give both people an AI system.

The difference does not disappear. A highly intelligent person can usually direct the system better, recognize errors faster and ask more sophisticated questions.

But the absolute scarcity of competent intellectual production falls dramatically.

A competent memo that once required a lawyer can increasingly be drafted by software.

An analysis that required an analyst can increasingly be produced interactively.

A programmer can generate code far faster.

Someone without specialized training can ask sophisticated questions about subjects that previously required years of accumulated knowledge just to enter.

This doesn't make intelligence worthless.

It changes what intelligence is competing against.

When everyone has increasingly good access to cognition-on-demand, the bottleneck moves.

The question becomes less:

Can you produce a competent answer?

And more:

What should we do?

Can you tell when the answer is wrong?

Will anybody trust you?

Can you persuade people to act?

Can you recruit unusually good people?

Can you convince someone to give you money?

Can you build a coalition?

Can you maintain relationships for ten years?

Can you choose a direction under uncertainty?

Can you make people want to work with you?

Those abilities were always valuable.

AI makes them relatively scarcer.

This may increase the value of traits intellectual culture finds uncomfortable

Consider extraversion.

For decades, knowledge work created increasingly valuable environments where a person could succeed despite spending enormous amounts of time alone.

Programming may be the canonical example.

The computer does not care whether you are charismatic.

A compiler doesn't care whether people enjoy having dinner with you.

If you produce excellent software, the output can speak for itself.

But once AI produces more of the output, value may migrate toward the parts of the job that still require humans.

Talking to customers.

Building a team.

Creating enthusiasm.

Finding collaborators.

Negotiating.

Selling.

Maintaining a network.

Reading a room.

That doesn't mean introverts lose. Plenty of introverts are socially sophisticated.

It means the relative economic penalty for weak interpersonal ability may increase.

Physical attractiveness belongs somewhere in this discussion too, although people understandably dislike talking about it.

Attractiveness isn't competence. It doesn't make someone trustworthy, intelligent or strategically skilled.

But decades of psychological research suggest that appearance affects first impressions and social treatment.

An attractive person can receive a small subsidy at the beginning of many human interactions: more attention, easier initial acceptance, greater willingness from others to engage.

Over thousands of interactions, even a modest advantage can compound.

More invitations produce more social experience.

More social experience produces better social calibration.

Better calibration creates more invitations.

In an economy where human interaction, presence, trust and persuasion become relatively more important, the value of that initial subsidy may rise rather than fall.

The same is true of charisma, voice, physical presence, humor and confidence.

AI cannot make another human enjoy being around you.

At least not yet.

The return of political skill

The largest neglected category may be political ability.

"Political" sounds dirty because we associate it with manipulation.

But politics exists anywhere multiple humans have partially conflicting interests.

A startup has politics.

A family has politics.

A university department has politics.

A film production has politics.

An open-source community has politics.

Politics begins whenever you cannot simply issue a command and expect compliance.

The politically skilled person understands repeated games.

They know that helping someone today may create a relationship rather than an immediate transaction.

They understand that publicly defeating an opponent may be worse than allowing them to save face.

They know when to take credit and when giving credit creates more value.

They understand that nominal authority and actual influence frequently diverge.

They can keep multiple relationships alive even when those relationships have conflicting interests.

The intellectual worldview tends to underestimate this because political knowledge is hard to formalize.

A technically brilliant person can look at an organization and ask:

"Why don't they simply do the optimal thing?"

The politically sophisticated person asks:

"Optimal for whom?"

That may become a more important question in the AI economy.

The strange reversal

This creates a historical irony.

The industrial and knowledge economies gradually constructed institutions capable of identifying people who were unusually good at formal cognition.

A child with high academic ability could be detected, accelerated, credentialed and routed into a profession where that ability produced enormous returns.

We became very good at this.

Now we are building machines largely out of the artifacts created by those same people.

The machine absorbs their books.

Their papers.

Their code.

Their documentation.

Their reasoning patterns.

Their vocabulary.

Their worldview.

And by reproducing some of the abilities that made this cognitive class unusually economically valuable, AI may reduce the scarcity premium attached to those abilities.

The 20th century moved economic value toward what standardized institutions could measure.

The 21st century may partially move it back toward what those institutions never learned to measure well.

Judgment.

Taste.

Courage.

Trust.

Charisma.

Attractiveness.

Leadership.

Negotiation.

Coalition building.

Narrative.

Reputation.

Agency.

Political skill.

The ability to make another human care.

This is not the death of intelligence.

Intelligence will remain enormously useful, particularly when paired with AI.

But intelligence may increasingly become infrastructure rather than differentiation.

A century ago, literacy itself distinguished people. Eventually literacy became expected.

Perhaps some forms of knowledge work are heading toward the same fate.

The interesting question is what becomes scarce afterward.

And we may discover that many of the most valuable abilities were sitting outside the classroom the whole time.

When we talk about new media, we usually ask the wrong question.

Will television kill newspapers?

Will the internet kill television?

Will AI kill Hollywood?

History suggests that institutions rarely disappear so neatly. Newspapers survived radio. Radio survived television. Universities survived mass culture. Television survived the internet.

The more important change happens somewhere else:

the geographic scale at which power is organized changes.

A medium can survive while losing the functions that once made it central.

That is what happened to American cities during the rise of national radio and television. And it may be happening to countries today as digital media becomes genuinely global.

The interesting historical sequence may be:

city → nation → globe

with each expansion producing a new round of consolidation.

Before national media, cities were much more complete worlds

It is difficult today to appreciate how culturally autonomous major cities once were.

St. Louis was culturally St. Louis.

Milwaukee was culturally Milwaukee.

Boston, Chicago, Philadelphia and New Orleans each had stronger independent information environments than their equivalents do today.

This was not simply a matter of regional accents or cuisine.

The differences had an economic infrastructure behind them.

A large city could support multiple competing newspapers, publishers, theaters, advertisers, political factions, department stores, local business elites and intellectual circles.

The newspaper was especially important because its economics were local.

A serious newspaper required expensive presses, reporters, editors, paper, distribution networks and enough advertisers and readers to pay for all of it.

That meant the natural market was usually a city or region.

By 1923, more than 500 American cities still had competing daily newspapers.

That is a remarkably decentralized information system compared with the America that emerged later.

A resident of St. Louis did not merely receive a slightly different version of the same national feed.

St. Louis had institutions whose job was to decide what St. Louis should know.

The telegraph globalized information before it globalized audiences

This makes the telegraph era especially interesting.

By the second half of the nineteenth century, information could already travel across continents almost instantaneously.

London bankers could communicate with New York.

Governments could communicate with embassies.

Reuters and other news agencies could move information through international cable systems.

In that sense, a global information network existed more than a century before the Web.

But it was a thin network.

The architecture looked roughly like:

global telegraph network → local newspaper/institution → local population

The telegraph moved information between elite nodes.

It did not directly deliver a rich stream of information into every household.

So the nineteenth century could simultaneously have:

global finance and diplomatic communication

while maintaining

highly local public cultures.

The global layer provided wholesale information.

Cities still controlled retail distribution.

Radio and television changed the geographic unit

Radio broke this architecture.

For the first time, a performance or message produced in one city could reach millions of ordinary people simultaneously.

A performer no longer needed to travel from city to city.

A national broadcaster could distribute the same program everywhere.

Television intensified this by adding images.

The economics changed dramatically.

Suppose St. Louis can spend $20,000 producing a local show for 500,000 viewers.

A national network can spend $1 million producing something for 40 million viewers.

The national production can be vastly more expensive and polished while still costing less per viewer.

At that point, local cultural preference has to be very strong to justify duplicating the expensive production infrastructure in every city.

The result was not that local institutions instantly disappeared.

Instead, they became nodes inside a national system.

A Cleveland television station still existed.

It still sold local advertising.

It still produced local news.

But much of the expensive programming came from New York or Los Angeles.

The local institution survived while losing upstream functions.

That distinction matters.

By around 1960, the winners were largely visible

The most intense geographic consolidation happened roughly between the 1920s and 1960s.

New York captured national advertising, publishing, network headquarters, finance and much of national news.

Los Angeles captured filmed entertainment.

Washington captured political information as the federal government itself became more important.

Chicago remained enormous.

St. Louis remained a real city.

Boston remained prestigious.

But the question of which cities controlled the national information system was increasingly settled.

That is why I would divide the twentieth-century media era into two phases.

1920–1960: consolidation

The geographic unit expands from city/region to nation.

Local competition declines.

National networks form.

Advertising becomes national.

Corporations operate nationally.

People move across the country in much greater numbers.

A handful of cities capture disproportionate command functions.

1960–2000: maturity

The national system deepens rather than fundamentally changing geographic scale.

CBS, NBC and ABC dominate television for decades.

Hollywood becomes structurally entrenched.

New York remains the center of advertising, publishing and finance.

Washington remains the political information capital.

Cable later fragments the audience, but mostly within the same American national market.

This is why 2000 makes more sense to me as the endpoint of the national mass-media regime than 1980.

In 1980, America was still overwhelmingly consuming media through nationally organized institutions.

Around 2000, the distribution architecture itself started escaping the nation.

The developing-world megacity boom belongs to the same national era

This also changes how we should understand the rise of cities such as:

São Paulo
Mexico City
Istanbul
Delhi
Jakarta
Lagos.

Their twentieth-century rise was not simply “cities getting big.”

It was fundamentally part of national development.

The sequence looked something like:

rural population
→ industrialization
→ migration into the dominant national city
→ national corporations
→ national advertising
→ national television
→ national capital and talent concentration.

São Paulo became extraordinarily powerful because Brazil was becoming an integrated national economic system.

Mexico City benefited from the same process in Mexico.

Jakarta concentrated Indonesian economic functions.

Delhi and Mumbai accumulated Indian functions.

This was the developing world's version of the same geographic transition the United States had undergone earlier.

But that creates an important possibility.

Their rise may belong to the national-media equilibrium, not necessarily the global one that follows it.

A city can keep gaining population while losing relative command importance.

Jakarta could add millions of people while its best founders, financiers, scientists and producers increasingly operate through global hubs.

Population and command power are not the same variable.

Population growth creates cities. Integration chooses command cities.

Tokyo demonstrates another side of this.

Japan is shrinking and aging.

Yet Tokyo can continue attracting people from the rest of Japan.

That means an aging society does not necessarily produce immediate decline in its dominant city.

It can produce stronger concentration.

A shrinking country can still have a growing apex because that apex is consuming talent and population from everywhere underneath it.

The same mechanism can operate globally.

The next stage does not require hundreds of millions of people to relocate internationally.

It may only require the most valuable tail of the distribution to move.

If the best fraction of:

founders
scientists
financiers
artists
engineers
executives
producers

can participate in one international labor and capital market, relatively small migration flows can produce enormous concentration of command power.

Jakarta can gain ten million ordinary residents while losing 50,000 unusually globally mobile people.

For many high-value industries, the second number may matter more.

The internet may be doing to nations what television did to cities

This gives us a useful historical periodization:

The dates are obviously approximate.

History does not obey eighty-year cycles.

But the structure is worth taking seriously.

Around 2000, something qualitatively changed:

distribution stopped naturally ending at the national border.

The internet was initially weak compared with television.

Then broadband, smartphones, social media, streaming and global platforms accumulated.

Now AI translation attacks language barriers.

Digital payments attack monetary friction.

Remote work attacks labor-market geography.

Future faster aviation may attack physical distance among globally mobile elites.

Each one makes national duplication harder to justify.

Cultural difference may not protect national systems as much as we assume

This is where historical analogies become uncomfortable.

People often assume French media will remain French because France is culturally distinctive.

Japanese media will remain Japanese because Japan is culturally distinctive.

Canadian media will remain Canadian because Canada is different from the United States.

But St. Louis was distinctive.

Texas was distinctive.

New England was distinctive.

Milwaukee's German-American culture was extremely distinctive.

Those identities survived national consolidation.

What disappeared was the economic requirement that each place maintain a complete independent hierarchy of cultural production and command.

Texas could remain Texan while consuming nationally produced television.

Boston could remain Boston while New York dominated national publishing and advertising.

Cultural identity survived underneath the larger system.

So the proper question is not:

Is France different from America?

It obviously is.

The question is:

Is France different enough that it economically requires an entirely separate full-stack media, technology, financing and talent system?

Those are very different claims.

The next consolidation could be much harsher than today's global-city lists suggest

If the analogy holds, today's debate over London, Paris, Singapore, Toronto, São Paulo, Mumbai or Amsterdam may eventually look like debates over which American regional city would dominate the twentieth century.

Many of them can remain:

rich
large
beautiful
culturally important
regionally dominant.

That does not make them apex command centers.

Chicago remained one of America's great cities.

It simply stopped competing with New York for certain national command functions.

London could eventually occupy a similar position in a fully integrated Atlantic economy: immensely prestigious and important, but increasingly downstream from the principal global command node.

The same could happen to much of Europe.

Its many important cities partly reflect centuries in which separate countries needed separate capital, banking, media and professional systems.

If those national systems become layers inside a global system, many of those duplicated command functions become economically redundant.

New York may therefore be stronger than raw population forecasts suggest

The safest mistake would be to assume the places leading today's newest industry will necessarily become the final winners.

That would have made Detroit look unbeatable during the rise of automobiles.

It would have made Chicago look structurally dominant during industrial America's rise.

San Francisco may be the world's dominant technology cluster today.

That does not guarantee it becomes the dominant city of the mature global digital regime.

Technology itself may eventually commoditize some of the scarce engineering and startup capabilities that currently sustain the Bay Area's advantage.

New York's specialization is more abstract.

It specializes in:

coordination.

Capital allocation.

Corporate control.

Finance.

Law.

Media.

Advertising.

Deals.

Status.

International talent.

Those functions become more—not less—valuable when many national systems are integrated into one larger market.

And if talented Europeans, Canadians, Indians and others increasingly enter a common English-speaking global elite labor market, New York's talent pool is no longer primarily American.

Europe can feed New York in the same way America's regions once fed New York.

That produces the familiar agglomeration loop:

talent moves to the strongest network
→ the network gets stronger
→ capital follows
→ firms follow capital
→ more talent moves.

China is the major exception because it built a wall

This framework also explains why China is different.

China's potential to maintain a separate command system does not come simply from having 1.4 billion people.

India also has roughly that scale.

The crucial difference is that China has deliberately prevented its information economy from being completely absorbed into the American-led digital system.

It maintains:

domestic platforms
capital controls
media censorship
political barriers
technology policy
military independence.

China therefore possesses a real structural barrier to consolidation.

India mostly does not.

An Indian founder can operate in English.

An Indian creator can use American platforms.

An Indian financier can participate in New York markets.

India becoming richer may therefore strengthen American/global command hubs rather than automatically producing an independent Indian apex.

Large population alone is insufficient.

There has to be some mechanism preventing absorption.

The really strange historical possibility

The most interesting implication is that the twentieth-century nation-centered media system may itself have been temporary.

For a few generations it seemed natural that every important country should possess:

national newspapers
national television networks
national movie stars
national advertising agencies
national entertainment industries
national business elites.

But perhaps this was simply the equilibrium created by radio and television.

Before it, major cities supported more independent information systems.

After it, global digital networks may support fewer independent national systems.

Seen this way, the twentieth century becomes one stage in a much longer geographic expansion:

religious/civilizational networks — enormous reach, extremely low bandwidth

printing — much richer information, more regional/national fragmentation

newspapers — highly detailed city publics

telegraph — global command communication returns, but mass audiences remain local

radio/television — rich mass communication expands to national scale

internet/digital — rich, instantaneous, personalized communication expands toward global scale.

The historically unusual thing about digital technology is not simply that it reaches the world.

Religions and telegraph networks already reached enormous territories.

Digital media combines:

global reach + audiovisual bandwidth + instantaneous updates + personalization + measurement + two-way participation.

Earlier systems usually had to trade geographic reach against informational resolution.

Digital systems increasingly do not.

We may currently be closer to 1935 than 1975

That is perhaps the most useful way to think about the present.

By 1960, the winners of America's national-media consolidation were already relatively obvious.

The next forty years largely deepened the system those winners controlled.

If 2000 marks the beginning of a comparable global transition, then today we are only about twenty-five years into it.

We should not necessarily expect the geographic hierarchy to be settled yet.

Today's prominent cities may be equivalent to the strong American regional centers that still looked formidable during the early years of national consolidation.

The interesting period may be roughly 2000–2040 or 2050, when the sorting happens.

After that, a mature global regime could become surprisingly stable.

The eventual result might not be hundreds of equivalent national media capitals.

It could be a tiny number of genuine global command nodes, with large national and regional cities operating beneath them.

The important lesson from the last transition is that the losers do not need to disappear.

St. Louis still exists.

Chicago still matters.

Boston is still prestigious.

Their cultures survived.

What changed was the hierarchy.

And that may be exactly what is beginning to happen to countries today.

Population growth creates cities.
Integration chooses command cities.
And every expansion in the geographic scale of media forces the hierarchy to be chosen again.

We usually tell economic history as a conflict between capitalists and workers.

That misses something important.

Capitalists are rarely one coherent class.

At the beginning of a major industrial transformation, there are usually at least two competing forms of capital.

There is old capital: land, inherited portfolios, protected markets, mature monopolies, licenses, financial claims and institutions whose value depends on preserving the existing economic order.

And there is new productive capital: railroads, steel mills, automobile factories, semiconductors, software, AI systems, robots, energy infrastructure—whatever production system is trying to become the next economic center.

The conflict between them is structural.

The new industrialist does not merely need customers.

He needs:

land
workers
energy
credit
infrastructure
favorable taxation
new education systems
different trade rules
sometimes entirely new property rights.

Those resources are not sitting unused.

Someone already controls them.

And that person often becomes richer by preventing the new system from reorganizing them.

That is why industrial revolutions so often become political struggles.

The old elite rarely says:

We are rentiers trying to preserve our declining economic share.

The new industrialist rarely says:

We would like to reduce the value of your assets and take control of the state.

They instead fight through political language:

liberty
tradition
free trade
protectionism
equality
property rights
national sovereignty
modernization
environmental protection
innovation.

These ideas can be sincerely believed.

But underneath them sits a harder question:

Which assets, industries and classes will the state organize society around?

And the middle class—farmers, engineers, professionals, skilled workers, homeowners, managers—often provides the numbers that decide the answer.

America’s Civil War was an extreme version of this conflict

The American Civil War was fundamentally about slavery. Nothing about the economic interpretation should obscure that.

But slavery was not merely a moral institution.

It was also the foundation of an enormously powerful property-owning class.

The antebellum South concentrated wealth in:

plantation land
enslaved people
agricultural exports
the political power required to protect those assets.

The emerging Northern economy was increasingly organized around something different:

manufacturing
railroads
free farms
wage labor
commercial cities
immigration
infrastructure.

These were not simply two regions with different lifestyles.

They represented competing models for the future American economy.

The Republican coalition of 1860 gives us a remarkable window into this.

It supported:

restricting slavery's expansion
protective tariffs for American industry
free western homesteads
internal improvements
a transcontinental railroad.

Put those policies together and a coherent political economy emerges:

land for independent farmers

• labor for an expanding free economy
• railways connecting a continental market
• protection for domestic manufacturers
• federal infrastructure supporting industrial growth.

Southern plantation interests did not necessarily benefit from that order.

A cotton exporter had good reasons to prefer cheap imported manufactured goods.

The Northern manufacturer wanted tariffs.

The planter wanted the political importance of plantation agriculture preserved.

The emerging industrial economy wanted western settlement, infrastructure and free labor to expand.

The conflict was ultimately about more than whether a particular machine would be adopted.

It was about:

Which economic system would control the continent?

The Southern planter class possessed enough political power to obstruct parts of the Northern developmental program.

Then the South seceded.

Its congressional veto disappeared.

The Homestead Act passed.

The Pacific railroad moved forward.

And after Union victory, slavery itself ceased to exist as a legal property system.

That was elite displacement in its most violent possible form.

The plantation elite did not voluntarily diversify into railway stocks and congratulate the new industrial order.

Its central asset was abolished.

Its political system was defeated militarily.

The United States that emerged after the war was increasingly governed by the developmental requirements of the Northern industrial economy.

Then the winners became the next old money

By 1900, America's emerging industrialists were no longer emerging.

Railroad owners, steel magnates, oil capitalists, bankers and utility interests had created colossal fortunes.

The new productive class had become the incumbent class.

This is where the process becomes cyclical.

The entrepreneur begins by demanding:

Open the system.

After winning, he increasingly demands:

Protect the system in which I won.

The late nineteenth-century industrialists had needed competition against older land and mercantile interests.

But once railways, oil, finance and industrial trusts became dominant, their owners increasingly benefited from:

concentration
barriers to entry
control of infrastructure
ownership pyramids
privileged access to finance.

Now another coalition began forming against them.

Farmers hated railroad pricing power.

Workers organized against industrial employers.

Smaller businesses opposed trusts.

Professional managers increasingly ran organizations that family owners had created.

Progressive reformers built new administrative institutions.

Eventually the federal government gained:

stronger antitrust powers
an income tax
securities regulation
estate taxation
utility regulation
labor law.

Then the Depression and the New Deal accelerated the transformation.

Again, the important point is not:

Capitalism disappeared.

The opposite happened.

American industrial corporations became some of the most powerful productive organizations in human history.

But the old proprietor's claim over those organizations weakened.

A company such as General Motors could become vastly larger while ownership became more dispersed, professional managers acquired more control, workers captured more income through unions and government taxed a larger part of the surplus.

The productive organization survived.

The ruling configuration around it changed.

That distinction matters enormously.

Displacing an old capitalist class does not require destroying the factories it owns.

Sometimes the new regime needs those factories desperately.

What changes is who controls the surplus and what the productive system is allowed to become.

Germany shows the process even more brutally

Imperial Germany initially complicates this story.

Germany industrialized extraordinarily quickly without first eliminating its old landed aristocracy.

The Prussian Junkers survived.

At the same time, a powerful industrial capitalist class emerged around:

steel
chemicals
electrical equipment
machinery
banking.

Instead of immediate displacement, Germany developed the famous “rye and iron” bargain:

agricultural landowners

• heavy industrialists.

The old landed elite received protection.

The new industrial elite received policies supporting industry.

This demonstrates an important limit to the argument.

The old elite does not have to disappear before industrialization can begin.

But something crucial had already changed:

industrial capital had become powerful enough that the landed elite could no longer rule alone.

The Junkers had to bargain with Krupp.

The new class had become an independent political center.

Then came thirty catastrophic years.

After WWI, Germany lost around 13% of its territory, including important eastern agricultural regions.

The aristocracy lost its privileged position under the monarchy.

Hyperinflation destroyed enormous quantities of financial wealth.

But the eastern landowning class still survived.

We know it survived because the Weimar state was still spending enormous amounts supporting distressed eastern estates during the late 1920s.

Then WWII finished the process.

Germany lost:

East Prussia
most of Silesia
most of Pomerania
other eastern territories.

For an aristocratic family whose fortune consisted of 6,000 hectares in East Prussia, this was not a tax increase.

The asset physically ceased to exist inside Germany.

And the remaining giant estates in the Soviet occupation zone were expropriated.

Roughly 3.1 million hectares entered land redistribution.

The old East-Elbian landed class was effectively removed from the economic structure of postwar Germany.

West Germany therefore began its new life with an unusual geography.

Its heartland consisted increasingly of:

the Ruhr
Rhineland commerce
southern manufacturers
Mittelstand companies
industrial workers
managers.

The classic Junker political economy was largely gone.

But productive industrial organizations remained reconstructable.

A Prussian estate lost to Poland could not be rebuilt in Frankfurt.

Siemens could.

BASF could.

Krupp could.

So postwar West Germany inherited an economy in which old rentier claims had been brutally stripped away while industrial knowledge remained.

That is a remarkably powerful form of institutional reset.

Qing China shows what happens when displacement never goes far enough

This is perhaps the more important comparison.

Qing China did not simply reject Western technology.

It modernized repeatedly.

It built:

arsenals
shipyards
modern military units
industrial enterprises
technical schools.

The Qing understood that European military and industrial technology mattered.

The problem was deeper.

The new productive system remained embedded inside the old political order.

Provincial authority remained fragmented.

Old bureaucratic structures controlled allocation.

Modern enterprises existed, but a new industrial capitalist class did not become sufficiently autonomous to reorganize the state around its needs.

Technology was absorbed.

The ruling system was not displaced.

Then compare Meiji Japan.

The Japanese did not merely purchase better artillery.

They:

abolished domains
eliminated samurai stipends
dismantled hereditary privileges
centralized taxation
reorganized the military
created national institutions
redirected capital toward industrial development.

Many former samurai survived the transition.

Some became:

officers
bureaucrats
teachers
managers.

But they survived by converting themselves into the new order.

The samurai system itself lost.

That distinction is critical.

Successful conversion often happens after displacement, not instead of it.

The old elite may survive personally.

Its old institutional privilege does not.

Ideology is how these coalitions explain themselves

These struggles are difficult to recognize because political coalitions do not describe themselves in balance-sheet language.

The Northern manufacturer did not campaign by saying:

We need to weaken plantation capital so manufacturing can dominate national accumulation.

He talked about:

free labor
union
national development.

The landowner did not say:

My asset value depends on restricting economic transformation.

He talked about:

property rights
tradition
constitutional limits.

The Progressive movement did not campaign under the slogan:

Let us transfer control from dynastic proprietors toward managers, workers and the administrative state.

It talked about:

fairness
democracy
monopoly power.

Later, globalization was defended through:

efficiency
openness
free trade
internationalism.

Its opponents talked about:

jobs
national industry
sovereignty.

The ideology matters.

But it also allows a much larger coalition to assemble around economic interests that would otherwise look narrow.

The middle class is often the decisive coalition partner

This is why the middle class should not be treated as one permanent political bloc.

A middle-class group asks:

Which economic regime makes my skills and assets more valuable?

A nineteenth-century independent farmer might side with Northern industrial capital because railroads increased land values and connected him to markets.

A skilled industrial worker might support industrial expansion because factories increased demand for his labor.

A twentieth-century engineer benefited from manufacturing corporations.

A programmer after 1990 benefited enormously from digital capital.

A homeowner in 2026 may instead benefit from preserving housing scarcity.

A doctor may benefit from credential restrictions.

An electrician may benefit from enormous AI datacenter construction.

A robotics technician may benefit from accelerated automation.

A lawyer might lose from the same process.

So “middle class” does not tell us enough.

The politically relevant question is:

Which part of the middle class is complementary to which form of capital?

That is how seemingly strange coalitions become understandable.

The 1980–2020 coalition was capital plus professional labor

The postwar corporate-industrial system eventually produced another transformation.

Global supply chains, telecommunications, financial integration and computers made capital increasingly independent of domestic factory labor.

A corporation could combine:

American capital
American management
American engineering and design
Asian or Mexican production
worldwide consumers.

This changed the domestic coalition.

The factory worker became less essential to American capital.

The:

programmer
banker
consultant
lawyer
manager
engineer

became more complementary.

For several decades, capital owners and educated professional workers often rose together.

The upper-middle professional class accumulated:

high salaries
valuable credentials
urban property
equities
institutional influence.

Meanwhile industrial labor weakened.

That arrangement is now itself becoming incumbent.

And AI may be beginning the next conflict.

AI capital does not need the same coalition

Today's emerging productive capital increasingly consists of:

AI models
semiconductors
datacenters
robotics
energy
advanced manufacturing.

It needs enormous amounts of:

electricity
land
transmission
cooling
chips
industrial construction.

Its domestic economic interests increasingly look less like the asset-light internet industry of 2010 and more like an earlier industrial capitalist class.

It wants:

build energy
build datacenters
build housing near productive regions
accelerate infrastructure
make new capital investment cheap
bring in exceptional technical talent
keep international markets open to digital exports.

And its complementary middle class may be changing.

It may include a relatively small elite of:

AI researchers
chip designers
systems engineers.

But the mass complementary group may increasingly consist of:

electricians
HVAC technicians
construction workers
robotics technicians
semiconductor workers
advanced manufacturing operators.

That is a potentially very different coalition from the one that dominated digital capitalism from 1990 to 2020.

And parts of today's upper-middle class may become the incumbent opposition

Consider the economic interests created by AI.

If machine intelligence dramatically lowers the cost of:

legal analysis
financial analysis
software development
education
medical diagnosis
administrative work,

then consumers gain.

AI capital gains.

But parts of the credentialed professional class lose scarcity rents.

Likewise housing.

New productive clusters need workers to live somewhere.

Productive capital benefits from:

abundant housing.

Existing metropolitan property owners may benefit from:

scarce housing.

Again, the conflict is not primarily ideological.

Their assets want different things.

An AI-era political struggle could therefore align:

new productive capital

• technical workers
• younger asset-poor people
• small entrepreneurs
• consumers

against some combination of:

property rentiers

• credential rents
• mature monopolies
• existing financial claims.

Whether that coalition actually forms is one of the major unresolved political questions of the next twenty years.

The United States appears to be testing displacement again

America currently has the strongest independent concentration of new digital/AI capital.

That matters politically as much as technologically.

Nvidia, frontier AI firms, robotics startups and the surrounding capital ecosystem are not merely divisions inside old banks, universities or industrial conglomerates.

They are becoming independent centers of wealth.

That gives them the potential to demand institutional changes.

The increasingly aggressive American debate around:

energy
permitting
datacenters
semiconductor production
industrial policy
AI exports

can be interpreted partly through this framework.

The new productive faction is beginning to tell the state:

The country's infrastructure must now be rebuilt around our requirements.

Whether it wins is not predetermined.

But the struggle has begun.

Europe faces a much harder conflict

Europe has spent two decades encouraging digital entrepreneurship.

But generating startups is not the same thing as generating an autonomous new capitalist class.

A mature European political economy already has enormous claims attached to:

pension systems
existing property
legacy industrial firms
banks
professional regulation
welfare commitments
safe household savings.

A truly aggressive new industrial strategy would require redirecting substantial resources toward:

AI
energy
semiconductors
risky venture investment
new infrastructure.

Something else would receive less.

That is the real tradeoff.

Europe can regulate AI.

It can subsidize AI.

It can build AI institutes.

But the harder question is:

Will Europe's emerging productive firms become powerful enough to impose losses on parts of the existing order?

If the answer is no, Europe may successfully use the next industrial technology while failing to produce the class that controls it.

That is the Qing problem in modern form:

modernization without displacement.

Japan may have already suffered half the revolution

Japan is particularly interesting.

Unlike Canada and some European countries, Japan allowed population decline and post-bubble asset destruction to severely weaken parts of its twentieth-century rentier economy.

Land values collapsed after the bubble.

Millions of homes became vacant.

Regional infrastructure became excessive relative to population.

Universities increasingly face shrinking student populations.

The traditional assumption that population growth would continuously validate existing property and institutional capacity disappeared.

In that sense, Japan has already paid part of the cost of de-rentierization.

But it has often handled new productive entrepreneurship through absorption:

startup
→ partnership with incumbent corporation
→ acquisition
→ integration into existing corporate structure.

That produces innovation.

It does not necessarily produce a new autonomous capitalist class.

Japan therefore presents a fascinating unresolved case:

old rentier structures have already taken enormous damage, but a sufficiently powerful replacement class has not yet clearly emerged.

Its next twenty years may look very different from its previous thirty.

China has several capitalist systems fighting underneath the Party-state

China may have the most complicated structure.

One enormous twentieth-century/early twenty-first-century coalition developed around:

local government
land sales
property developers
banks
household real estate.

That produced an enormous property-rent system.

Another powerful faction consists of:

exporters
factories
industrial supply chains
manufacturing capital.

Then a newer productive faction emerged around:

digital platforms
EVs
batteries
AI
robotics
advanced technology.

Unlike the United States, however, China has a political authority sitting above all of them.

The Party-state can decide:

property gets less credit
EV manufacturing gets more
a digital billionaire becomes too politically independent
semiconductors receive strategic capital.

China is therefore not allowing an uncontrolled class displacement.

It is attempting state-managed elite rotation.

That may be very powerful.

It may also become dangerous if political allocation repeatedly selects the wrong productive faction.

What happens when the rentier class wins?

The failure is usually not technological backwardness.

That is why it is hard to see.

The country may have:

excellent universities
cutting-edge laboratories
advanced machinery
startup accelerators
AI adoption
talented engineers.

But underneath that modern appearance:

savings continue flowing toward existing property
credit continues favoring incumbent firms
new companies get purchased before becoming politically independent
credential barriers remain intact
land remains protected
tax policy preserves old claims
infrastructure serves the previous economy.

The new technology is present.

The old political economy remains sovereign.

That country becomes a consumer of the industrial revolution rather than its center.

It can remain prosperous for a long time.

Venice remained extraordinary after losing its commercial supremacy.

Rome remained culturally important after losing imperial power.

A country does not become poor overnight simply because it stops controlling the productive frontier.

But over decades, the hierarchy changes.

The places controlling the new production system accumulate:

capital
technological capability
high-value firms
military capacity
political influence.

The former center increasingly accumulates:

historic assets
financial claims
consumption
tourism
preservation.

The difference compounds.

Industrial revolutions therefore require losers

This is the part political rhetoric usually avoids.

We like technological progress to be a story in which:

entrepreneurs get richer
workers get richer
homeowners get richer
pensioners get richer
old companies remain valuable
new companies become valuable
government collects more revenue.

History rarely works that cleanly during fundamental transitions.

A new productive regime reorganizes scarcity.

If housing becomes abundant, some property scarcity disappears.

If intelligence becomes abundant, some professional scarcity disappears.

If new companies capture markets, incumbents lose them.

If savings move into frontier investment, fewer resources remain for old assets.

If political power moves toward one coalition, another loses influence.

Industrial revolutions are therefore not merely periods when new things become valuable.

They are periods when old things become relatively less valuable.

And the people who own those old things fight back.

That is the real politics of AI

The most important AI-policy questions may therefore have little to do with model benchmarks.

Watch instead:

Who gets the electricity?

Who gets the land?

Which industries receive cheap capital?

Which professions lose licensing protection?

Which assets receive tax protection?

Which workers become complementary to the new capital?

Which startups remain independent?

Which ones are absorbed?

Which incumbent class is willing to accept lower relative wealth?

Those are the actual fronts of the industrial struggle.

Because the defining moment of an industrial revolution is not when somebody invents the new machine.

It comes later.

It is when the owners of the new productive system become strong enough to say:

The economy will now be organized around our needs.

And the owners of the previous system realize:

If they win, we lose.

That is when technological change becomes class struggle.

And that is when an industrial revolution truly begins.

In my last piece, After the Frontier, I looked at what tends to happen once an expansionary civilization reaches the end of its geographic frontier.

The basic pattern was:

territorial expansion → consolidation → prosperity → wealthy stagnation → relative decline → collapse

But while putting together the historical timelines, France kept bothering me.

France did not seem to have one clean frontier cycle.

It had several.

France built a major empire in North America, then lost most of it in 1763.

A few decades later, it was expanding across Europe under Napoleon.

That territorial system collapsed in 1815.

Yet only fifteen years later France invaded Algeria, beginning another century of expansion across Africa and Asia.

Then that empire disappeared too.

If losing an empire marks the end of an expansionary civilization, France should have been finished more than once.

It wasn't.

That made me wonder whether I had been looking at the wrong variable.

Perhaps the decisive question is not simply:

How much territory does a country still control?

It is:

What survives after the territory is lost?

In France's case, one thing survived almost every reset:

Paris.

The colonial map changed repeatedly. Political regimes came and went. France suffered military defeats, revolutions and territorial contractions.

But Paris remained one of the world's major concentrations of capital, administration, science, education, finance and culture.

That suggests a second hypothesis to add to the frontier cycle:

A civilization can lose its territorial perimeter and still generate another outward wave if the productive and administrative core that created the first one remains globally competitive.

France may be one of the clearest historical cases of that happening repeatedly.

France as a Repeated Experiment

France is useful because its history does not give us one clean imperial cycle.

It gives us several.

The first two resets are particularly striking.

The Treaty of Paris in 1763 forced France to give up its mainland North American territories. French Canada was gone, and Britain's victories had also badly reduced French ambitions in India.

If imperial territory itself were the main source of French power, this should have been devastating for generations.

Instead, within thirty years, France was producing the largest territorial upheaval Europe had experienced in centuries.

The particular political regime had changed completely.

The geographic direction of expansion had changed.

But the productive-demographic-administrative center generating that power was still there.

Then came 1815.

The Napoleonic territorial system collapsed. France was pushed back to a much smaller European core.

Again, one might imagine the expansionary era was finished.

Yet in 1830, only fifteen years later, France seized Algiers, beginning the second major French colonial expansion. Cambridge's history of French colonialism explicitly identifies the invasion of Algeria in 1830 as the beginning of this second nineteenth-century imperial wave.

By the end of the nineteenth century, the direction had changed completely:

not Europe, but Africa and Asia.

France acquired Algeria, expanded through West and Central Africa, took Tunisia, built its Indochinese empire and established possessions elsewhere.

By 1900 it had firm control over roughly twenty newly acquired colonies.

So the recurring structure was:

territorial system collapses

→ France contracts toward its core

→ core survives

→ France accumulates again

→ another outward wave appears somewhere else.

The Constant Was Not the Frontier. It Was Paris.

This is the part I find most interesting.

The frontier kept moving.

The origin point didn't.

French Canada could disappear.

Napoleonic Europe could disappear.

Algeria and Indochina could disappear.

But for centuries Paris remained one of the world's principal concentrations of:

government
finance
science
education
culture
elite manpower
administrative capacity
infrastructure

France was also extraordinarily centralized.

French science in the nineteenth century was characterized not just by strong state administration but by the geographic concentration of resources, personnel and influence in Paris.

And Paris remained internationally competitive even after major French defeats.

Around 1840, London had become the leading financial center in the world.

But the next city was Paris.

Around 1875, after decades of industrialization elsewhere and just four years after France's defeat by Germany, the hierarchy was still:

London first, Paris second.

That ranking persisted through immense political instability:

monarchy,

revolution,

republic,

empire,

another revolution,

another empire,

military defeat.

The government kept changing.

The metropolitan capital remained.

This matters because an empire is not produced by square kilometers.

It is produced by a system capable of mobilizing:

surplus capital + administrators + engineers + military capacity + information + logistics.

The outer territory is partly the result of that system.

It is not necessarily the system itself.

The Core Hypothesis

So I would distinguish between two things:

The imperial perimeter

This is what appears on the map.

It can disappear quickly.

The productive-administrative core

This is the machinery capable of generating another perimeter.

It includes:

• accumulated physical capital;
• financial capital;
• sophisticated state administration;
• universities and scientific institutions;
• engineering capability;
• educated human capital;
• transportation and communications systems;
• military-industrial capability;
• commercial networks;
• institutional prestige.

The hypothesis is:

Regenerative potential∝relative strength of the surviving core\boxed{ \text{Regenerative potential} \propto \text{relative strength of the surviving core} }

But the important word is relative.

A city does not need to become poorer for its expansionary capacity to fall.

Its competitors merely need to become stronger faster.

A more complete version might be:

A strong metropolitan core can generate another outward wave when it has enough surplus and administrative capacity, and when there is some frontier whose value exceeds the cost of reaching and controlling it.

This is not intended as a literal econometric equation.

It is a way of clarifying the variables.

Why Relative Strength Matters More Than Absolute Wealth

Consider France around 1789.

France had roughly 28 million inhabitants and was the most populous European country.

That meant that even after losing Canada, the metropolitan core remained enormous relative to the competing states around it.

France had lost territory.

It had not lost the underlying demographic and productive scale capable of raising huge armies and administering a great state.

This helps explain why territorial defeat in 1763 did not permanently lower France into the category of a secondary country.

But the nineteenth century gradually changes the equation.

France keeps becoming richer.

Paris keeps accumulating capital.

Yet:

Britain industrializes faster.

Then:

Germany unifies and industrializes.

Then:

the United States becomes continental-scale.

Russia also grows.

France is therefore perfectly capable of another imperial expansion while simultaneously experiencing relative decline against stronger emerging cores.

That distinction is essential.

A country can regenerate without recovering its previous rank.\boxed{ \text{A country can regenerate without recovering its previous rank.} }

France after 1815 came back.

France after 1871 came back.

But each comeback occurred in a world containing stronger competitors.

The ceiling kept moving.

Strong Neighbors Did Not Prevent French Expansion

This also corrects another simplistic version of the model.

A productive core does not necessarily expand geographically into the territory immediately beside it.

France almost always had formidable neighboring powers.

Depending on the period:

Britain
Spain
Austria
Prussia/Germany
Russia

were capable of imposing enormous costs on French expansion.

Yet France still reached:

North America
the Caribbean
Algeria
West Africa
Central Africa
Madagascar
Indochina
the Pacific.

Why?

Because a frontier does not need to be contiguous.

A maritime power can skip expensive neighbors and search for cheaper opportunities thousands of kilometers away.

Our earlier formulation captured this better:

A globally competitive productive core does not require a contiguous frontier. It requires some frontier whose expected return exceeds the cost of projecting power to it.

Germany sitting next to France did not make Senegal farther away in the economically relevant sense.

If Germany was extremely expensive to challenge while a distant territory contained valuable assets and much weaker resistance, the maritime frontier could be cheaper than the continental one.

So:

blocked European frontier

could produce:

overseas frontier.

That is exactly what nineteenth-century France did.

1871 Makes the Mechanism Especially Visible

France's defeat by Germany in 1870–71 is interesting because France subsequently becomes more territorially expansionary overseas.

The defeat did not destroy Paris.

It changed France's relative position inside Europe.

And after 1871, colonial ideology became much more prominent among French political elites. France accelerated its expansion overseas; by the end of the century it controlled a large new collection of territories in Africa and Asia.

This suggests a counterintuitive possibility:

Competition can redirect expansion rather than end it.

If the easiest European frontier becomes prohibitively expensive, surplus generated by the core can seek opportunities elsewhere.

The map changes direction.

The underlying expansionary capacity survives.

Then 1945 Changes the Rules

The last French contraction is different.

France emerged from World War II severely damaged.

Then the colonial perimeter collapsed:

Indochina
Morocco
Tunisia
Algeria
most of sub-Saharan Africa.

This time there was no third large territorial empire.

But this is where the distinction between capacity and outcome becomes important.

The absence of a territorial comeback does not prove that the metropolitan core ceased being productive.

Quite the opposite.

Postwar France experienced the Trente Glorieuses, roughly the quarter-century of extraordinarily rapid growth after WWII. Across Western Europe, 1947–73 was characterized by sustained productivity and consumption growth; France's own growth included exceptionally large productivity gains.

Paris survived another imperial contraction.

And it remains globally important today.

Oxford Economics' 2025 Global Cities Index describes Paris as the fifth-largest metropolitan economy in the world and second-largest in Europe.

That is remarkable persistence.

The French-controlled map became dramatically smaller.

But the metropolitan core remained near the global frontier.

What changed after 1945 was the international environment.

Direct territorial acquisition became vastly less institutionally acceptable.

Nuclear weapons raised the costs of conflict between major states.

The United States and Soviet Union dominated the security order.

Decolonization transformed political legitimacy.

France therefore expressed much of its remaining outward capacity through different mechanisms:

European institutions,

capital,

diplomacy,

military interventions,

cultural networks,

nuclear deterrence,

corporations,

development relationships.

So the hypothesis should not be:

Strong capitals always create new territorial empires.

It should be:

A strong surviving core preserves the capability for another outward wave; the form that wave takes depends on the opportunities and constraints of the international system.

Core Strength Is Necessary-ish, Not Sufficient

That distinction matters.

A globally competitive metropolitan core is probably not sufficient for territorial expansion.

You also need some combination of:

political cohesion
projection capacity
available opportunities
security pressure
military asymmetry
administrative competence
lack of effective rival interdiction

And the international rules must permit it, or weaken sufficiently that the benefits exceed the costs.

But the absence of a competitive core may be much more decisive.

If the empire disappears and the core itself becomes technologically obsolete, financially marginal or administratively incapable, it becomes very difficult to generate another wave.

That gives us a potential predictor:

The more useful question is not how much territory an old empire lost, but whether the metropolitan machine left behind remains globally competitive.

That was the central conclusion of our earlier comparison of Paris, Tokyo and the major American metropolitan cores.

The United States: What If the Perimeter Shrinks but the Core Doesn't?

This makes the United States interesting.

Suppose American global influence contracts substantially over the next century.

Fewer military commitments.

Less control over international institutions.

A smaller share of world GDP.

Asia becomes the principal center of world production.

Perhaps the dollar is less dominant.

None of that automatically destroys:

New York
Boston
Washington
the Bay Area
Los Angeles
Seattle
Texas

The United States does not even have a single Paris.

It has a network of metropolitan cores.

The United States also differs from France in an important way: it does not depend on a single dominant metropolitan core.

By nominal metropolitan GDP, its major cities remain extraordinarily large relative to the leading metropolitan economies of Latin America:

The figures come from a global compilation of metropolitan GDP estimates; years and metropolitan definitions differ somewhat by country, so the rankings should be read as order-of-magnitude comparisons rather than perfectly standardized statistics.

The important point is that the American equivalent of Paris is not one city. It is a network of globally large productive cores inside the same political system. Even the largest metropolitan economies in Latin America remain much smaller than several separate American centers.

That is an extraordinary concentration of independent productive centers inside one political system.

So imagine American global influence shrinking while the productive capacity of its major metropolitan centers remains largely intact.

to the same degree.

Historically, that does not look like the end of the story.

It looks more like France after one of its contractions.

The first decades might involve political trauma and arguments over lost status.

But if the productive cores remain near the world frontier, they continue generating:

capital,

technology,

military capability,

corporations,

migration pull,

scientific output,

cultural influence.

Eventually that surplus has to express itself somewhere.

Perhaps first inside North America.

Perhaps through the hemisphere.

Perhaps later through entirely different parts of the world.

The specific geography is impossible to forecast.

The structural point is simpler:

a shrinking perimeter does not imply a permanently shrinking radius.\boxed{ \text{a shrinking perimeter does not imply a permanently shrinking radius}. }

France demonstrated that repeatedly.

Japan May Be an Even Cleaner Experiment

Japan provides another remarkable case.

Its imperial perimeter disappeared almost instantaneously in 1945.

Korea was gone.

Taiwan was gone.

Its continental position in China was gone.

Its Pacific territorial system was gone.

Yet the productive geography inside Japan remained:

Tokyo–Nagoya–Osaka

and was rebuilt.

Tokyo remains one of the world's largest urban economies. Oxford Economics estimated it as the second-largest metropolitan economy in the world in its 2024 analysis; its 2025 index still places Tokyo among the world's top ten cities overall.

Japan then regenerated enormous external economic influence without rebuilding formal territorial administration.

Japanese manufacturing, finance, technology and infrastructure investment radiated outward through Asia.

Again:

perimeter destroyed

→ core rebuilt

→ outward influence returns.

The French example warns against assuming that any future outward expansion from Tokyo would have to proceed sequentially into Korea, China or Russia.

France did not have to conquer Germany before reaching Indochina.

Strong neighboring states can make the nearest frontier the least attractive frontier.

A maritime productive core can search globally.

Whether Japan ever again expresses external power territorially is impossible to predict.

Today's international order strongly discourages it.

But over a horizon of centuries, it would also be strange to declare the territorial form permanently extinct simply because the current institutional order is eighty years old.

The more defensible claim is:

Tokyo retains substantial regenerative capacity because the productive core survived the empire.

What political form that capacity eventually takes is contingent.

The Map Is the Output, Not the Machine

This may be the broader lesson from France.

We tend to look at historical maps and assume the map itself was the power.

France in 1750 looks powerful because its colors appear in North America.

France in 1812 looks powerful because French administration covers much of Europe.

France in 1930 looks powerful because enormous portions of Africa and Asia are painted French.

Then the colored areas disappear and we say:

France declined.

That is true in one sense.

But it misses the deeper continuity.

The machine sitting underneath those maps remained disproportionately concentrated around Paris.

When one territorial system disappeared, Paris did not need to rediscover administration, finance, education, science and military organization from zero.

It already possessed them.

It simply needed another opportunity.

So perhaps:

Empire = productive core + historically available frontier.

Take away the frontier and the empire contracts.

Leave the core intact and another frontier may eventually produce another wave.

Destroy or permanently marginalize the core, and the cycle becomes much harder to restart.

A Better Way to Think About Imperial Decline

This suggests three very different kinds of decline.

France experienced the first repeatedly.

It gradually experienced the second as Britain, Germany, the United States and others accumulated capital faster.

But it avoided catastrophic long-term destruction of the third.

That may explain its extraordinary persistence.

Paris remained sufficiently competitive that France repeatedly retained the capability to matter again.

The Hypothesis

So the proposition I would take from the French case is:

The best predictor of an old expansionary power's ability to generate another outward wave may not be how much territory it retains, but whether its surviving metropolitan productive-administrative core remains large and competent relative to the rest of the world.

Not absolute size.

Relative capability.

Not the old frontier.

The machine capable of finding another frontier.

France repeatedly lost the map while keeping Paris.

Japan lost its empire while keeping Tokyo.

A future United States could lose much of its global perimeter while retaining New York, Boston, California, Texas and its other productive cores.

None of this tells us exactly where another outward wave would go, when it would occur, or whether it would take territorial, financial, security or institutional form.

But France gives us a strong reason not to assume:

imperial contraction = permanent civilizational retreat.

Sometimes the outer shell disappears.

The core survives.

And after enough time, the core starts pushing outward again.

I started thinking about this after watching Jaadugar: A Witch in Mongolia.

What struck me was not the anime itself, but the world behind it.

Eight hundred years ago, the Mongols were not a peripheral people. They were one of the central forces of Eurasian history. Mongol armies and successor states stretched from China across Central Asia and into Persia, connecting enormous parts of the continent under a political and military system that contemporaries could hardly ignore.

Today, Mongolia is a small country between Russia and China, and the Mongol world is mostly encountered as history.

That contrast interested me.

How does a civilization go from being one of the great expansionary forces of its age to becoming peripheral to the world system?

From there I started looking at other civilizations that once seemed almost unstoppable: the Ottoman Turks, Portugal and Spain, Russia, and eventually Britain.

The pattern I became interested in was not simply “empires rise and fall.” That is too vague.

It was something more specific:

What happens after a civilization reaches the end of its geographic frontier?

Across several very different empires, a rough pattern appears:

frontier expansion → consolidation → prosperity → wealthy stagnation → relative decline → possible contraction or reinvention

The surprising part is that the end of expansion often does not immediately produce decline.

Frequently, the civilization becomes richest shortly before or after its frontier closes.

The decline comes later.

The Basic Cycle

The key mistake is to imagine that:

frontier closes → decline begins immediately.

Historically it often looks more like:

frontier closes → development boom → prosperity → long plateau → decline.

The golden age is often the harvest from the preceding expansion.

Historical Examples

The exact dates are debatable, but the broad comparison is more useful than pretending that every empire followed an identical calendar.

These dates should not be interpreted mechanically.

The more important observation is the shape.

Portugal, Spain, the Ottomans and Russia all eventually transitioned from:

adding new territory

to:

developing and maintaining an enormous inherited territorial stock.

For some time, this produced prosperity.

Eventually maintaining the accumulated system became more important than extending it.

Why Prosperity Often Comes After Expansion

Imagine a country that has just acquired enormous territory.

Initially much of it is economically underdeveloped.

It possesses:

• farmland that is not fully cultivated;
• mineral deposits that are not yet mined;
• rivers without dams;
• settlements without railways;
• cities that have barely begun growing;
• forests and energy resources;
• new ports and trade routes;
• populations that have not yet been fully integrated into national markets.

The next generation therefore does not need to conquer anything.

It can simply develop what the previous generation acquired.

So the economic sequence becomes:

territorial accumulation

→ infrastructure

→ capital accumulation

→ urbanization

→ industrialization

→ mass prosperity

This is why maximum territorial expansion and maximum prosperity do not necessarily occur at the same time.

The expansion creates the asset base.

The consolidation phase monetizes it.

The Three Parts of the Post-Frontier Period

The most useful distinction is probably not between expansion and decline.

It is between three different kinds of consolidation.

1. Prosperous Consolidation

This is the best part of the cycle.

The frontier has mostly closed, but there is an enormous amount left to build.

The society invests heavily in:

railways
roads
electricity
housing
factories
ports
universities
mines
industrial cities

Capital per person rises quickly.

Real wages rise.

The middle class expands.

Ordinary people feel that each generation will have more than the previous one.

This often becomes the era later remembered as the country's golden age.

2. Wealthy Stagnation

Eventually most of the obvious development has already happened.

The country now owns an enormous accumulated stock:

houses
infrastructure
companies
financial wealth
universities
pensions
public institutions
industrial systems

The political economy changes.

A younger society asks:

What can we build?

A mature wealthy society increasingly asks:

How do we protect and divide what already exists?

The marginal dollar increasingly goes toward:

None of the activities on the right are necessarily bad.

But the economic character of society changes.

The stock remains huge.

The rate at which the stock grows slows.

This is when a country can remain extremely comfortable while gradually losing relative position.

Britain, 1920–1970: The Most Useful Example

Britain may be the best analogue for thinking about what a post-peak wealthy society actually feels like from the inside.

Between roughly 1920 and 1970 Britain:

• lost much of its empire;
• lost its position as the dominant world power;
• saw sterling lose enormous international value;
• lost industrial position relative to the United States and Germany;
• experienced a growing brain drain toward North America.

Yet British life did not simply collapse.

By the later part of the period ordinary people had:

• higher real incomes;
• better housing;
• cars;
• televisions;
• household appliances;
• improved medicine;
• expanded education;
• more leisure.

So Britain demonstrated a crucial possibility:

relative national decline can coexist with improving individual living standards.

The country can become less important while its citizens become more comfortable.

At the same time, elite institutional capital proved remarkably durable.

Oxford and Cambridge remained prestigious.

London remained important in finance and culture.

British scientists remained respected.

British musicians and actors continued to become global stars.

Britain stopped being the center of the system without becoming irrelevant.

That is probably a better image of mature imperial decline than dramatic collapse.

Where the United States, Canada and Australia May Be Today

If we treat their classic geographic frontiers as largely closed around 1890–1910, then all three are now roughly a century or more into the consolidation period.

A very simplified timeline would look like this:

The interesting comparison is therefore:

America/Canada/Australia around 2025

may have some structural similarities to

Britain around 1920.

Not because their economies or political systems are identical.

But because both involve societies sitting on an enormous accumulated stock after the end of a long territorial-development cycle.

What the Next Fifty Years Might Look Like

If the analogy works, the next fifty years would not necessarily look like economic disaster.

They might look surprisingly prosperous.

Suppose AI and robotics become the equivalent of electricity, automobiles and mass manufacturing during Britain's post-peak period.

Then:

national relative power ↓

while simultaneously:

ordinary material life ↑↑

The US, Canada and Australia could experience extraordinary improvements in:

• robotic household labor;
• autonomous transportation;
• medicine;
• construction;
• education;
• logistics;
• entertainment;
• personalized services;
• manufactured goods.

And those benefits might eventually diffuse far beyond the major metropolitan centers.

A suburban or regional household in 2070 could command an astonishing amount of machine labor.

That could make smaller cities and suburban regions significantly more attractive.

So one possible trajectory is:

less globally dominant country

but

much more technologically comfortable population.

That is not contradictory.

Britain already demonstrated the basic pattern.

The Geography Could Change Too

Today's assumption is that economic opportunity will continue concentrating indefinitely into cities such as:

New York
San Francisco
Toronto
Vancouver
Sydney

But technological diffusion could reverse part of that.

If AI reduces the value of physical proximity to specialized intellectual labor, while robotics reduces the cost of providing physical services, then regional cities become more viable.

You could eventually have:

smaller city

• cheap automated transportation
• AI medical and professional services
• robotic construction and maintenance
• excellent digital infrastructure

= very high living standard

without the city itself being globally important.

Britain's twentieth century offers a useful precedent.

National influence could concentrate in London while ordinary prosperity spread through suburbs and regional Britain.

The future Anglosphere might similarly combine strong elite centers with widely distributed technological consumption.

Global Rank Can Fall While Domestic Life Improves

This may be psychologically harder than economists usually acknowledge.

Imagine an American in 2070.

His house contains technology unimaginable in 2025.

His medical care is better.

Transportation is easier.

AI performs enormous amounts of intellectual labor for him.

Robots perform physical labor.

Yet he also knows that his country is no longer the unquestioned center of the world economy.

Perhaps the global hierarchy looks something like:

China / India

followed by

United States

followed by several other large economies.

The precise ranking is less important than the change in relative position.

Americans would notice:

• where the largest factories are;
• where the largest cities are;
• which countries build the biggest infrastructure;
• which currencies increasingly matter;
• where ambitious young people go;
• whose companies dominate new industries;
• whose culture foreigners imitate.

That matters psychologically.

People consume not only goods but also status.

So an American could simultaneously believe:

“My life is better than my grandparents'.”

and

“My grandparents lived in a more important country.”

Britain experienced something remarkably similar.

Elite Institutions Usually Decline More Slowly Than National Power

One of the strongest historical patterns is that accumulated prestige is extremely persistent.

Britain lost industrial and geopolitical supremacy.

It did not lose Oxford and Cambridge.

London remained a major financial center.

British scientists, writers and musicians remained internationally important.

The same could happen in the United States.

Even after America's relative economic rank falls:

Harvard may still be Harvard.
MIT may still be MIT.
Stanford may still be Stanford.
New York may still be one of the world's financial centers.

Foreign elites may continue sending their children there.

They may continue keeping money there.

They may continue using American legal, financial and academic institutions.

But the world's most ambitious builders may no longer necessarily believe they must permanently move to America.

A talented Indian entrepreneur might study at MIT and then return to India because the largest growth opportunity is there.

That is similar to what gradually happened to Britain.

The old center remains a place for:

education
prestige
finance
culture
networks

while the center of new physical growth shifts elsewhere.

The Full Cycle

The model can therefore be summarized like this:

The most important dividing line is therefore not simply:

expansion versus decline.

It is:

creating new assets versus protecting accumulated assets.

The Question for the Anglosphere

The United States, Canada and Australia already possess enormous inherited stocks:

territory
natural resources
infrastructure
companies
financial systems
universities
housing
institutions
human capital

That can sustain very high living standards for a long time.

Britain demonstrates that losing relative global position does not automatically destroy ordinary prosperity or elite institutions.

So if the historical analogy holds, the next fifty years might look less like collapse and more like:

extraordinary technological comfort

• strong inherited institutions
• increasing importance of accumulated family wealth
• continued prestige in finance, science, education and culture
• less centrality to world production
• greater nostalgia for the period of national supremacy
• a gradual shift in ambitious global talent toward younger growth centers

The central question is whether AI, robotics and other technologies create a genuine new frontier.

If they merely make existing asset owners richer, then they may accelerate wealthy consolidation.

If they create:

new industries
new cities
new fortunes
new physical development
new mobility
new opportunities for people without inherited wealth

then they may actually restart the cycle.

That gives us two possible paths:

Frontier closes → prosperity → stagnation → decline

or

Frontier closes → prosperity → new technological frontier → renewed expansion.

The United States may be the first major civilization with a realistic chance of repeatedly substituting technological frontiers for geographic ones.

Whether that works again over the next fifty years may matter more than whether America remains number one in any particular annual ranking.

The most vulnerable country in 2050 may not be the poorest.

It may look perfectly normal.

It has thirty or forty million people. Apartment towers. Universities. Airports. Commercial banks. Hospitals. A national electricity grid. A professional military. Perhaps fifty expensive fighter aircraft and a few hundred armored vehicles.

Its citizens expect electricity around the clock, modern healthcare, pensions, safe cities, functioning infrastructure, and very few military casualties.

But the country imports its advanced semiconductors.

It rents much of its cloud infrastructure.

Its most capable AI systems come from foreign companies.

Its satellites are limited or foreign.

Its military software depends on somebody else’s technological ecosystem.

Its economy produces little at the global technological frontier.

It has accumulated nearly all the fixed costs of modern civilization without acquiring the rents generated by owning its most important technologies.

This may become an increasingly dangerous place to be.

For most of the twentieth century, we imagined economic development as a ladder:

village
→ infrastructure
→ industrialization
→ middle-income nation-state
→ advanced economy.

Technological progress may be turning that ladder into something closer to a barbell.

At the top, a small number of enormously capital-intensive technological systems could become richer, more integrated, more automated, and more capable of securing themselves than any societies in history.

At the other end, cheap technology could allow surprisingly small communities to provide much of ordinary civilization for themselves without reproducing the enormous centralized institutions that rich countries built during the twentieth century.

Between them sits the conventional modern state.

Too expensive to become primitive.

Too technologically dependent to become sovereign.

Too centralized to disappear.

And increasingly easy to see, disrupt, tax, coerce, or raid.

Fifty years of shrinking minimum scale

Something important happened during the last half-century that we rarely describe as one phenomenon.

The minimum efficient scale of technological capability collapsed.

Think about what an individual or small organization gained access to between 1975 and today.

Computation:

mainframe
→ personal computer
→ smartphone.

Communications:

national telephone company
→ internet
→ encrypted global communications.

Software:

giant proprietary systems
→ PCs
→ open source
→ cloud services.

Media:

television station
→ website
→ smartphone camera and global distribution.

Navigation:

military or professional surveying infrastructure
→ GPS in every phone.

Manufacturing:

industrial control systems
→ inexpensive sensors, microcontrollers, CNC machines, 3D printers.

Mechanical power:

large industrial machinery
→ cheap electric motors, battery tools, electric vehicles, drones.

Finance:

physical banking infrastructure
→ internet banking
→ mobile money
→ cryptocurrency and stablecoin networks.

Energy may now be going through the same transition.

For most of the industrial era, useful electricity implied an enormous centralized system:

generator
→ transmission network
→ substations
→ distribution grid
→ household.

Solar panels and batteries create a fundamentally different topology.

A household, farm, workshop, clinic, or village can increasingly produce meaningful electricity locally.

The World Bank estimates that off-grid solar could be the least-cost way to provide electricity to roughly 41% of the people who would otherwise remain without access by 2030. Off-grid systems already provided a majority of new electricity connections in sub-Saharan Africa during 2020–22.

The important point isn't that solar is environmentally friendly.

It is that electricity generation is becoming divisible.

That is politically consequential.

Silicon Valley was an early beneficiary

Much of what we call the internet startup revolution was really an economic consequence of declining minimum efficient scale.

A software entrepreneur in 1980 needed substantial capital merely to obtain computing resources.

A software entrepreneur in 2005 could buy a laptop.

By 2015, a tiny company could rent enormous computing infrastructure from Amazon or Google by the hour.

Software itself had almost zero marginal reproduction cost.

A handful of programmers could therefore create products serving millions of people.

This generated an explosion of bottom-up economic experimentation:

startups
independent developers
online commerce
freelance knowledge workers
small media companies
global software businesses.

But Silicon Valley was never actually independent of centralized civilization.

It sat on top of:

semiconductor fabs
universities
national research funding
venture capital markets
reliable electricity
property law
global finance
telecommunications infrastructure.

Centralized technological systems produced cheap tools that decentralized economic agency.

That relationship may now be entering a much more extreme phase.

The frontier is reconcentrating

While the outputs of technology become increasingly distributable, producing the frontier itself is becoming extraordinarily capital intensive.

Advanced semiconductor fabrication requires facilities costing tens of billions of dollars.

Frontier AI requires vast quantities of:

advanced chips
electricity
networking
data centers
specialized engineers
capital.

Satellite constellations require enormous launch, manufacturing, software, and ground infrastructure.

Advanced robotics increasingly depends on an entire ecosystem of:

models
sensors
precision manufacturing
batteries
semiconductors
industrial software.

The technological frontier is therefore moving upward in scale.

AI provides an unusually clean demonstration of both processes occurring simultaneously.

Stanford’s AI Index found that the cost of using AI at roughly GPT-3.5-level benchmark performance fell from about $20 per million tokens in late 2022 to $0.07 by late 2024—a decline of more than 280-fold in roughly eighteen months. Meanwhile, frontier training compute and energy requirements continued rising rapidly.

That is almost the entire thesis in miniature:

creating frontier intelligence becomes more centralized

while

using yesterday’s intelligence becomes radically cheaper.

The same pattern exists with solar panels.

Building the world's dominant photovoltaic manufacturing ecosystem requires gigantic industrial scale.

Using the resulting panel requires a roof.

The same pattern increasingly applies to chips:

creating the semiconductor requires an industrial civilization.

Using it may require a $200 device.

So technological concentration at the top does not imply technological helplessness at the bottom.

The opposite may happen.

The top becomes more powerful precisely by manufacturing increasingly capable tools cheaply enough for the bottom to use.

Civilization may have a much smaller minimum viable population

This leads to a strange question.

How many humans are actually required to sustain a reasonably advanced material life?

Not to manufacture everything.

To live well while remaining resilient.

Perhaps the interesting unit isn't a country of forty million.

Perhaps it is a community of ten thousand.

Imagine ten thousand people living in a town surrounded by productive farmland in 2050.

They do not attempt autarky.

They don't fabricate their own semiconductors or manufacture aircraft.

Instead, they localize things that are bulky, essential, or repeatedly consumed while importing things that are technologically complex but physically small.

That distinction matters enormously.

Energy

The community has:

rooftop solar
a larger community solar field
batteries
a local microgrid
perhaps hydro, wind, biomass, or some dispatchable backup.

Its critical facilities can operate even when the wider national grid fails.

Homes have electricity.

Workshops function.

Water pumps operate.

Food remains refrigerated.

Communications remain online.

The economic implication is much larger than merely lower electricity bills.

The community has eliminated one of the great dependencies of twentieth-century civilization:

the requirement that a distant national institution continuously operate an enormous synchronous machine for local life to function.

Food

Surrounding farmland provides much of the community's food.

But this is not subsistence agriculture.

Agriculture becomes increasingly mechanized and eventually partially autonomous:

electric tractors
automated irrigation
agricultural drones
machine-vision weed control
robotic equipment
AI agronomy
cold storage
local food processing.

A small percentage of the population can therefore feed everyone else.

The community still imports:

fertilizer inputs
replacement equipment
specialized seeds
electronics
machine components.

But interruption of international trade does not immediately mean starvation.

That distinction—between trade for optimization and trade for survival—may become strategically important.

Water

Water infrastructure can also become surprisingly local:

wells
reservoirs
filtration
rainwater collection
solar pumping
wastewater treatment.

This will vary radically by geography.

Ten thousand people in an arid desert face a completely different problem from ten thousand people beside abundant freshwater.

But again, the relevant technologies are becoming increasingly modular.

Housing and construction

This is where 3D printing fits—but probably in a less science-fictional way than “press a button and print a city.”

Current 3D concrete printing can automate meaningful portions of construction, and modular prefabricated printed structures are already being demonstrated. But the technology still relies heavily on conventional reinforcement, foundations, plumbing, electrical work, roofs, windows, finishing, and human quality control. Standards remain a major obstacle to widespread structural deployment.

The larger shift is therefore local digital fabrication:

automated concrete placement
prefabricated modules
CNC cutting
standardized components
AI-assisted design
local timber or earth construction
3D-printed fixtures and replacement parts
robotic excavation and material handling.

A community doesn't need to manufacture every construction material.

It needs enough local equipment and knowledge to convert:

concrete
wood
steel
glass
insulation

into buildings cheaply and repeatedly.

AI could make that easier by turning architectural, structural, electrical, and maintenance knowledge into widely available software assistance.

Construction remains physical.

But less specialized human knowledge may need to be permanently resident in the community.

Education

This is where AI could change geography much more profoundly.

Historically, advanced education generates enormous economies of scale.

A town of ten thousand cannot employ:

physics professor
Korean-language teacher
calculus specialist
programming instructor
mechanical engineer
historian
statistics professor
specialist tutor

for every conceivable subject.

A city of ten million can.

That is one reason knowledge economies concentrate geographically.

But suppose every student has a competent personalized AI tutor.

Then the school's scarce human resources shift toward:

childcare
motivation
supervision
physical education
group projects
socialization
mentoring.

The information layer becomes globally available.

The community may still need excellent teachers.

It no longer needs every possible specialist physically present.

That potentially lowers the minimum population required to offer sophisticated education by orders of magnitude.

Healthcare

Healthcare provides the harder version of the same argument.

A town of ten thousand will not support:

neurosurgery
transplant medicine
pediatric oncology
every surgical specialty.

But most healthcare isn't neurosurgery.

A local health system could plausibly combine:

physicians
nurses
laboratory technicians
pharmacists
AI-assisted diagnostics
imaging
ultrasound
remote consultation
automated medical records
emergency stabilization.

Rare conditions move outward to regional centers.

Routine medicine stays local.

That is already how many small communities work.

AI increases the range of expertise available to the people who remain physically present.

Again, the technology does not replace the human institution.

It raises the amount of civilization each skilled human can support.

Local production and repair

A resilient community would probably care much more about repair than modern consumer societies do.

Its industrial layer might include:

machine shop
welding
electronics repair
woodworking
CNC equipment
additive manufacturing
battery repair
vehicle maintenance
agricultural machinery repair.

This does not create an independent industrial economy.

It creates graceful degradation.

When an imported machine fails, the community has a meaningful probability of repairing or adapting it instead of becoming immediately dependent on the original manufacturer.

Local AI makes the repair shop more capable because technical documentation, diagnostics, programming help, and engineering assistance become cheap.

A capable mechanic with AI potentially substitutes for several narrower specialists.

Transportation

Inside the community:

electric motorcycles
bicycles
small EVs
agricultural vehicles
autonomous delivery systems

may be enough.

Between communities, one serious physical connection remains useful:

road
railway
river
coast.

And possibly a small airstrip.

I would not assume battery-electric aviation solves everything. Aircraft punish battery weight severely, and present electric-aviation research still concentrates heavily on short-range applications.

But the argument doesn't depend on batteries.

Whatever wins—

hybrid aircraft
synthetic fuel
hydrogen
electric short-haul aviation
conventional turboprops—

a small airstrip connects ten thousand people to the global high-complexity economy.

Aircraft are excellent for:

people
medicine
electronics
urgent machine parts.

They are terrible for:

cement
grain
fertilizer
steel.

So resilient communities still want one cheap bulk transport connection.

They just don't require a gigantic national transportation bureaucracy.

Finance becomes thinner too

A sophisticated local economy doesn't necessarily require a national branch-banking architecture.

Money could exist in multiple forms:

cash
conventional digital banking
mobile money
stablecoins
remittance networks
informal credit.

Bitcoin matters historically because it demonstrated that scarce digital assets can exist outside national monetary administration.

But stablecoins may be more consequential for ordinary communities.

A worker abroad can earn in a high-productivity economy and transfer purchasing power directly to a household thousands of kilometers away.

The household then buys:

solar equipment
machinery
smartphones
medicines
electronics.

A poor country's national financial system becomes less important as an intermediary between:

global productive economy

and

local household.

That is potentially a profound shift in development economics.

What actually needs to be imported?

Once you think this way, the list becomes surprisingly short.

A technologically capable ten-thousand-person community might locally provide most of its:

food
water
electricity
housing
basic construction
schooling
primary healthcare
transportation
childcare
elder care
ordinary security
repair
entertainment
local administration.

The imported layer is disproportionately composed of highly complex products:

semiconductors
computers
advanced batteries
pharmaceuticals
specialist machinery
medical equipment
precision bearings
some chemicals
vehicles
telecommunications equipment.

These represent enormous technological complexity but surprisingly little physical volume.

The community doesn't need a semiconductor fab.

It needs a shipment of chips.

It doesn't need an automobile industry.

It needs replacement vehicles every several years.

It doesn't need a pharmaceutical industry.

It needs reliable access to medicines.

The community therefore remains deeply connected to global civilization without reproducing global civilization locally.

That is not autarky.

It is low-fixed-cost civilization.

Premodern social scale, modern technological capability

Ten thousand is also interesting for another reason.

It is large enough for substantial specialization:

farmers
doctors
nurses
mechanics
electricians
builders
teachers
programmers
security personnel
shopkeepers
administrators.

But small enough that much of economic and social life can still operate through:

reputation
family
repeated interaction
religious institutions
local associations
community norms.

Modern industrial civilization replaced much of this with enormous anonymous institutions.

That trade made sense because technological sophistication required scale.

But what happens when sophisticated technology becomes usable at much smaller scales?

You could get something historically strange:

premodern social density + modern material capability.

The community does not have to recreate Sweden.

It doesn't need several layers of national bureaucracies simply to deliver electricity, knowledge, banking, media, and basic expertise.

The technological frontier increasingly arrives as products.

This is not necessarily a libertarian paradise

A thick local society has costs.

It may be:

religious
conformist
patriarchal
nepotistic
intolerant of unusual behavior
dominated by several powerful families.

There may be far less anonymity than in New York.

If your marriage collapses, everyone may know.

If your business reputation collapses, there may be nowhere else locally to go.

A strong community can protect individuals from state failure while simultaneously exerting intense social control.

So the argument is not:

decentralized society = freedom.

It is:

decentralized society can be robust.

Those are different things.

Why the bottom could become surprisingly stable

A coherent local society might actually avoid several pathologies of weak middle-income states.

Consider the struggle to control a conventional capital.

The national government controls:

electricity
television
banks
civil-service jobs
large infrastructure budgets
military procurement
international aid.

Winning national political power therefore offers enormous rents.

That encourages:

coups
patronage politics
ethnic winner-take-all competition
central corruption.

But imagine electricity is mostly local.

Economic support comes substantially from family businesses and diaspora networks.

Information travels through decentralized communications.

Education increasingly comes from global digital resources.

Local security is primarily local.

Now capturing the national capital gives you less.

Political competition can potentially become less existential.

There is less reason to fight over control of the national broadcaster if nobody needs it.

Less reason to capture the energy ministry if villages largely generate their own electricity.

Less reason to control every government job if the government isn't the dominant employer.

The state can become thin precisely because society has become thick.

That does not guarantee peace.

But it changes the economics of political control.

Violence changes rather than disappears

A distributed society may be difficult to conquer but it is not necessarily peaceful in the liberal Western sense.

Violence could become more localized:

family disputes
clan conflict
territorial feuds
contested trade corridors
protection payments
occasional raids
competition between networks.

The key distinction is that the violence may no longer require destroying the entire society.

The community's:

energy
food
housing
communications
productive life

do not depend on capturing one distant capital.

Peace can therefore be strong inside trusted communities while relations between communities become rougher.

That resembles much of premodern political history.

But with drones and smartphones.

The return of the raid

A low-fixed-cost society does not generate enormous technological rents.

It therefore has a different relationship with the wealthy external world.

Historically, frontier societies frequently interacted with richer settled societies through combinations of:

trade
migration
mercenary work
tribute
protection
raiding.

Modern raiding would not usually mean stealing machinery from a semiconductor fab.

A semiconductor fab is nearly useless without its engineers, chemicals, electricity, maintenance systems, and global suppliers.

The valuable target increasingly becomes the interface:

people
cargo
transit routes
resource sites
continuity of operation.

Economic extraction takes forms such as:

ransom
protection payments
tolls
extortion
resource rents.

Transnational criminal markets already demonstrate that organizations do not need one sovereign hierarchy to coordinate across countries.

Different networks can specialize in:

finance
logistics
local access
political protection
transportation.

The system behaves more like a market than an army.

The distributed political world may increasingly behave this way as well.

Humans remain important because physics remains important

This is where extremely rich technological societies and low-fixed-cost societies may diverge most sharply.

The technological top will increasingly substitute machines for human risk.

Its people become expensive.

A lost worker represents:

years of education
high lifetime earnings
political liability
insurance costs
pension commitments
enormous family investment.

A wealthy society will spend increasingly large amounts to prevent human exposure.

Robots inspect infrastructure.

Autonomous vehicles patrol.

Machines perform dangerous industrial work.

Military forces increasingly push sensors and weapons outward while humans remain farther away.

But robotics has a physical problem.

Software travels at approximately zero marginal cost.

Robots do not.

Robots have:

mass
batteries
moving parts
maintenance requirements
weather constraints
communications requirements.

They encounter:

mud
rocks
water
dust
trees
stairs
unpredictable humans.

Robotic logistics research still treats reliability, networking, contested environments, and physical integration as major constraints.

So even an extraordinarily advanced society may become capable of seeing much of the world without becoming capable of economically policing all of it.

Global eyes, local hands

This could define the high-tech political order.

Satellites watch huge areas.

AI continuously searches imagery.

Financial networks identify transactions.

Telecommunications reveal relationships.

Cyber systems operate globally.

Information becomes cheap to move.

Physical control remains expensive.

So the advanced technological civilization may eventually conclude:

We can see that territory.

We know roughly who controls it.

We know who trades with them.

We do not intend to govern it.

Its security policy turns inward.

Protect:

borders
ports
major cities
financial systems
citizens
strategic supply chains.

Make participation in domestic economic life increasingly authenticated.

Employment connects to verified identity.

Banking connects to identity.

Property ownership connects to identity.

Critical infrastructure connects to trusted hardware and software.

Robotics provides physical enforcement where the economics justify it.

The successful high-tech state becomes something like a fortress civilization.

Not necessarily authoritarian.

But highly legible.

The cost of living anonymously inside it rises dramatically.

The middle owns the targets

Now return to the perfectly normal country we started with.

It cannot operate like the local community.

Its population has been urbanized and socially atomized.

Millions depend on:

national electricity
commercial banking
centralized healthcare
formal employment
imported food systems
complex transportation.

Extended family structures may have weakened.

Local communities no longer provide most welfare or security.

People expect the state to work.

But the state cannot operate like the technological fortress either.

It doesn't own:

frontier AI
advanced semiconductor manufacturing
major satellite networks
globally competitive cloud infrastructure
robotic industrial ecosystems
advanced intelligence networks.

It imports them.

So it has inherited the cost structure of the top without acquiring the top's rents.

And it has abandoned much of the social resilience of the bottom.

This is the middle trap.

The middle receives invoices

The technological top earns rents from:

chips
models
cloud
finance
satellites
robotics
intellectual property.

The distributed bottom consumes their older outputs cheaply.

The middle purchases frontier capability at retail while maintaining an expensive twentieth-century society.

It needs:

universities
hospitals
pensions
bureaucracy
professional military
national grids
highways
banks
expensive cities.

It must also purchase:

chips
AI
cloud
weapons
industrial machinery
capital.

The top receives monopoly or oligopoly rents.

The bottom receives technological spillovers.

The middle receives invoices.

And then it has to defend everything those invoices paid for.

Russia and Ukraine show pieces of the problem

Russia and Ukraine are not pure examples of this future.

Russia has nuclear weapons, missiles, space capabilities, a substantial industrial base, and a serious military technology sector.

Ukraine has a highly innovative technology ecosystem supported by a functioning state and an enormous Western technological backend.

But their war illustrates an important feature of the middle-state problem.

They must defend:

cities
power generation
railways
factories
bridges
apartment buildings
telecommunications
recognizable territorial government.

Cheap precision systems make these assets increasingly targetable.

But neither society can simply abandon them.

The result is enormously expensive warfare in which technological adaptation can dramatically improve lethality without necessarily producing cheap political control.

Ukraine's direct physical reconstruction needs are already measured in hundreds of billions of dollars. The cost lies disproportionately in exactly the things a complex territorial state cannot simply abandon: housing, transport, energy, and other fixed infrastructure.

A decentralized community can lose a drone workshop and rebuild another.

A modern state cannot casually lose a power station serving two million people.

We have been here before

The industrial revolution created a surprisingly similar technological barbell.

At the top it created:

battleships
industrial artillery
railways
telegraphs
huge arsenals
modern state taxation.

The great powers weren't powerful merely because they owned guns.

They possessed entire industrial systems capable of:

producing weapons
moving armies
financing wars
replacing losses.

Some states tried to buy the outputs without reproducing the system.

Qing China purchased modern ships and weapons and built arsenals, but modernization remained institutionally fragmented.

The Ottoman Empire built substantial modern military capabilities but increasingly depended on European finance to fund them.

Several Latin American countries purchased world-class European warships despite lacking anything resembling Britain's underlying naval-industrial ecosystem.

Brazil could buy a dreadnought.

It could not buy the civilization that continuously produced dreadnought power.

That was one side of industrialization.

The other side was equally important.

Factories also produced enormous quantities of increasingly cheap:

rifles
ammunition
explosives
radios
trucks.

The minimum cost of organized resistance fell.

An anti-colonial movement did not need:

a battleship fleet
an aircraft industry
a giant steel sector.

It needed enough inexpensive technology and human organization to make foreign rule expensive.

The stronger side needed the capacity to impose continuous control.

The weaker side only needed enough capability to prevent control from becoming cheap.

That asymmetry helped transform twentieth-century politics.

AI and distributed energy may be creating the next version.

Development may no longer be a ladder

The twentieth-century development model assumed poor societies would gradually reproduce the infrastructure of wealthy ones.

Build the grid.

Build the university.

Build the banks.

Build the bureaucracy.

Build the highways.

Build the industrial base.

Urbanize.

Construct a large middle class.

Eventually become rich.

That pathway may no longer be the only technologically viable one.

A community could potentially move from:

weak state + low material capability

to:

weak state + surprisingly high local capability

without passing through every intermediate institution.

We've already seen narrow forms of this.

Some countries largely skipped universal landline networks and went directly to mobile phones.

Some communities may skip universal grid electricity and go toward distributed solar.

Finance can partially leapfrog bank branches.

AI may allow expertise to leapfrog large professional institutions.

Automated fabrication may allow portions of construction and manufacturing to leapfrog highly specialized labor systems.

These are not complete replacements.

But they don't need to be.

Each one removes another reason why ten thousand people require a state designed for forty million.

The changing minimum efficient scale of civilization

That may ultimately be the most important technological story here.

Industrial modernity increased the minimum efficient scale of sophisticated civilization.

You needed:

factories
grids
universities
bureaucracies
national transportation
large capital markets.

That favored enormous integrated states.

The technologies of the last fifty years have quietly begun reversing portions of that relationship.

Cheap computing reduces the scale required for information processing.

The internet reduces the scale required for communication.

AI reduces the scale required for expertise.

Solar reduces the scale required for electricity generation.

Batteries reduce the scale required for energy reliability.

Drones reduce the scale required for aerial capability.

Digital fabrication reduces the scale required for some manufacturing and construction.

Cryptographic networks reduce the scale required for trusted digital coordination.

None of these eliminates civilization's industrial core.

In fact, the industrial core becomes more concentrated than ever.

But fewer people need to live inside that core's organizational structure to consume what it produces.

That creates the strange possibility of two simultaneously successful worlds.

The fortress and the village

At one pole:

enormous capital
advanced AI
semiconductor fabs
satellites
robotics
integrated finance
highly legible citizens
low human exposure.

The society becomes extraordinarily productive and extraordinarily expensive.

At the other:

solar panels
batteries
cheap electronics
local AI
drones
family networks
community institutions
local agriculture
repairability
low fixed costs.

The society is much poorer.

But it may be surprisingly difficult to break.

One pole maximizes integration.

The other maximizes resilience.

And neither looks much like the ordinary twentieth-century nation-state.

The real losers may be the normal countries

I would not put precise probabilities on this world.

Nor will every rich country become a fortress, every poor community become resilient, or every middle-income country fail.

There is also a genuinely terrible fourth equilibrium:

weak state

• weak community
• cheap technology.

There, drones empower gangs.

AI helps scammers.

Solar powers armed compounds.

Communications make predatory organizations more capable.

Cheap technology does not create trust.

It amplifies whatever social structure already exists.

So the bottom itself bifurcates:

distributed resilience

versus

distributed predation.

But if even a substantial minority of lower-income societies achieve the first equilibrium, the conventional development hierarchy starts looking much stranger.

The richest societies become nearly impossible to penetrate because their entire domestic environment is technologically integrated.

The resilient local societies remain much poorer, but increasingly difficult to dominate because very little depends on a single vulnerable system.

The exposed middle gets neither advantage.

It has:

valuable citizens
expensive buildings
visible infrastructure
complex welfare obligations
sophisticated consumption expectations
imported technology
weak local resilience.

It is wealthy enough to possess targets.

Not wealthy enough to make those targets safe.

The future may be U-shaped

The usual assumption is that poor societies want to become middle-income societies and middle-income societies want to become rich.

Economically, that will remain true.

Politically and strategically, something more complicated may happen.

The stable equilibria could increasingly sit at opposite ends.

At the technological top:

enormous fixed investment
enormous upstream rents
machine coordination
domestic legibility
high individual value
low tolerance for physical risk.

At the resilient bottom:

low fixed costs
imported commodity technology
local energy
local food
human relationships
informal welfare
cheap machines
redundant institutions.

Between them:

centralized infrastructure
imported intelligence
expensive human capital
weak communities
high expectations
limited technological sovereignty.

The technological top owns the frontier.

The resilient bottom buys its discarded miracles cheaply.

The middle tries to reproduce the appearance of the top without owning the economics underneath it.

That may be the least sustainable position of all.

For fifty years, technology has lowered the minimum scale at which people can become productive.

The next fifty may lower the minimum scale at which people can sustain something approaching civilization.

If that happens, the nation-state will face pressure from both directions.

At the top, technological systems become larger than most countries can afford to reproduce.

At the bottom, communities become capable of doing more without their countries.

And the central political question of the twenty-first century may become unexpectedly simple:

How much state do ten thousand technologically equipped people actually need?

A useful starting point is Wheat Waffles’ 2022 video “40 Differences in Treatment Between a Sub5, Normie & Chad,” which has been viewed more than 1.1 million times.

The video divides men into three crude attractiveness categories and argues that each group experiences a meaningfully different dating market:

Chads — highly attractive men
Normies — ordinary-looking men
Sub-5s — clearly unattractive men

Wheat Waffles claims that, based on his own face-rating data:

9% of men are “Chads”
77% are “normies”
14% are “sub-5s”

These are not scientifically established population statistics. They come from a self-selected face-rating service, so the exact percentages should be treated skeptically.

But the model itself is interesting because it makes a stronger claim than “looks matter.”

It argues that men in these three groups experience different market regimes.

A Chad does not merely get somewhat better results.

Women sometimes approach him.

They make interactions easier.

They laugh at jokes that might not work for another man.

They find excuses to touch him.

He can supposedly use mediocre dating-app photos and still get matches.

He may obtain sex without first supplying commitment.

The video summarizes his position as:

women compete for his attention.

The normie experiences something else.

Women generally treat him neutrally at first.

He gets rejected frequently.

He needs better photos.

Social skill matters.

Fitness matters.

Shared interests matter.

“Game” matters.

Warm introductions work better than cold ones.

He may need to provide commitment before obtaining the same sexual access a much more attractive man receives easily.

The video summarizes his position as:

he competes with other men for women.

And then there is the “sub-5.”

In the video’s harshest formulation, even improvement has low returns.

Confidence may be read as arrogance.

Flirting can become creepy.

The gym is interpreted as compensation.

Dating apps produce essentially nothing.

The video claims that these men are effectively excluded from much of the market.

Its summary is:

they compete merely to participate.

Again, this is deliberately exaggerated internet language, not a scientifically demonstrated caste system.

But strip away the terminology and there is a serious underlying question:

What happens when a market maximizes individual freedom and exit, but bargaining power is distributed very unequally?

That question applies surprisingly well to both employment and dating.

And the labor market provides a useful analogy.

There are two genuinely good positions in a free labor market

Suppose you live in a country where employers can terminate workers relatively easily.

There are still two ways to feel economically secure.

The first is individual market power.

You are highly demanded.

Recruiters contact you.

Several companies would hire you.

You can negotiate compensation.

You can reject bad managers.

If your employer fires you tomorrow, it is unpleasant, but you are not terrified that you will spend the next twelve months sending out 400 applications.

Your employer has an exit option.

But so do you.

You have leverage because your outside option is strong.

The second good position is institutional protection.

Maybe you have tenure.

Maybe you work in a strongly unionized sector.

Maybe employment law makes dismissal difficult.

Maybe termination requires cause, notice, severance, arbitration, seniority rules, or collective bargaining.

You personally may not be exceptional.

But you don't need to be.

The institution gives ordinary workers bargaining power they could never obtain individually.

These are two very different forms of security:

Market security: “You can fire me. Someone else wants me.”
Institutional security: “You cannot cheaply fire me in the first place.”

Either can work.

The terrible position is having neither.

The employee with neither protection nor demand

Imagine a respectable middle-class professional.

He has a degree.

He has seven years of experience.

He is competent.

But his labor market is oversupplied.

His employer can eliminate his job next week.

And if that happens, a comparable replacement could take six to twelve months.

He applies to hundreds of positions.

Most never respond.

A few send automated rejections.

He gets several interviews.

One company puts him through five rounds and chooses somebody else.

Eventually he gets another job.

Then the whole risk resets.

What advice does this person usually receive?

learn another technical skill
become better at Excel
learn to code
get a certification
improve communication
network more
accumulate another three years of experience.

All of these things may improve employability.

But notice what they do not provide.

They do not provide structural security.

An additional certification might increase the probability of getting an interview from 4% to 6%.

That is useful.

But the worker is still living in a market where:

his employer can terminate him easily
replacement employment is difficult to obtain.

The underlying power relationship hasn't changed much.

This is why “be indispensable” is not actually the same thing as job security.

An employee can know every internal system.

His coworkers can depend on him.

Management can like him.

He can work extremely hard.

Those things reduce the probability that the company fires him.

They do not remove the company's ability to do it.

And if being fired would be catastrophic, that changes his behavior long before anyone actually fires him.

He becomes cautious.

Don't antagonize management.

Keep delivering.

Keep learning.

Stay useful.

Avoid negotiating too aggressively.

Do not become expensive.

Do not let your performance slip.

This may look like psychological insecurity.

But some of it is simply rational bargaining behavior.

If one side can terminate the relationship relatively cheaply and the other side needs twelve months to recover, they do not have equal leverage.

Now apply the same framework to dating

Modern relationships increasingly maximize individual choice.

You can date whom you want.

You can reject whom you want.

You can leave a boyfriend or girlfriend.

And modern marriage, despite legal obligations around property, children and support, generally cannot compel somebody to continue being romantically attached to you.

That is an enormous expansion of individual freedom.

But free exit has distributional consequences.

The employment market has two ways to protect a weak participant:

high market demand
or institutional protection.

Modern dating retains the first.

It has largely weakened the second.

There is no romantic labor union.

There is no collective bargaining agreement governing affection.

No government can appropriately order someone:

“You must continue loving this person.”

So at the most fundamental level, the contemporary relationship operates under something close to:

continued participation by mutual consent.

Either side can leave.

That makes the quality of your outside option extremely important.

And now the Chad/normie/sub-5 distinction starts looking less like internet slang and more like a crude theory of bargaining power.

Chad: the worker recruiters are fighting over

The Chad is analogous to the highly demanded professional.

His biggest advantage isn't merely that he gets more matches.

It is that losing one relationship does not threaten his access to the entire romantic market.

If a girlfriend leaves him:

other women are interested
he gets inbound attention
dating apps work
approaching works
casual opportunities exist
rebuilding may be relatively quick.

So his internal negotiating position is fundamentally different.

He can credibly say:

“I want this relationship, but I do not need this relationship at any cost.”

That doesn't mean he is emotionally indifferent.

A billionaire can love a company he owns.

A star employee can genuinely care about his job.

The point is that loss is survivable.

That creates optionality.

And optionality creates leverage.

The video actually captures this repeatedly.

It claims women may:

approach attractive men
initiate touch
laugh more readily
tolerate weak photos
accept casual arrangements
become frustrated when he does not text back
compete for his attention.

If even part of that is directionally true, the important implication isn't simply:

Chad gets more sex.

It's:

Chad operates on the choosing side of the market.

He is the scarce worker.

Normie: employed, but structurally insecure

The normie occupies a stranger position.

He is viable.

He can date.

He can marry.

He is not excluded.

But according to the video's model, his success depends much more on what he adds to his baseline appearance.

Fitness.

Career.

Humor.

Social competence.

Good photographs.

Style.

Confidence.

Shared interests.

Persistence.

Timing.

Commitment.

The video repeatedly contrasts the normie with the Chad this way.

The Chad can allegedly post a bad mirror selfie and still get results.

The normie needs strong photos.

The Chad can attract women largely through appearance.

The normie needs “a few other things going for him.”

The Chad may not need “game.”

The normie benefits substantially from learning it.

The Chad sometimes receives female pursuit.

The normie generally has to initiate.

This means the normie resembles the competent middle-class employee.

He can obtain the position.

But he has to compete hard for it.

And losing it may be expensive.

Suppose an ordinary man spends a year dating before finding a woman he genuinely wants.

He initiates dozens of conversations.

He goes through rejections.

He plans dates.

He develops the relationship.

Eventually they become exclusive.

Now compare their outside options.

If the relationship ends, perhaps rebuilding something comparable takes him another year.

Maybe longer.

He therefore has a powerful incentive to retain the relationship.

This produces advice that sounds remarkably similar to employment advice:

stay attractive
keep earning
maintain confidence
remain interesting
don't become complacent
keep dating your wife
keep improving yourself.

Again, none of this is necessarily bad advice.

The problem is what kind of advice it is.

It is the romantic equivalent of:

keep improving your professional skills so your employer continues to value you.

That lowers termination risk.

It does not eliminate structural vulnerability.

If your outside option remains poor, you are still highly exposed to unilateral exit.

Sub-5: the worker in a collapsed labor market

The video's most controversial category is the “sub-5.”

Its claim isn't merely that unattractive men have worse outcomes.

It claims their return on investment collapses.

Going to the gym has little impact.

Confidence is reinterpreted negatively.

“Game” barely works.

Dating apps produce almost no legitimate matches.

Cold approaches fail.

Warm approaches fail.

Money may buy access but not the same underlying desire.

That is an extremely strong claim, and I do not think the evidence establishes such a hard threshold.

But economically, the model is easy to understand.

Imagine someone living in a region where demand for his occupation has collapsed.

People tell him:

improve your résumé.

He does.

get another certificate.

He does.

learn another software tool.

He does.

become more confident in interviews.

He does.

But there are still 800 applicants for every decent job.

At some point, marginal improvement is no longer the obvious strategy.

The rational question becomes:

Why am I continuing to optimize inside this market?

And that leads to a very different set of options.

Self-improvement is not the same thing as structural security

This distinction matters.

If you're poorly positioned in the labor market, learning Excel may help.

But there are larger strategies:

move to another city
move to another country
change industries
enter a protected profession
retrain into a scarce occupation
leave the labor market and organize your life differently.

Dating has analogous strategies.

“Gymmaxxing” is basically human-capital investment.

So is earning more.

So is becoming more socially skilled.

These may improve your rank.

But if you're still in a market where your bargaining power remains low, they haven't changed the structure.

A man can add fifteen pounds of muscle and still have weak outside options.

A man can increase his income from $80,000 to $150,000 and still discover that highly attractive men receive spontaneous desire that money does not replicate.

There is no obvious salary where:

$X income = Chad.

And if an ordinary man needs top-one-percent income merely to approach the negotiating position an attractive man received genetically, that itself demonstrates the imbalance.

Strategy one: become genuinely scarce

The cleanest solution is obvious.

Become the romantic equivalent of a highly recruited employee.

Become sufficiently desirable that women compete for access to you.

That maximizes individual optionality.

You no longer need institutional protection because you can afford termination.

But this solution has the same problem as telling every worker:

become a world-class machine-learning engineer.

It cannot be population-wide advice.

Scarcity is relative.

Everyone cannot simultaneously occupy the top 10%.

Strategy two: move markets

This is where geography becomes much more interesting than generic self-improvement.

Suppose a man is mediocre in Toronto.

There are thousands of similar men.

His ethnicity is ordinary.

His profession is ordinary.

The sex ratio isn't favorable.

Women with his preferred characteristics receive enormous attention.

Now put exactly the same man somewhere else.

Perhaps:

his nationality is rare
his education carries more status
his income is unusually high locally
his appearance fits local preferences better
the sex ratio is more favorable
competing men are fewer.

He hasn't changed.

His relative market position has.

This is geomaxxing in its most economically coherent form.

It is exactly like moving from a city where 5,000 programmers compete for 100 openings to one where companies are desperate for programmers.

You do not need to become globally elite.

You need to become relatively scarce in your actual market.

In that sense, the objective is to become locally Chad-like.

Strategy three: reintroduce institutional protection

There is another possibility that modern dating discussions often ignore:

religious or tightly embedded communities.

Religion can function, imperfectly, like a labor institution.

Not because it legally traps anybody in marriage.

Modern personal autonomy prevents a true romantic equivalent of employment tenure.

But embedded communities can still create:

mediated introductions
family involvement
strong norms around fidelity
reputational consequences for cheating
expectations that marriage survives temporary dissatisfaction
community mediation during conflict
pressure against opportunistic exit.

That changes the relationship from a completely isolated bilateral market transaction.

There are third parties.

There are norms.

There is an institution.

The analogy isn't perfect, but it is considerably closer to:

union + professional guild + mediation mechanism

than to Tinder.

This is why simply dating someone who checks “Christian” or “Jewish” on an app is not the same thing.

The protection comes from actual embeddedness.

If neither partner participates in the institution, there is nothing to enforce the norm.

Strategy four: deliberately choose a market where you are on the strong side

There is another uncomfortable possibility.

Instead of maximizing the quality of the partner you can barely obtain, choose within a pool where you are highly demanded.

The employment analogy is straightforward.

Suppose you can barely get a $200,000 position where every employer has twenty equally qualified replacements.

Or you can take a $150,000 role in a market where five companies desperately want your skills.

The first maximizes headline compensation.

The second may maximize bargaining power.

Dating has a similar frontier.

A man can continually pursue the most attractive women who will occasionally accept him.

That may place him at the bottom of their feasible choice set.

Or he can select somewhat differently and become unusually desirable within that pool.

This doesn't mean dating someone you find unattractive.

That creates a different problem.

It means recognizing that:

maximum obtainable partner quality and maximum relationship security are not necessarily the same optimization target.

Strategy five: exit

And there is one solution both labor-market and dating discussions routinely underweight.

Stop participating.

When a region has terrible employment prospects, not everyone becomes a software engineer or emigrates.

Some people reduce career ambition.

They live with family.

They take low-paying but abundant work.

They organize life around children, parents, church, community, hobbies, or leisure.

They effectively decide:

the career market isn't where I'm going to compete for status.

There is a romantic equivalent.

If someone concludes:

courtship requires too much effort
his outside options remain weak
relationship termination would be extremely costly
he dislikes the bargaining structure
changing geography or institutions isn't attractive

then one coherent response is:

stop organizing adulthood around obtaining a relationship.

Keep friends.

Keep family.

Build a business.

Become excellent at a craft.

Travel.

Compete professionally.

Play sports.

Create things.

Mentor younger people.

Build wealth.

Develop intellectual interests.

This is not the comforting advice:

“Stop looking and love will find you.”

For an ordinary man with little inbound attention, that may simply mean nobody comes.

True exit requires accepting that possibility.

It means:

I may not have a long-term romantic relationship, and I am deliberately choosing the alternative allocation of my life anyway.

That is a real compromise.

But staying in a market where you dislike the terms is also a compromise.

Free markets maximize choice. They do not maximize security.

That is the larger point.

Liberal societies place enormous value on individual optionality.

Employers can change employees.

Employees can quit employers.

People can choose partners.

People can leave partners.

This produces freedom.

But the value of freedom depends partly on your bargaining position.

For the highly demanded worker:

freedom is leverage.

For the worker nobody else wants:

freedom mostly means the employer is free to replace him.

The same formal rule creates radically different lived experiences.

That may also be true in dating.

For the Chad:

“Either of us can leave”

can feel liberating.

He has alternatives.

For the normie:

“Either of us can leave”

may feel considerably more precarious because rebuilding the relationship could require another long cycle of rejection, courtship and search.

And for someone at the bottom of the market, the loss may feel potentially irreversible.

This suggests that much mainstream relationship advice answers the wrong question.

It asks:

How can you become a better partner so that your relationship succeeds?

That is equivalent to asking:

How can you become such a good employee that your company doesn't fire you?

Useful question.

But not the only one.

The structural questions are different:

How strong is my outside option?
Am I in the right market?
Can I enter a market where I'm relatively scarce?
Is there an institution that meaningfully protects commitment?
Am I maximizing partner quality at the expense of bargaining security?
And if I dislike all available versions of this market, should I participate at all?

Those are less romantic questions.

But they may be more honest ones.

Volkswagen is considering a restructuring that could ultimately eliminate as many as 100,000 positions. Roughly 50,000 reductions have already been agreed across the group, while CEO Oliver Blume says another roughly 50,000 may be required to bring Volkswagen’s cost structure closer to competitors. The additional figure is not yet a formal headcount target; Blume describes it as a benchmark for the scale of savings required. VW says its overhead costs remain more than 30% above comparable rivals.

Reuters — VW weighs up to 100,000 job cuts, four plant closures in biggest overhaul yet

That is an extraordinary number.

Volkswagen employed 662,942 people worldwide at the end of 2025, including its Chinese joint ventures. About 284,000 worked in Germany. So 100,000 positions would correspond to roughly 15% of the company’s global workforce.

It is tempting to read this simply as a Volkswagen story:

VW was too bureaucratic.
It was slow on electric vehicles.
Chinese EV makers got better.
Trump imposed tariffs.
Therefore VW needs layoffs.

All of those things matter.

But I think the more interesting interpretation is broader.

Volkswagen may be one of the first places where the economics of the post-Cold War European industrial system are being visibly repriced.

And to understand why, the wrong comparison is:

Germany vs. China.

The better comparison is:

the Germany-centered European supply chain vs. the Chinese supply chain.

Those are the actual competing industrial organisms.

Germany never competed with China using German workers alone

Imagine Germany around 2005–2015.

A German automaker could combine:

• German engineering
• German machine tools
• German management and capital
• Czech assembly
• Polish components
• Hungarian engines and electronics
• Slovak production
• relatively cheap Russian energy
• frictionless access to a huge European market.

That was an extremely powerful arrangement.

Germany itself was expensive, but the production system surrounding Germany was not.

The fall of communism, followed by EU enlargement, effectively gave German industry access to a large nearby labor reservoir that was:

geographically close,
increasingly skilled,
politically stable,
integrated into the EU market,
and dramatically cheaper than German labor.

Volkswagen’s own employment map illustrates how integrated this became. At the end of 2025 VW employed about:

• 36,800 people in Czechia
• 20,100 in Poland
• 12,300 in Hungary
• 12,000 in Slovakia

in addition to its enormous German workforce.

This was not just Volkswagen outsourcing.

The entire Central European auto ecosystem developed around the same logic.

So Germany’s true competitive unit was something closer to:

German industrial core + lower-cost Central European manufacturing hinterland.

In that sense, the EU created a partial equivalent to what China already possessed inside one country.

China’s low-cost hinterland was its own countryside

China’s version was much larger.

The classic image of Chinese industrialization is a factory in Shenzhen or Dongguan full of workers.

But those workers did not originally come from Shenzhen.

For decades, China could draw workers out of poorer rural and inland areas and move them into industrial clusters along the coast.

The hukou system, the enormous rural-urban income gap and China’s sheer population created an industrial labor reservoir on a scale Europe never approached.

Even in 2025, after decades of urbanization, China still counted 301.15 million rural migrant workers.

Of those, 28.2% worked in manufacturing.

That implies roughly 85 million migrant manufacturing workers alone.

Think about that number for a moment.

That is not China's entire manufacturing workforce.

It is just the migrant-worker component.

China's total population was still 1.405 billion at the end of 2025, with about 851 million people aged 16–59. Some 451 million residents were still classified as rural.

The EU, by comparison, had 452 million people at the beginning of 2026.

So China's total population is a little more than three times larger.

But the deeper difference historically was not merely population.

It was the ability to keep pulling lower-income workers into more productive industrial employment.

Europe had a smaller version:

Germany → Czechia / Poland / Slovakia / Hungary.

China had:

Shanghai / Guangdong / Zhejiang / Jiangsu
← workers from Henan / Anhui / Sichuan / Hunan / Guangxi / inland China.

And as coastal wages rose, production could itself move inland.

In other words:

Europe internationalized its low-cost hinterland. China internalized it.

That distinction matters.

The Czech and Polish workers are no longer that cheap

The European model worked beautifully partly because the wage gap was enormous.

But successful convergence eventually destroys labor arbitrage.

In 2025, manufacturing labor costs per hour were approximately:

Those are still enormous discounts relative to Germany.

A Polish manufacturing worker does not suddenly cost as much as a German one.

But Poland at €17/hour is very different economically from the Poland that entered the EU in 2004.

Likewise Czechia.

Likewise Slovakia.

Likewise Hungary.

This is not a policy failure on their part. It is what successful development looks like:

foreign investment
→ factories
→ productivity
→ skills
→ tighter labor market
→ higher wages.

But from Germany’s perspective, one of its industrial shock absorbers is gradually disappearing.

The old system was:

expensive Germany + very cheap Eastern Europe.

Increasingly it is:

very expensive Germany + medium-cost Eastern Europe.

That is a much less compelling cost structure against Asia.

Compare that with the Chinese factory labor base

China is no longer the ultra-cheap country of 2002 either.

Chinese wages have risen enormously.

But they remain much lower than Central European industrial labor costs.

China's National Bureau of Statistics says rural migrant workers employed in manufacturing earned an average 5,126 yuan per month in 2025.

JETRO’s 2025 survey of Japanese companies in Asia found an average monthly base salary of about $629 for a regular manufacturing worker in China with three years of experience.

Those figures should not be mechanically compared with the European €15–€20-per-hour figures.

The European number is total employer labor cost per hour, including non-wage costs. The Chinese figures are worker income/base salary. Different surveys cover different workers, locations and benefits.

So saying:

“China costs exactly one-quarter of Poland”

would be false precision.

But the direction and order of magnitude are difficult to miss.

The European automotive system is attempting to retain mass industrial production using labor that costs:

roughly €15–€20 an hour even in its lower-cost core production countries,

while Chinese factories still have access to huge pools of workers whose direct wages remain only a fraction of that.

And there is another difference.

Chinese official statistics reported enterprise employees working around 48.6 hours per week on average in 2025. Again, this is not perfectly comparable internationally, but it gives some idea of the labor-input environment.

Europe has moved in essentially the opposite direction for decades.

This is where Europe’s demographics become more dangerous

Both China and Europe are aging.

China's population is already shrinking.

So it would be wrong to tell a simple story in which:

young China defeats old Europe.

China is not young anymore.

In 2025, 23% of China's population was already 60 or older.

But the two regions entered aging with very different industrial structures.

Europe has combined:

aging

• unusually low working hours
• very high labor costs
• an Eastern European labor pool rapidly converging toward Western wages.

China has:

aging

• substantially longer working hours
• much lower factory wages
• a vastly larger remaining workforce
• massive internal migration infrastructure
• deeper manufacturing clusters.

China therefore has demographic problems without yet having European labor economics.

That distinction matters.

China's cheap-labor advantage is eroding.

Europe's eroded earlier.

The supply chain itself is the advantage

The wage comparison also misses the most important part of China's advantage.

A Chinese automaker does not simply hire a cheaper assembly worker.

It operates inside an ecosystem containing:

batteries
electric motors
power electronics
displays
semiconductors
castings
steel
chemicals
machine tools
tooling
logistics
software engineers
component suppliers

at enormous scale and often within a few hours of one another.

So the competition is not:

€49 German worker versus ¥5,126 Chinese worker.

It is:

European system cost per competitive vehicle versus Chinese system cost per competitive vehicle.

That is much harder for Europe to solve.

Europe built an impressive continental supply chain of its own.

Germany was the high-value hub.

Czechia, Slovakia, Poland and Hungary supplied lower-cost manufacturing.

France, Italy, Spain, Austria and northern Italy added additional engineering, components and production depth.

For a period, Europe had something resembling a 450-million-person integrated factory.

But China built the same concept with 1.4 billion people, one national government, one internal market and much greater industrial density.

That difference becomes increasingly important as European labor costs converge upward.

Energy makes the comparison worse — although China is not energy independent

Europe and China are both large energy importers.

But their situations are not symmetrical.

The EU imported a net 57% of its energy needs in 2024. Germany's dependency was about 67%.

China imports enormous quantities of oil and gas too.

But China also produced about 4.73 billion tonnes of coal in 2025. The IEA explicitly describes domestic coal production as a cornerstone of China's energy security. China is simultaneously building solar, wind, nuclear and power infrastructure at enormous scale.

Germany's old model was unusually elegant:

cheap Russian energy

• cheap Central European labor
• German technology
• Chinese demand.

Look at what has happened to each component.

Cheap Russian energy:

largely gone.

Cheap Central European labor:

progressively less cheap.

Chinese demand:

increasingly replaced by Chinese competition.

German technology:

still excellent, but no longer uniquely ahead in automobiles.

That is the context in which Volkswagen is discussing 100,000 positions.

This is why the VW story feels sudden even though it isn't

Structural problems often appear all at once because firms can live with them for years.

Suppose a German factory was built in 2008.

The capital is already sunk.

Workers are already trained.

Suppliers already exist.

China is buying cars.

Energy is cheap.

Even if demographic projections look terrible for 2035, closing the factory today makes no sense.

Then fifteen years pass.

Now management has to decide whether to invest billions in the next generation of that plant.

At precisely that moment:

China becomes a competitor.
European volumes weaken.
electricity and gas remain expensive.
Central European wages have risen.
EV architecture requires fewer mechanical components.
tariffs make exports harder.
the workforce is aging.

Suddenly the calculation changes.

The factory did not become uneconomic overnight.

The option to postpone confronting its economics expired.

That is why restructuring tends to arrive as a flood.

Germany may be entering its Japan phase — but Europe is the better comparison with China

There is an obvious Japan analogy.

Japan in the 1990s entered aging with:

high wages
world-class manufacturing
excess industrial capacity
mature domestic demand.

Over the following decades, Japanese firms closed plants, automated, moved production abroad and specialized more heavily in areas where accumulated industrial know-how remained valuable.

Japan remained an industrial power.

But it employed far fewer people producing ordinary manufactured goods domestically.

Germany may be approaching something similar.

BMW, Mercedes, Siemens, BASF, Porsche and VW do not need to disappear for German industrial employment to fall substantially.

Japan demonstrates that distinction very clearly.

But Japan is no longer the right comparison with China.

Modern Japan increasingly resembles a giant high-value industrial node embedded inside the broader Asian production system.

It supplies sophisticated:

machinery
vehicles
components
materials
robotics
specialty chemicals

into much larger markets around it.

In that sense, Japan is closer to a very large, unusually industrial version of Switzerland than to China.

The more interesting macro comparison is:

China's continental manufacturing system

versus

Europe's continental manufacturing system, historically organized around Germany.

And on that comparison, Europe has a problem.

Europe's low-cost hinterland got rich

Perhaps the simplest way to describe the entire story is this.

Europe once had:

Germany + its own emerging-market manufacturing zone.

China had:

coastal China + its own enormous emerging-market countryside.

Both systems used lower-cost workers to support higher-value industrial centers.

Both are aging.

Both are seeing their poorer regions converge.

But China's labor reservoir was dramatically larger, its internal market is dramatically larger, and its supply-chain density has become much deeper.

The Czech Republic becoming richer is good news for Czech citizens.

Poland becoming richer is good news for Poland.

But economically, it means those countries progressively stop playing the role they played for German manufacturers in the 2000s.

Eventually the question becomes:

If German labor costs €50 an hour, Czech labor costs €20, Polish labor costs €17, and Chinese factories can combine much cheaper labor with equally sophisticated batteries, electronics and increasingly strong engineering, what exactly is Europe charging the premium for?

There are good answers.

Europe still has exceptional brands, engineering, machinery, industrial know-how and research capabilities.

But the answer can no longer simply be:

because German industry is technologically superior.

That assumption is what China has spent the last twenty years attacking.

Volkswagen's proposed restructuring is therefore more than a corporate layoff story.

It may be an early sign that Europe is finally being forced to answer a question it could postpone during the golden years of globalization:

How much mass manufacturing can a high-wage, low-hours, aging continent sustain once its low-cost neighbors are no longer very cheap and its largest former customer has become its largest industrial competitor?

A 100,000-job restructuring at Volkswagen would be one answer.

Probably not the last one.

Why some of the people best equipped to build the next world can look strangely unimpressive while the current one is still winning.

There is a familiar way to become elite.

Pick a difficult field. Enter a selective institution. Survive increasingly competitive filters. Accumulate responsibility. Allow the institution to certify your progress.

The sequence might be:

analyst → manager → partner

or:

engineer → senior engineer → staff → principal

or:

PhD → research scientist → lab leader

By the time someone reaches 35 or 40, the record is easy to read.

The institution has done the accounting.

It tells the world:

This person passed our filters.
This person became more valuable.
This person belongs here.

There is another kind of person whose development is much harder to see.

He may start companies, learn software, work as an engineer, enter unfamiliar industries, move between countries, study markets, follow new technologies, abandon projects that stop making sense and repeatedly question which game is worth playing in the first place.

From the outside, this can look much worse.

The conventional professional appears to be compounding.

The other person appears to keep starting over.

For much of my adult life, I assumed the difference was simple:

they specialized; I explored.

I no longer think that is the right distinction.

The deeper difference is that one person's exploration happens inside a system that knows how to credential it.

The other's may not.

And that matters because institutions do much more than train people.

They make accumulation visible.

Institutions explore too

The conventional specialist is often not narrow at all.

A corporate lawyer might work across M&A, governance, employment, regulation and international transactions.

A management consultant can move from fisheries to housing to pharmaceuticals to banking to energy.

An engineer at a frontier technology company might work on infrastructure, models, robotics, hardware and product.

These are broad intellectual careers.

The difference is that their breadth takes place inside a stable coordinate system.

The lawyer continues practicing law.

The consultant continues solving business problems.

The researcher continues operating inside a recognized scientific frontier.

A large institution effectively makes an extraordinary offer:

Explore widely. We will preserve the continuity.

The McKinsey consultant who moves from airlines to semiconductors does not become an amateur every six months.

Every project adds to the same ledger:

experience, compensation, colleagues, clients, promotion, reputation and the McKinsey credential itself.

The researcher can radically change research questions while retaining the laboratory, peer network, publication system, equipment and professional identity underneath him.

The lawyer moves among cases while accumulating years at the firm.

Institutions subsidize exploration.

More importantly, they certify that the exploration counted.

Independent careers have no equivalent accounting system.

If I build one company, teach myself a new technology, enter another market and later decide that a more important technological transition is happening somewhere else, much of the previous accumulation may become illegible.

I might not actually have returned to zero.

But there is no institution standing behind me telling everyone that I have advanced from Level 6 to Level 7.

To an outsider, the trajectory may simply look incoherent.

That difference is easy to mistake for a difference in ability.

There are two levels at which a person can explore

The usual distinction between “generalist” and “specialist” misses something important.

A more useful distinction is between exploring within a game and exploring which game should exist.

The first is within-paradigm exploration.

The lawyer asks:

Which legal problem should I solve?

The consultant asks:

Which business problem should I solve?

The machine-learning researcher asks:

Which problem on the frontier of machine learning should I attack?

These can be extraordinarily difficult questions.

But the basic coordinate system has already been supplied.

Paradigm exploration asks something more destabilizing:

Which game is worth playing at all?

Should the important company be in software, media, robotics, finance or manufacturing?

Is value going to accrue at the model layer, the application layer, infrastructure or some layer that does not yet have a name?

Should an organization employ 10,000 people or 500 people plus autonomous systems?

Is an existing industry worth optimizing, or is the important opportunity to replace its architecture entirely?

Should I optimize for an established domestic market or build something global from the beginning?

Those questions are much harder to credential.

There may be no exam.

There may be no promotion.

There may not even be an accepted vocabulary for describing the problem yet.

Sometimes this kind of thinking produces a new industry.

Sometimes it produces a smart person who spends 15 years being interested in things.

At year ten, those two people can look remarkably similar.

That uncertainty is real.

The elite are not always legible before they win

There is a powerful hindsight effect in how we tell stories about consequential people.

Once someone succeeds, the past reorganizes itself.

The failed project becomes “an important learning experience.”

The strange job becomes “where he acquired a crucial skill.”

The unrelated interest becomes “the source of an interdisciplinary breakthrough.”

The years of ambiguity become “preparation.”

If the same person never produces the later achievement, those experiences receive different names:

lack of focus, wandering, unfinished projects, inability to commit.

History edits backward.

This makes successful unconventional careers look much more intentional than they felt while they were happening.

Henry Ford is an obvious example.

Before Ford Motor Company, he had been a machinist's apprentice, repaired steam engines, worked as an engineer for Edison, experimented with vehicles and participated in two earlier automobile ventures.

Ford Motor Company was founded in 1903, when Ford was 39.

At 25, his trajectory did not yet explain itself.

Later, it did.

The same is true more generally of entrepreneurship.

Technology culture has trained us to imagine that exceptional founders reveal themselves very early:

a brilliant teenager programs;

a 21-year-old drops out;

a 24-year-old raises venture capital;

a 28-year-old becomes a billionaire.

Those people exist.

But they are not the only model.

A large study using U.S. Census data found that the average founder age among the fastest-growing 0.1 percent of new ventures was 45. Previous experience in the relevant industry was also strongly associated with success.

That does not mean wandering until middle age is a strategy.

It means the market for consequential talent is not identical to the market for youthful legibility.

Breadth is not the same thing as shallowness

There is an obvious objection to all of this.

Perhaps the unconventional person is not secretly developing some rare capability.

Perhaps he simply never stayed with anything long enough to become good.

That happens.

There is nothing inherently admirable about variety.

A person can spend decades collecting ideas without accumulating mastery.

Breadth becomes valuable only when different experiences begin to change one another.

Software changes how you understand organizations.

Economics changes how you understand software markets.

Entrepreneurship changes how you distinguish elegant ideas from things customers will actually pay for.

Living in different countries changes how you perceive institutions and cultural assumptions.

Manufacturing changes how you think about physical constraints.

AI changes what you believe organizations themselves can become.

Eventually, several previously separate models may begin to interact.

That is not the same as knowing a little about everything.

It is combinatorial depth.

Research on innovation points in this direction.

Generalists appear particularly useful in uncertain environments because they can recombine knowledge across components, while specialists often contribute deeper optimization inside particular components.

Studies of repeat inventors also find that many move among technological domains rather than remaining permanently fixed in one. But they usually do not jump randomly across the entire intellectual universe. They tend to move into areas connected to things they already know.

And there is a penalty for jumping too far.

Scientists and inventors who move a great intellectual distance from their accumulated expertise tend, on average, to perform worse after the transition.

That is important.

The lesson is not:

Follow every curiosity.

It is:

Build a sufficiently large map that you can see connections unavailable from one coordinate—and accumulate enough depth that the connections survive contact with reality.

Think like a fox. Build like a hedgehog.

Isaiah Berlin made famous the distinction between the fox and the hedgehog.

The hedgehog knows one big thing.

The fox knows many things.

Philip Tetlock later found something resembling this distinction in forecasting. Thinkers who combined multiple perspectives, updated probabilities and resisted forcing reality into one grand theory often forecast better than highly ideological “hedgehog” thinkers.

Nate Silver later popularized the distinction in The Signal and the Noise.

But there is a mistake hidden in the metaphor.

A person can be a fox epistemically without being a fox economically.

I can want economics, engineering, history, politics, technology, psychology and manufacturing in my head while still building one company.

I can collect models broadly while allocating capital narrowly.

I can change my mind easily about theories while being extraordinarily persistent about an objective.

In fact, this may be one of the most powerful combinations:

Think like a fox. Build like a hedgehog.

Take information from everywhere.

Commit resources somewhere.

This distinction matters because the highest form of breadth is not permanent movement.

It is synthesis.

Institutions are not neutral containers

There is another reason I distrust the assumption that the conventional career is “safe” while the unconventional one is risky.

Specialization is itself a bet.

If you spend 15 years inside an institution, you are making a concentrated investment in the proposition that:

this institution, this industry and this definition of valuable expertise will continue to matter.

Often that is an excellent bet.

Sometimes it becomes stranded.

The people trapped inside declining industries are rarely stupid.

Quite often they are exceptionally intelligent people who became extraordinarily good at solving the problems rewarded by the previous technological regime.

That distinction matters.

An institution does not merely teach you skills.

It teaches you:

what problems matter;

what competence looks like;

who deserves respect;

what risks are reasonable;

what evidence counts;

what career moves constitute progress;

and what sort of future is plausible.

The same coordinate system that makes exploration cheap also places boundaries around what is likely to be explored.

The McKinsey consultant can explore twenty industries.

But she is much less likely to spend ten years asking:

Why should companies buy management consulting in this form at all?

The frontier researcher may explore enormous intellectual territory while taking for granted that the frontier represented by his institution is the frontier that matters.

The elite institution gives its members an enormously powerful map.

That map is an advantage until the terrain changes.

Then it can become a source of correlated error.

The world occasionally changes the scoring system

This is where the argument becomes particularly relevant now.

Artificial intelligence is not merely another productivity tool.

It may alter the relative scarcity of different forms of cognition.

For decades, advanced economies placed enormous premiums on people capable of absorbing information, manipulating abstractions, analyzing organizations, producing recommendations and coordinating complex knowledge work.

That environment rewarded lawyers, consultants, financiers, managers, software engineers, researchers and other highly trained cognitive specialists.

AI does not make these people irrelevant.

But it may change what is scarce.

If machines become increasingly capable of operating inside established bodies of knowledge, then the premium on human beings may shift toward questions such as:

Which body of knowledge should we be using?

Which assumptions no longer hold?

Which two industries are about to collide?

Which capability has become newly possible?

Which organizational form makes sense now?

What should we build that did not make economic sense five years ago?

Those are not necessarily specialist questions.

They are often questions about changing the level of analysis.

The specialist naturally asks:

How can we make this system better?

The person moving across domains is more likely to ask:

Why does the system have this form at all?

This is both his advantage and his greatest source of stupidity.

Sometimes the outsider sees a constraint insiders have stopped noticing.

Sometimes he simply does not understand why the constraint exists.

Boundary-crossing innovation therefore carries both higher upside and higher uncertainty.

That is exactly what we should expect.

The conventional elite optimize within maps

There is nothing illegitimate about this.

Every civilization needs people who become extraordinarily good at operating its most important systems.

Elite institutions select talented people, expose them to difficult problems, surround them with ambitious peers and give them increasingly consequential responsibilities.

That produces real competence.

But there is another kind of elite whose value becomes most obvious during periods when the map itself is unstable.

Call them the uncredentialed elite.

The phrase does not mean people without university degrees.

Many will have degrees.

Some will have worked at elite institutions.

“Uncredentialed” means something more specific:

Their most valuable capability cannot be fully certified by the institutions that currently exist because part of that capability consists of seeing beyond the categories those institutions use.

They may combine domains that normally produce separate professions.

They may enter industries through side doors.

They may be unusually willing to rebuild their model when reality contradicts the accepted one.

They may have high agency because nobody has provided a predefined sequence of steps.

Their advantage is not that they avoided institutions.

It is that their identity was never completely enclosed by one.

And this creates a strange economic problem.

Before they produce something undeniable, it can be difficult to distinguish them from dilettantes.

There is no standardized credential for:

saw the next industrial architecture before it became obvious.

There is no promotion ladder for:

combined five previously unrelated domains into a new company.

There is no exam for:

correctly identified that the old prestige hierarchy was attached to a declining technological regime.

The credential arrives afterward.

It is the thing they built.

Exploration has no automatic accounting system

This is perhaps the largest psychological advantage of institutional careers.

They continually resolve ambiguity.

You make associate.

You get promoted.

You publish the paper.

You become partner.

Your compensation rises.

The milestones certify that the previous years counted.

The unconventional path provides much weaker reassurance.

It gives you options rather than promotions.

You can become substantially more capable while looking almost unchanged from the outside.

That makes the path psychologically difficult.

It also makes self-deception unusually easy.

Perhaps you really are building an unusual combination of capabilities.

Perhaps you simply enjoy novelty and have constructed an elaborate intellectual justification for avoiding commitment.

There is no perfect test.

That uncertainty cannot be solved philosophically.

Eventually reality has to arbitrate.

Customers.

Machines.

Capital.

Scientific results.

Organizations.

Revenue.

Products.

People willing to follow you.

Things that work.

The uncredentialed elite cannot remain permanently uncredentialed.

At some point, capability must become consequence.

The goal is not to keep exploring

This is where I disagree with the romantic version of the generalist argument.

The objective is not maximum breadth.

It is not a life spent collecting experiences.

It is not permanent optionality.

The objective is convergence.

At first, the pieces may look unrelated.

Software.

Economics.

Technology.

Entrepreneurship.

Different countries.

Different industries.

Failures.

Relationships.

New ideas.

Then a sufficiently important problem appears and several of those pieces suddenly become relevant at once.

The path begins to collapse inward.

Not because the earlier breadth was a mistake.

Because it has finally found something worth concentrating on.

The strongest model I can come up with is:

Explore broadly enough to discover something non-obvious.
Learn deeply enough to know when you are wrong.
Concentrate enough resources to make the insight real.

All three are necessary.

Cross-domain thinking without reality becomes intellectual entertainment.

Depth without the ability to question the surrounding paradigm can produce extraordinary optimization of a world that is disappearing.

And exploration without eventual commitment becomes permanent adolescence.

The rare combination is different:

broad search + deep learning + concentrated execution.

The credential is the new reality

Perhaps this explains why some consequential people emerge later than expected.

Their advantage was never simply that they spent more years climbing one ladder.

They accumulated a combination of models, capabilities, failures, relationships and experiences that could not easily have been designed in advance.

For years the combination looked inefficient.

Then the world changed.

A technology appeared.

An industry reorganized.

A previously impossible company became possible.

A problem emerged for which the strange combination suddenly made sense.

And somebody who had looked less optimized than his peers found himself unusually well positioned.

Before the opportunity appeared, the career looked incoherent.

Afterward, everyone called it strategy.

That is one of the peculiarities of genuinely nonlinear careers.

The conventional elite receive their credentials before the outcome.

The uncredentialed elite receive theirs afterward.

Their credential is the institution they created.

The technology they built.

The company that changed an industry.

The system that other people subsequently learn to navigate.

Until then, the market may discount them.

Sometimes correctly.

Sometimes very badly.

And during periods when the old maps stop working, that distinction can become enormously important.

Because the next elite may not initially look like a better version of the current elite.

They may look like people who were playing the wrong game.

Right up until the game changes.

Imagine I offered you an investment with the following terms.

There is a substantial chance that everything works normally for decades.

But there is also a meaningful chance that, sometime during your working life, the system suffers a catastrophic failure.

In that state, you could lose most of your locally held wealth. Your income could disappear. Your property could become inaccessible or worthless. Crime and political violence could rise dramatically. Your ability to move money or leave the country could become restricted.

In the worst version, there is a non-trivial risk of injury or death.

Would you put essentially your entire net worth into it?

Probably not.

Yet this is approximately how most people construct their lives.

Their home, citizenship, career, bank accounts, pension, property, professional network, social network and legal rights all sit inside one country.

We diversify stocks because putting everything into a single company feels obviously reckless.

Then we put almost our entire life portfolio inside one jurisdiction.

The usual response is:

Sure, but developed countries don't collapse.

I used to find that intuitively persuasive.

Then I started looking at the time scale.

Eighty peaceful years feels normal because it is almost all we remember

Western Europe and Japan have now lived through roughly 80 years without another catastrophe comparable to the first half of the twentieth century.

For people alive today, that begins to feel like the natural state of developed societies.

Our parents lived through it.

Our grandparents often lived through most of it.

Institutions survived.

Elections happened.

Markets grew.

Wars happened somewhere else.

That experience creates a powerful intuition:

The country has been stable for my entire life, therefore stability is probably the normal state of the country.

But that conclusion becomes much less obvious when the event you are trying to estimate may occur only once every century or two.

If a catastrophic political event naturally happens on a 100-, 150- or 200-year time scale, then observing 30 or 50 peaceful years tells you almost nothing.

Even 80 years may not tell you very much.

The postwar period could represent a structural break in human history.

Or it could simply be a very long peaceful interval.

Those are radically different interpretations, and we do not yet have enough history to distinguish them confidently.

That is the problem.

So I tried a different question

Instead of asking:

Does America look unstable today?

Or:

Are European institutions healthy?

I wanted a more mechanical question:

Historically, how long have countries broadly similar to today's developed countries actually gone between serious conflicts on their own territory?

I took a rough reference group:

• United States
• Canada
• United Kingdom
• France
• Netherlands
• Belgium
• Switzerland
• Sweden
• Denmark
• Norway
• Germany
• Italy
• Spain
• Portugal
• Japan
• Australia
• New Zealand

Then I looked backward roughly 250 years.

The event definition was deliberately simple:

Organized war-level conflict on what is now the country's territory, involving the government or a rival political authority, and reaching roughly the conventional 1,000-battle-death threshold.

This is not an actuarial model.

Borders change. Definitions change. Some countries have unusual histories. A conflict in a remote colony is not psychologically equivalent to tanks entering the capital.

But I was not trying to calculate a probability to four decimal places.

I wanted to know whether the historically sensible baseline looked more like:

1%

10%

or

50%

over a human-scale horizon.

The result surprised me.

The historical peace spells were much shorter than intuition suggests

In the first-pass sample, the historical survival curve looked roughly like this:

Again, these are rough historical reference-class numbers, not forecasts for any particular country.

But they change the prior.

Only about 31% of historical peace spells in this sample reached 100 years.

An 80-year stretch of peace is already unusual.

The current postwar period is not historically ordinary.

That does not mean another conflict is "due."

It means something more important:

We should not casually assume that another 50 peaceful years is the default simply because the last 80 were peaceful.

A baseline around a 50% chance of a catastrophic domestic conflict over the next 50 years no longer strikes me as obviously extreme.

It may be wrong.

But it is at least in the range that the long-run historical record forces us to take seriously.

The working conflict table

The table below is essentially what sat behind the earlier calculation. A † means a pre-1816 episode that had to be manually added rather than coming from COW.

COW's chronology does indeed contain the post-1816 backbone we were using: French insurrections, Belgian independence, Carlist wars, the Italian unification conflicts, U.S. Civil War, Schleswig wars, Meiji Restoration, Satsuma Rebellion, Australian Aboriginal wars, WWI, the Spanish Civil War and WWII.

What I actually did with those rows

The important thing is that I did not count "number of wars per country."

I turned the history into peace spells.

Take the United States as the easiest example:

That final 80-year observation is fundamentally different from the first three.

We know the 29-, 46- and 76-year spells ended.

We do not know how long the current one will ultimately last.

So simply averaging all the numbers would bias the result downward by pretending today's unfinished spells ended in 2025.

That's why I used a survival-analysis concept.

Kaplan-Meier, in plain English

At every duration where one of the historical peace spells ended, you ask:

Among all peace spells that had survived this long, what fraction failed here?

You compound those survival fractions as duration increases.

The resulting function, usually written S(t), just means:

What fraction of peace spells would historically be expected to still be alive after t years?

Then:

1 minus that number = probability that the spell has ended in a qualifying conflict by that duration.

The advantage is that Switzerland's current 200+ year spell, America's current spell, Europe's post-1945 spell, etc., still contribute information without us pretending we know when they will end.

I rebuilt the arithmetic explicitly just now

Using the broad working coding above, merging overlapping episodes, I get 48 completed peace spells plus 17 currently censored spells.

So my earlier statement of 47 completed should be treated as a first-pass counting discrepancy, not a sacred number.

The completed spells alone look approximately like this:

And the reconstructed Kaplan-Meier curve gives approximately:

The obvious objection: today's countries are different

This is where most geopolitical analysis moves immediately.

America is rich.

Western Europe is democratic.

Japan has strong institutions.

NATO exists.

Nuclear deterrence exists.

Modern states have enormous administrative capacity.

All true.

And all relevant.

"Strong institutions" is not an operational explanation

When someone tells me:

America has strong institutions.

My brain wants to ask:

Strong against what exact failure mode?

What happens if an election result is not accepted?

What happens if political legitimacy splits across institutions?

What happens if a severe recession coincides with a military confrontation?

What happens if different levels of government refuse to cooperate?

What happens if a financial shock, technological shock and constitutional crisis arrive together?

What happens if today's political coalition is replaced by one that does not respect the same informal constraints?

These are not predictions.

They are stress tests.

In software, nobody serious says:

The architecture looks healthy, therefore failure probability is negligible.

You ask:

• Where are the single points of failure?
• Which systems fail together?
• What happens under load?
• What assumptions are hidden?
• Which dependencies are outside your control?
• What happens when the happy path breaks?

That is how I increasingly think about countries too.

A policy is not capability

This is also where my operator perspective differs from a lot of financial or geopolitical commentary.

A financial analyst may look at:

• capital spending;
• market size;
• valuation;
• supply chains;
• corporate margins.

A geopolitical analyst may look at:

• alliances;
• military budgets;
• elections;
• treaties;
• state capacity.

Those are important.

But building technology teaches you to ask one additional question:

What actually has to happen for this outcome to become real?

That question matters enormously.

A country can announce industrial policy without building industry.

A company can announce an AI strategy without changing a single employee workflow.

A military can spend enormous amounts without building a functioning procurement system.

A government can have emergency plans that nobody knows how to execute.

A country can look institutionally stable until several supposedly independent systems fail at the same time.

The interesting variable is not only what exists on paper.

It is whether the underlying system can execute under stress.

That is much harder to observe from the outside.

And that creates another problem: political fragility is partially hidden

If a country's true probability of breakdown were obvious three years in advance, country risk would be relatively easy to manage.

People would leave.

Capital would move.

Governments would respond.

Allies would intervene.

Political actors would change strategy.

The crisis itself might be prevented.

The fact that many catastrophic political events are not widely anticipated is therefore not surprising.

The system reacts to its own prediction.

And much of the relevant state is hidden.

You can observe elections.

You can observe GDP.

You can observe institutions.

But you cannot directly observe:

• how political elites behave under an unprecedented shock;
• whether the military remains unified;
• how much legitimacy institutions actually retain;
• how quickly social trust can collapse;
• which event suddenly coordinates previously fragmented opposition;
• which apparently stable equilibrium is close to a tipping point.

This looks familiar to anyone who has built complex systems.

The worst failures often do not come from the variable you were watching.

They come from an interaction between several variables you thought were independent.

So I don't want to predict the crisis

I want to remove the single points of failure

This is the part where the startup method becomes useful.

When building a product, I don't start by trying to predict every possible user behavior.

I ask:

What is the cheapest experiment that reveals the biggest unknown?

And:

What is the minimum architecture that prevents one failure from destroying the whole system?

Applied to country risk, the question becomes:

What are the minimum changes that prevent one country's failure from destroying most of my life?

That is a much more useful question than:

Will America have a civil war in 2057?

I have no idea.

Neither does anyone else.

But I can inspect my own architecture.

Run the failure analysis on your own life

Imagine your country enters a catastrophic political or military crisis.

Don't discuss it abstractly.

Walk through the system.

Income

Can you still earn if your domestic economy stops functioning normally?

Can you work legally elsewhere?

Is your earning power portable, or is it dependent on local licensing, customers or institutions?

Money

Where are your assets legally held?

Not what companies do you own.

Where is the account?

Which country's courts govern it?

What currency do you depend on?

What happens if capital controls appear?

Housing

How much of your net worth is concentrated in one physical property market?

If you suddenly needed to leave, could you sell?

Would there be buyers?

Legal mobility

Where can you actually live without asking permission?

Not:

I could probably get a visa.

But:

I have a legal right to enter, stay and work.

Network

If you moved tomorrow, would you know anyone?

Could you find work?

Could your family function there?

Do you understand the institutions?

Timing

How much of this can be done after the crisis becomes obvious?

That last question may be the most important one.

The option becomes valuable before everyone knows they need it

Suppose country risk becomes obvious.

Suddenly millions of people want the same things:

• foreign residency;
• foreign bank accounts;
• foreign currency;
• liquid assets;
• flights;
• housing abroad;
• safe jurisdictions.

The price of escape rises precisely when you need it.

Some escape routes simply disappear.

This is why waiting for obvious evidence may be irrational.

The value of the option comes from possessing it before the state of the world is revealed.

That is the same logic behind redundancy in engineering.

You do not install the backup generator after the electricity goes out.

Is an 80% downside realistic?

Not for the average war.

That distinction matters.

A normal recession is not an 80% loss.

Even many wars do not destroy 80% of national wealth.

But I am not interested in the average bad year.

I am interested in the catastrophic branch.

The branch where several things happen together:

• physical conflict;
• currency destruction;
• expropriation;
• capital controls;
• forced displacement;
• loss of income;
• property destruction;
• collapse in public safety.

If most of your economic life sits inside the affected country, an 80% personal loss is not difficult to imagine.

And financial wealth is only part of the damage.

There is also:

• physical injury;
• death risk;
• family disruption;
• lost years of education;
• destroyed careers;
• psychological trauma;
• political humiliation;
• loss of national status;
• decades of reduced opportunity after defeat.

Countries can rebuild GDP.

People do not necessarily get those years back.

Now the expected-value problem becomes uncomfortable

Suppose your rough long-horizon probability of a catastrophic country event is around 50%.

Suppose the catastrophic personal downside is around 80%.

I am not claiming those are exact numbers.

Use your own.

The important point is simply that if both numbers are large, then the expected exposure is large.

Yet many intelligent people devote almost no effort to reducing it.

Operationally, their life looks like:

• one citizenship;
• one country of residence;
• one property market;
• one banking system;
• one currency;
• one local career;
• one domestic professional network;
• one pension regime.

Then they buy an international ETF and call themselves diversified.

Their securities portfolio may be diversified.

Their life architecture is not.

The correct response is not a bunker

It is redundancy

This distinction matters.

If I believed the rational response was spending every Sunday stockpiling canned food, I would not find the idea particularly interesting.

The better strategy looks much more like designing a fault-tolerant system.

Build a second legal option.

Have assets outside one jurisdiction.

Develop income that can travel.

Know another city well enough that moving there is operationally possible.

Maintain relationships outside your home country.

Keep enough liquidity that you do not need to sell everything during the panic.

Understand which decisions become impossible once a crisis begins.

The objective is not to live as though catastrophe is imminent.

The objective is to make sure catastrophe in one location does not automatically become catastrophe everywhere in your life.

The best hedges are productive in the normal world too

This is where the economics become especially attractive.

A second-country professional network can create business opportunities even if nothing goes wrong.

A second language can increase your earnings.

International assets can improve ordinary portfolio diversification.

A second citizenship can create employment, travel and educational options.

Portable skills increase negotiating leverage.

Living in multiple countries improves information.

So these are not necessarily dead insurance expenditures.

They can produce upside in the normal world and become extraordinarily valuable in the catastrophic world.

That is exactly the kind of option I like.

Low carrying cost.

Useful under ordinary conditions.

Massive payoff in a rare bad state.

This is really an engineering problem

The more I think about country risk, the less useful I find the usual political question:

Is my country safe?

"Safe" is too binary.

A better question is:

How many independent things have to go wrong before my life becomes unrecoverable?

If the answer is one, you have a fragile system.

If one government, one banking system, one property market or one legal jurisdiction can simultaneously destroy most of your wealth, income and mobility, then you have an architectural problem.

In software, I would never design an important system that way.

I am increasingly unconvinced that I should design my life that way either.

The postwar world may continue for another century

I hope it does.

Maybe modern democracy permanently reduced internal war.

Maybe nuclear weapons permanently reduced great-power conflict.

Maybe economic integration changed the process.

Maybe the next 80 years look just like the last 80.

But the historical record does not give me enough confidence to treat that outcome as virtually guaranteed.

That is the important distinction.

You do not need to predict catastrophe.

You do not need to believe catastrophe is the most likely outcome.

You only need to believe that:

the probability is meaningful,

the downside is enormous,

and

the cost of reducing your exposure is reasonable.

Once those three conditions hold, doing nothing becomes the strange decision.

Don't spend your time arguing endlessly about whether the system will fail.

Identify the failure modes.

Find the single points of failure.

Test your assumptions.

Build redundancy where it is cheap.

Then keep operating.

Because country diversification is not fundamentally about pessimism.

It is about refusing to let one system determine every important outcome in your life.

The most consequential thing Japan did to the Soviet Union in the 1980s was not military.

Japan did not threaten Moscow. It did not win an arms race against the USSR. It did not force the Soviet Union out of Eastern Europe.

What Japan did was more psychologically destructive:

it made the Soviet model look obsolete.

By the 1970s, Soviet leaders could still describe their country as one of the world's great modern industrial civilizations. The USSR possessed nuclear weapons, spacecraft, advanced aircraft, enormous steel and machinery industries, world-class mathematics and physics, and one of the two superpower militaries.

Its economic failures could still be rationalized.

The United States was richer, but Soviet citizens were told that America achieved its wealth through capitalism, inequality, unemployment, consumerism and exploitation.

The USSR was supposedly pursuing different goals.

Then Japan became impossible to ignore.

Here was a country that had been physically devastated in 1945, possessed few natural resources, maintained a relatively small military, and had a fraction of Soviet territory.

Yet by the 1980s Japan was producing some of the world's best:

• automobiles;
• semiconductors;
• consumer electronics;
• industrial robots;
• machine tools;
• ships;
• cameras;
• precision components;
• public infrastructure.

Japan became the world's largest automobile producer around 1980. By the late 1980s Japanese firms controlled roughly half of the global semiconductor market. Japanese nominal GDP rose from roughly $1.1 trillion in 1980 to more than $3 trillion by 1990.

The comparison became politically poisonous because Japan did not merely outperform the USSR.

Japan invalidated the Soviet explanation for why the USSR underperformed.

If the Soviet model really represented industrial modernity, why could Japan build better cars?

Why did Japan have better electronics?

Why were its factories more productive?

Why were its trains better?

Why were ordinary Japanese households surrounded by technologies Soviet engineers could not provide their own population?

Eventually the comparison stopped being:

Japan is doing unusually well.

It became:

Something is fundamentally wrong with us.

That shift mattered.

Once Soviet citizens and policymakers began believing the problem was not an isolated policy mistake but the system itself, reform became destabilizing rather than restorative.

Gorbachev did not create Soviet economic doubt. He inherited it.

And when the political system finally cracked, Soviet economic prestige did not merely decline gradually.

It collapsed.

Russia inherited nuclear weapons, aerospace expertise, world-class scientists, enormous energy reserves, sophisticated military industries and a permanent UN Security Council seat.

But economically, the old status was gone.

Nobody in 2000 thought:

United States, Japan, Germany, Russia — four roughly equivalent centers of advanced economic power.

Russia had retained extraordinary capabilities.

It had lost the presumption of systemic competence.

That loss proved remarkably permanent.

China may be playing the same role for Europe

This is the analogy I think matters for the 2020s.

The conventional way to describe China's rise relative to Europe is quantitative:

China's economy is getting larger.

China manufactures more.

China exports more goods.

But that misses the stage we are now entering.

China already won the quantitative manufacturing contest years ago.

By 2024, China accounted for about 32% of global manufacturing value added.

The United States was around 15%.

Japan roughly 6%.

Germany around 5%.

South Korea around 3%.

China alone therefore produced more manufacturing value added than the United States, Japan, Germany and South Korea combined.

That is not the important new development of the 2020s.

Europe had already learned to live psychologically with Chinese scale.

The comforting explanation was straightforward:

China has quantity. Europe has quality.

China had factories.

Europe had engineering.

China had cheap labor.

Europe had technology.

China assembled.

Europe designed.

China competed on cost.

Europe competed on sophistication.

That story allowed Europe to preserve its status even after the quantitative balance had shifted decisively toward China.

The dangerous development of the 2020s is that China is beginning to remove the qualitative escape hatch as well.

The sequence looks something like this

That third stage is the important one.

It is the equivalent of Japan's 1980s.

Cars are especially dangerous because Germany built part of its identity around them

China manufacturing more toys than Europe never threatened European prestige.

Chinese companies becoming competitive with Germany in automobiles does.

Germany's implicit industrial proposition for decades was not:

We make a lot of cars.

It was:

We know how to make the world's best cars.

Mercedes.

BMW.

Porsche.

Volkswagen.

Audi.

The automobile was an unusually visible demonstration of European engineering competence.

Then the technological basis of the car changed.

Electric vehicles placed greater importance on:

• batteries;
• power electronics;
• software;
• electric motors;
• charging;
• electronics supply chains;
• manufacturing integration.

And China built the ecosystem.

In 2025 China produced roughly three-quarters of the world's electric cars.

It produced more than 80% of global battery cells.

Chinese companies also dominated cathode and anode materials, and Chinese battery producers supplied more than half of the EU's own EV battery market.

That is not simply another case of China making a cheaper European product.

The embarrassing possibility is that China has become better positioned for the technological regime that replaces Europe's most prestigious industrial product.

Europe can tariff Chinese EVs.

But protection itself communicates something.

Twenty years ago, European companies wanted China to open its market to superior European industrial products.

Now Europe increasingly worries about protecting its own market from Chinese manufactured goods.

The direction of anxiety has reversed.

Green technology creates an even deeper legitimacy problem

Europe spent decades defining itself partly through environmental leadership.

Climate regulation became one of the clearest distinctions between the European and American models.

The implicit European argument was:

America may have Silicon Valley, but Europe is building the sustainable society of the future.

Then something unexpected happened.

Europe became one of the world's strongest advocates of the green transition.

China became its industrializer.

In 2025 China added almost 500 gigawatts of renewable power capacity in a single year.

More than 60% of all renewable capacity installed globally that year was Chinese.

That included roughly:

• 370 GW of solar;
• 117 GW of wind.

The entire European Union added about 85 GW.

China had already exceeded its own 2030 wind-and-solar target in 2024, six years early.

And China does not merely install clean energy.

It manufactures much of the physical equipment required for the global transition:

• solar modules;
• batteries;
• EVs;
• power electronics;
• increasingly grid equipment;
• much of the relevant processing capacity.

This produces an ideologically awkward result.

Europe can still say:

We cared about climate first.

China can increasingly answer:

We built the system.

That is a more damaging comparison than losing market share in an ordinary industry.

Green industrial policy was supposed to demonstrate the advantages of coordinated European government action.

Instead, China may demonstrate those advantages more convincingly.

Infrastructure attacks Europe on another point of pride

Europe has many excellent transportation systems.

Paris, Madrid, Vienna, Copenhagen, Amsterdam and Zurich remain among the world's best urban environments.

The important comparison is therefore not whether every Chinese train is nicer than every European train.

It is the ability to build.

At the end of 2025, China had roughly 50,000 kilometres of high-speed railway.

Its modern high-speed network was negligible less than twenty years earlier.

Chinese cities operated more than 11,000 kilometres of urban rail, and the country continues adding hundreds of kilometres each year.

Europe, meanwhile, can take years or decades to approve, fund and construct major infrastructure projects.

Each delay has a perfectly reasonable explanation:

• environmental review;
• local consultation;
• property rights;
• fiscal constraints;
• historic preservation;
• procurement requirements;
• fragmented jurisdictions.

Every explanation may individually be valid.

But political legitimacy operates on accumulated impressions.

After the tenth explanation, voters may stop hearing:

Our system embodies higher standards.

They may start hearing:

Our system cannot execute.

That distinction is politically explosive.

Science is becoming harder to use as Europe's fallback argument

For years Europe had another defense.

Perhaps America commercialized technology better.

Perhaps China manufactured it more cheaply.

But Europe still possessed extraordinary universities, laboratories and scientific institutions.

That remains true.

What is changing is the relative trajectory.

China's R&D intensity reached about 2.7% of GDP in 2024.

The EU has remained close to 2.1%.

Measured using purchasing-power parity, China's total R&D expenditure has reached approximately American scale, while European spending has fallen relative to the United States over the past decade.

Scientific output has shifted dramatically as well.

China now leads the Nature Index in natural-science research output.

In energy and environmental research, the OECD found China's share of the world's top 10% most-cited publications rose from around 15% in 2012 to nearly 40% in 2022.

The EU moved in the opposite direction:

27% to roughly 15%.

This matters because it weakens another comforting interpretation:

China manufactures Western discoveries.

Increasingly, China discovers things too.

AI may complete the psychological transition

Artificial intelligence provides perhaps the clearest example of prestige reclassification happening in real time.

A decade ago, it was plausible to imagine three technologically consequential regions:

United States — Europe — China

The contemporary AI discussion increasingly looks different:

United States — China

and then everyone else.

Stanford counted 40 notable American AI models in 2024.

China produced 15.

Europe combined produced only three.

By 2025, the numbers had risen to 59 for the United States and 35 for China.

China also leads in AI publication volume, citations and patent output, while Chinese frontier models have rapidly closed benchmark gaps with American systems.

The remarkable thing is not simply that Europe is behind.

It is that Europe is disappearing from the comparison.

People ask:

Can China catch the United States in AI?

They increasingly do not ask:

Can Europe catch China?

That grammatical change is how status loss begins.

China is also destroying the cheap-labor explanation

Perhaps China's manufacturing dominance could still be dismissed if it rested on millions of inexpensive factory workers.

But China is rapidly automating.

In 2024 Chinese factories installed approximately 295,000 industrial robots.

That was about 54% of every industrial robot installed worldwide that year.

China already had more than two million industrial robots in operation.

More importantly, Chinese robotics manufacturers captured 57% of their own domestic market, compared with less than 30% historically.

The manufacturing giant is therefore beginning to manufacture the machines that automate the manufacturing giant.

This creates a very different competition.

Europe is no longer facing:

cheap Chinese labor versus sophisticated European automation.

It is increasingly facing:

Chinese scale + Chinese automation + Chinese engineering + Chinese suppliers.

That is how a quantitative advantage becomes qualitative.

Why China is more psychologically dangerous to Europe than America

Europe has always possessed an ideological answer to the United States.

America could say:

We are richer.

Europe could answer:

We chose stronger welfare states.

America:

We built Google, Apple, Microsoft, Nvidia and Meta.

Europe:

We chose stronger regulation, public services and labor protection.

America:

Our capital markets are deeper.

Europe:

We prefer a less financialized society.

Whether one agrees with those answers is almost beside the point.

They allow Europe to interpret underperformance as a trade-off between models.

China creates a much harder comparison because China competes on many of the dimensions Europe itself claims to value.

Europe values strong government.

China has a vastly stronger state.

Europe values industrial policy.

China uses industrial policy far more aggressively.

Europe values public transportation.

China builds enormous public transportation systems.

Europe values decarbonization.

China installs renewable infrastructure at unparalleled scale.

Europe argues that the state should protect strategic industries.

China retained an enormous industrial base.

Europe argues that markets alone underinvest in long-term technology.

China mobilizes capital on a scale few European governments could contemplate.

Europe argues that scientific excellence should not depend entirely on Silicon Valley venture capital.

China increasingly produces frontier science.

So the European defense becomes much harder.

The question is no longer:

Why isn't Europe more like America?

Europe has answered that question for thirty years.

The dangerous question becomes:

Why can China execute so many of Europe's stated priorities better than Europe can?

That is the question Japan helped force on the Soviet Union.

Demographics make this far more serious than an ordinary industrial decline

Europe is encountering this comparison at exactly the wrong demographic moment.

The EU's working-age population is projected to decline substantially over the coming decades.

The European Commission expects the population aged 20–64 to fall by roughly 10% between 2025 and 2050.

At the same time, Europe's old-age dependency ratio is projected to increase from roughly 38% today to more than 55%.

In simple terms:

fewer workers will support more retirees.

That makes productivity growth much more important.

An aging society can remain extremely prosperous if:

• each worker becomes dramatically more productive;
• domestic capital earns strong returns;
• businesses invest heavily;
• technological industries expand;
• talented immigrants arrive;
• ambitious young people stay.

Europe's problem is that several of those mechanisms are simultaneously weak.

And Europe is not short of accumulated wealth.

EU households hold roughly €10 trillion in bank deposits.

The euro area remains a net creditor to the rest of the world, with a positive net international investment position of around €1.8 trillion in early 2026.

This is one of the strangest parts of the European story.

Europe has capital.

Europe does not consistently convert that capital into high-return European productive assets.

The Draghi competitiveness report estimated that Europe needs roughly €750–800 billion in additional investment every year merely to achieve its existing strategic objectives.

So the demographic problem and the industrial problem reinforce each other.

Europe needs higher productivity precisely when its productive machinery appears weakest.

This is where the 2030s become dangerous

The Soviet analogy becomes particularly interesting after the psychological break.

Soviet economic stagnation existed long before 1991.

But once people lost faith in the model, behavior changed.

When opportunities appeared, people exited.

Scientists left.

Entrepreneurs left.

Capital left.

Some of the Soviet Union's most internationally mobile human assets joined other systems.

Europe does not need a political collapse for something analogous to happen.

Europe has free movement of capital and people already.

Which means the adjustment can occur quietly.

A French machine-learning researcher joins an American AI lab.

A Dutch founder incorporates in Delaware.

A German battery engineer works for CATL.

A European startup moves its headquarters after Series B.

European pension capital buys American technology stocks.

A European manufacturer puts its next large factory in the United States because energy, subsidies and growth prospects are better.

No single decision constitutes collapse.

Every decision may be perfectly rational.

But together they can produce a dangerous feedback loop:

This is the European scenario I would worry about in the 2030s.

Not Soviet-style food shortages.

Not the dissolution of the European Union.

Not Europe becoming poor overnight.

Something more plausible:

a rich society entering a self-reinforcing cycle of declining confidence, outward capital flows, talent migration and institutional paralysis.

China has terrible demographics too

This analogy should not be taken too far.

China itself is aging extraordinarily quickly.

Its population has begun shrinking.

Its fertility rate is extremely low.

It may eventually face an even more severe aging problem than Europe while possessing substantially less wealth per person.

China therefore does not demonstrate that its model solves every problem.

It does not.

What China can demonstrate is narrower and politically more relevant:

Europe's failures in infrastructure, industrial investment, technological scaling and physical deployment are not inevitable consequences of being an advanced society with a large state.

That matters.

Europe cannot simply say:

Mature economies naturally stop building.

China can build.

Europe cannot simply say:

Environmentalism inevitably means industrial sacrifice.

China industrialized environmental technology.

Europe cannot simply say:

A large state necessarily suppresses technological ambition.

China's state is much larger and more interventionist.

And because Europe is aging, discovering that its institutions struggle to produce productivity growth makes the demographic problem much worse.

The deeper analogy

The comparison therefore isn't:

Japan destroyed the USSR.

Nor:

China will destroy Europe.

It is this:

Japan, 1980–1990

Japan became a comparison object that made Soviet economic weakness impossible to explain away.

The USSR still had enormous capabilities.

But policymakers and citizens increasingly understood that something fundamental was wrong with the system.

Once that belief spread, Soviet economic prestige collapsed remarkably quickly.

Russia retained islands of technological excellence afterward.

It never recovered the USSR's former status as one of the world's central economic models.

China, 2020–2030

China may become the comparison object that makes European institutional weakness impossible to explain away.

Europe will still possess:

• ASML;
• Airbus;
• world-class pharmaceutical companies;
• elite universities;
• excellent cities;
• sophisticated engineering firms;
• enormous accumulated wealth.

But those could increasingly look like exceptions inside a system losing relative capability, rather than evidence that Europe remains one of the world's defining productive systems.

That distinction is enormous.

The politically dangerous moment is when the explanation changes

Europe today still largely discusses its problems separately.

German industry has an energy problem.

European AI has a compute problem.

Startups have a venture-capital problem.

Infrastructure has a permitting problem.

Defense has a fragmentation problem.

Manufacturing has a Chinese-subsidy problem.

Demographics are a pension problem.

Each has its own report.

Its own ministry.

Its own policy programme.

But imagine another five years in which China continues advancing simultaneously in:

• EVs;
• batteries;
• solar;
• robotics;
• AI;
• scientific research;
• nuclear power;
• public transportation;
• shipbuilding;
• drones;
• advanced manufacturing;
• electricity infrastructure.

Eventually the public may stop believing there are twelve unrelated problems.

A much simpler interpretation becomes available:

Maybe the model itself is no longer capable of producing competitive outcomes.

That was the psychologically decisive development in the late Soviet period.

The USSR's problems did not suddenly appear in 1989.

What changed was the interpretation of those problems.

They stopped looking temporary.

They stopped looking sector-specific.

They stopped looking like the necessary cost of a superior social system.

They began looking systemic.

That is the real risk China poses to Europe.

2030 may matter more psychologically than economically

China does not need to become richer per capita than Germany.

It does not need to defeat Europe militarily.

It does not need Europeans to admire the Chinese political system.

Japan never needed Soviet citizens to become Japanese nationalists.

China merely has to make one proposition increasingly difficult to deny:

A society that Europe once considered technologically backward can now execute many of the things Europe claims to value better than Europe can.

If enough Europeans internalize that conclusion, the political consequences may be much larger than another decade of mediocre GDP growth.

Because status decline becomes permanent when people stop treating the existing hierarchy as temporary.

Japan helped push the Soviet Union across that psychological threshold during the 1980s.

The old Soviet claim to economic-modernity status never recovered.

China may be pushing Europe toward a similar threshold during the 2020s.

And if Europe's Soviet moment is the decade from 2020 to 2030, the most consequential effects may not become obvious until the decade after.

The 2020s would be the decade of recognition.

The 2030s would be the decade when capital, companies, talent and political behavior begin adjusting to what people have recognized.

Why China’s 2020s may resemble Korea’s 2010s—and why the 2030s could feel very different

There is a particular national mood that appears near the end of a long ascent.

The country is no longer merely getting richer. Its success starts being reflected back at it by foreigners.

Its companies are no longer dismissed as imitators. Foreign competitors study them. Its cities become places ambitious people want to see. Its technology becomes associated with the future. Its diaspora starts wondering whether leaving was necessarily the upward move. Even cultural products that once seemed provincial begin traveling.

For a while, almost every international comparison seems to move in the same direction.

South Korea had something close to that moment in the 2010s.

China may be having it now.

The analogy I want to make is not that China will become another Korea. China is nearly eleven times Korea’s population, a nuclear power, a permanent member of the UN Security Council, and the world’s second-largest economy at market exchange rates.

The analogy is about the mood around peak relative ascent:

The interesting part is what happens after the mood turns.

Because Korea’s experience suggests that the post-peak period does not begin when factories close or culture disappears.

It begins when success stops hiding dependence.

Korea in the 2010s: 3% growth felt much bigger than 3%

South Korea was already a rich country by 2010. It was not experiencing Chinese-style catch-up growth.

Yet between 2009 and 2019, Korean real GDP still grew by an average of 3.1% a year, versus just 1.8% across the OECD. Korea was mature enough to be taken seriously but still growing quickly enough to keep closing gaps with richer countries.

That was an unusually satisfying position.

Samsung overtook old Japanese electronics champions and became one of the two defining smartphone companies of the era. Hyundai and Kia moved from discount brands toward serious competitors to Toyota and Volkswagen. Korean shipbuilders had displaced Japan. Korean displays and memory chips sat at the technological frontier.

And Seoul itself carried the message.

An American or European arriving in the early 2010s could encounter faster broadband, better mobile infrastructure, enormous subway systems, newer urban districts and Samsung devices everywhere. Korea did not need to explain that it had modernized. You could see it.

Then culture arrived.

Gangnam Style in 2012.

BTS and Blackpink.

K-beauty.

Korean television.

Parasite.

Then, just over the boundary of the decade, Squid Game.

None of these things individually created Korea’s national confidence. What mattered was that industrial, technological and cultural evidence all pointed in the same direction at once.

Japan, Korea’s century-long benchmark, appeared stagnant.

China was gigantic but still clearly below Korea in many high-status technologies.

Europe was struggling through the euro crisis.

America was still recovering from 2008.

Korea kept gaining positions.

That was the essential feeling.

A young Korean could complain about jobs, housing or Korean society and still believe something quite different about the country itself:

Korea will matter more ten years from now than it does today.

That is national ascent.

Korea’s 2020s are the post-miracle X-ray

Now compare that with Korea today.

The country has not collapsed. In fact, 2026 is an unusually strong year because of the AI semiconductor boom.

But the underlying slope has changed dramatically.

Growth was just 1.0% in 2025. The OECD expects the semiconductor cycle to push it temporarily to 2.6% in 2026, then back to 1.9% in 2027. Compare that with the 3.1% average of the pre-pandemic decade.

Korea’s nominal GDP in 2025 was about $1.87 trillion, placing it roughly 13th in the world and accounting for only about 1.6% of global GDP. That is a very successful country. It is not a system-sized economy.

And once growth stops doing the psychological work, that small scale becomes harder to ignore.

The industrial reversal is particularly stark.

Korean battery companies looked like obvious winners of the EV transition only a few years ago. Yet their share of the European EV-battery market fell from nearly 80% in 2022 to 60% in 2024, largely because Chinese manufacturers were gaining. By 2025, Chinese battery producers had moved above 50% of the European market.

Shipbuilding tells a similar story. Korea remains exceptionally strong in LNG carriers and other sophisticated vessels. But China now has roughly 45% of global shipyard capacity and about 60% of the global orderbook. UNCTAD notes that gas carriers are essentially the only major contracting segment in which China did not rank first in 2024.

So the national industrial vocabulary changes.

In 2012:

What can Korea take from Japan next?

In 2026:

Can Korea hold HBM?

Can it preserve high-end shipbuilding?

Can Korean batteries recover against China?

Can Hyundai avoid the fate of European automakers?

The word that quietly enters the conversation is still.

Korea is still world-class in memory.

It is still formidable in shipbuilding.

It still has major battery companies.

That is a different psychology from ascent.

Then the dependencies become visible

The more interesting reversal is not even Chinese competition.

It is that Korea’s 2010s success had created an image of much greater autonomy than Korea actually possessed.

Take finance.

Korea built Samsung, Hyundai, SK Hynix and LG, but it never built a currency or capital market remotely equivalent to its industrial position.

In MSCI’s 2026 market-accessibility review, Korea still lacks a fully deliverable offshore won market. MSCI continues to treat access to Korea differently from developed markets despite decades of industrial success.

Take energy.

Roughly 70% of Korea’s crude oil imports still come from the Middle East. In 2026 Seoul was sufficiently concerned about supply security that it negotiated deeper crude-storage and emergency-supply arrangements with the UAE.

Take America.

The United States increasingly does not simply want to buy Korean industrial products. It wants Korean industry physically located in America.

The current U.S.–Korea economic arrangement involves a Korean investment commitment of roughly $350 billion in the United States—an enormous figure beside Korea’s $1.87 trillion annual economy.

Samsung can become a stronger global semiconductor company by building more fabrication capacity in America.

Hyundai can protect its U.S. market by building cars in Georgia.

LG Energy Solution can become a more important battery company by producing inside North America.

But Korea then has to confront an uncomfortable distinction:

The success of Korean companies is no longer automatically the success of Korean territory.

That is a classic post-peak problem.

And finally there is Japan.

As recently as the 2010s, national ascent made it easier for Korea to imagine that it was gradually escaping its old strategic relationship with Japan.

In 2026, the logic is running the other way.

South Korea and Japan have agreed to deepen cooperation on oil and LNG stockpiles, emergency energy swaps and security, and in August 2026 their officials pledged closer macroeconomic coordination after an unusually coordinated foreign-exchange intervention.

The history did not disappear.

The strategic room did.

That, to me, is the most interesting thing about post-peak Korea.

The country did not suddenly acquire these dependencies in the 2020s.

The rise had made them feel less important. The slowdown makes them visible again.

Now look at China in the 2020s

China today sits in almost the opposite psychological position.

Its economic growth is already slowing, but its international evidence of ascent is still accumulating.

China accounted for 32% of all global manufacturing value added in 2024.

That is not merely first place.

It is more than the United States, Japan, Germany and South Korea combined.

In 2025, China produced nearly 75% of the world’s electric cars. It produced more than 80% of the world’s battery cells. Chinese battery companies supplied almost three-quarters of the batteries deployed in electric cars globally.

Those are extraordinary numbers.

And importantly, they are changing prestige rather than merely production statistics.

BYD is no longer interesting because China can manufacture cheap cars.

Huawei is no longer interesting merely because it survived sanctions.

Chinese AI models are no longer evaluated solely as cheaper imitations of American models.

Chinese drones, batteries, EVs, robots and industrial systems increasingly force foreign competitors to benchmark against China.

That is the Korea-2010s feeling.

The old status assumption—

foreign means technologically superior—

starts breaking down.

And the shift is beginning to show up outside factories.

China recorded 35.2 million foreign visits in 2025, up strongly from the previous year. More than 30 million foreign entries used visa-free policies, up almost 50% in one year.

Meanwhile 495,000 Chinese students returned from overseas study in 2024, up 19.1% in a single year. Since 2012, the Ministry of Education says 5.63 million overseas students have returned.

Even foreign opinion is becoming less uniformly hostile among younger people.

Pew’s 2026 polling found favorable views of China among 61% of Canadians aged 18–34, versus 39% among those over 50. In Britain the split was 54% versus 37%; in France, 51% versus 23%; in Mexico, 70% versus 44%. This is not yet Korean-style cultural affection, and geopolitics obviously distorts the comparison. But it does suggest that younger outsiders are updating their mental picture of China faster than older cohorts.

This is what I mean by peak esteem.

Not that everything inside China is going well.

Not that Chinese people are uniformly optimistic.

And certainly not that China has reached maximum economic power.

It means that around the late 2020s, China may receive the maximum psychological dividend from several decades of accumulated rise.

Foreigners discover the factories after they have already been built.

They discover Shenzhen after it is already rich.

They discover Chinese EVs after the supply chain has already been assembled.

They discover Chinese AI after the research ecosystem has already matured.

Prestige arrives late.

Which is exactly why it can peak just as the underlying growth machine is slowing.

The IMF currently projects Chinese growth falling from 5.0% in 2025 to 3.4% by 2030.

So the China of 2030 could simultaneously be growing at barely one-third of its old miracle-era rate and enjoying the highest international technological prestige in its modern history.

That would be very Korean.

Then comes the 2030s

This is where the analogy becomes more interesting than the standard “India rises, China ages” story.

The IMF has modeled Chinese potential growth at only around 2.8% a year during 2031–40 without major structural reforms, down from roughly 3.8% during 2025–30. It also models a much better reform path, so this is not destiny. But it gives us a plausible baseline for what a post-peak decade actually looks like.

The OECD reaches a related conclusion from another direction: East Asia’s share of global output continues rising until the early 2030s, then begins declining. China drives most of that reversal.

That would make the early 2030s a surprisingly clean candidate for the turn.

And once the turn happens, I think the Korean pattern reappears.

Chinese companies start succeeding outside China

Today, the strange thing about Chinese EV dominance is how Chinese it still is.

In 2024, less than 2% of the electric cars made by Chinese manufacturers were produced outside China.

That will almost certainly change.

The battery industry already shows the direction.

In 2024 China held roughly 85% of global battery manufacturing capacity. Based on committed projects, the IEA expects geographical diversification to reduce that to roughly two-thirds by 2030.

But Chinese ownership does not disappear. In Europe specifically, Chinese companies’ share of battery manufacturing capacity could rise from less than 10% in 2024 to more than 30% by 2030.

That is the post-peak paradox in numerical form.

Chinese firms become more global.

Chinese territory becomes less dominant.

By the 2030s the same logic could spread from batteries into cars, robotics, machinery, solar equipment and electronics.

Europe tells Chinese manufacturers: build here.

Brazil tells them: build here.

Indonesia tells them: build here.

India tells them: if you want this market, localize.

Chinese corporations may comply and prosper.

But China itself captures less of the next factory, the next industrial worker and the next round of capital investment.

Korea is beginning to confront this distinction through Samsung, Hyundai and LG.

China could face it at continental scale.

The export machine stops looking purely like power

China’s enormous trade surplus currently looks like evidence of industrial superiority.

And it is.

But there is another interpretation that becomes more important if domestic growth weakens.

A country that produces far more manufactured goods than its own consumers absorb needs foreigners willing to buy the difference.

The IEA already estimated that Chinese electric-car production exceeded domestic demand by 20% in 2025, helping push Chinese EV exports above 2.5 million vehicles, twice the previous year’s level.

During ascent, this is frightening to everyone else:

China can overwhelm our industries.

During the post-peak period, foreign governments begin recognizing the other side of the bargain:

China needs access to our consumers.

That gives them bargaining power.

Tariffs.

Local-content requirements.

Joint ventures.

Technology conditions.

Investment screening.

Production localization.

The exact instruments will differ by country.

But the structural change is important.

The industrial machine China built to reduce foreign dependence can itself become dependent on foreign market access.

Then China discovers how incomplete its financial ascent was

This may become one of the most psychologically uncomfortable comparisons of the 2030s.

China is already the manufacturing center of the world.

Yet in the third quarter of 2025, the renminbi accounted for less than 3% of global trade settlement, around 8% of trade finance, and only about 2% of global foreign-exchange reserves.

That gap is easy to tolerate when China is growing at 5%, factories are gaining share and capital wants exposure to the Chinese growth story.

It becomes much more conspicuous in a 2–3% growth world.

Imagine China in 2037: still producing perhaps more industrial goods than any country in history, yet Chinese firms and wealthy households still care intensely about access to dollar assets, global financial centers and foreign capital.

Korea built Samsung without building a globally important won.

China may discover that even manufacturing on a superpower scale did not automatically create a Chinese financial system of comparable international weight.

That would be a genuine post-peak revelation.

Resource dependence starts looking different too

China is much more resource-secure than Korea. It has coal, enormous renewable capacity, nuclear power and continental depth.

But the world's largest industrial system consumes resources on a scale no other country has to manage.

Oil and gas from abroad.

Iron ore.

Copper.

Bauxite.

Agricultural commodities.

Feed.

Various specialty minerals.

A large portion arrives across maritime routes extending through the Indian Ocean and beyond.

During the rise, that looks like globalization.

During a period of strategic competition and slower growth, it starts looking more like constraint.

China can build an enormous navy.

It cannot cheaply make the Persian Gulf, Indian Ocean, Southeast Asian sea lanes and every major commodity-producing region politically Chinese.

That difference matters.

The post-peak realization may be:

We became the world’s largest industrial power without becoming sovereign over the global system that feeds that industry.

That is not Korea’s precise problem.

But it is the Chinese equivalent.

And China may suddenly need countries it spent the 2020s intimidating

This is perhaps the closest analogy of all to Korea rediscovering Japan.

An ascending China can afford to imagine that time solves strategic problems.

Japan will eventually adjust.

South Korea will increasingly fall into China’s economic orbit.

Europe cannot afford to decouple.

Southeast Asia will accommodate the regional giant.

American relative power will keep eroding.

But a post-peak China has to calculate differently.

If growth is structurally below 3%, manufacturing is slowly decentralizing, foreign capital has alternatives and the international coalition around China is hardening, then Beijing has more reason to ask:

Can we really afford permanent hostility with Japan?

Can we afford Europe treating China primarily as a security problem?

Can we afford India becoming a permanent strategic adversary?

Can we afford Southeast Asia systematically reducing its reliance on Chinese supply chains?

That is the diplomatic equivalent of Korea needing Japan again.

Nationalist grievances do not disappear.

The country simply has less freedom to indulge all of them simultaneously.

The Korea–China symmetry is sharper than it first appears

Put the two cycles side by side:

This is why I think the analogy is more useful than another demographic forecast.

We already know China is aging.

We already know India will become larger.

Those observations tell us remarkably little about what post-peak China will feel like.

Korea gives us a more interesting model.

The post-peak decade is the decade when a country discovers that many of the powers it thought it had acquired were actually conditional.

Industrial power depended on markets.

Corporate power did not guarantee domestic production.

Manufacturing strength did not create monetary power.

Technology did not create energy independence.

Military capability did not eliminate alliance or geopolitical constraints.

And historical antagonisms that felt affordable during ascent suddenly become expensive.

That is what Korea’s 2020s are beginning to reveal.

The real change is the question people ask

A Korean in 2013 could reasonably ask:

How much more important will Korea be in ten years?

A Korean in 2026 increasingly has reason to ask:

Which parts of Korea’s current position can we still hold in ten years?

That is the turn.

China may cross it in the early 2030s.

A Chinese person in 2027 can still plausibly look ten years ahead and assume that China will be more technologically important, more industrially dominant and harder for other countries to resist.

By 2037, the psychologically characteristic question may be different:

How much of what we achieved by 2030 can we preserve?

China does not have to collapse for that to happen.

Its cities can remain magnificent.

Its companies can remain globally important.

Its military can continue getting stronger.

Its culture may actually become more internationally influential.

Its citizens can keep getting richer.

Relative decline requires something much simpler:

ten years stop reliably making China more powerful relative to everyone else.

Once that happens, the entire interpretation of the country changes.

A factory moved overseas is no longer merely globalization. It looks like hollowing out.

A wealthy family moving assets abroad is no longer merely diversification. It looks like lost confidence.

A compromise with Japan is no longer tactical pragmatism. It looks like reduced room for maneuver.

Dependence on foreign markets is no longer proof of export strength. It becomes vulnerability.

And a technological sector China merely retains no longer feels like another conquest.

It becomes something that must be defended.

That is what I mean by China’s Korea moment.

The 2020s may be the decade when China receives the maximum prestige dividend from forty years of ascent. The 2030s may be the decade when it discovers just how many of the apparent freedoms created by that ascent were never entirely its own.

For most of the last forty years, American capitalism has seemed to move in one direction.

More competition.

More financial discipline.

More layoffs.

More outsourcing.

More venture capital.

More winner-take-all technology markets.

More pressure on every employee, company and city to justify its existence economically.

A company announces that artificial intelligence will allow it to eliminate 20% of its workforce and Wall Street celebrates.

A startup destroys an incumbent and Silicon Valley celebrates.

A forty-year-old employee whose productivity no longer justifies his compensation gets fired.

A company sitting on excess labor or capital is told to return it to shareholders.

For roughly forty years, the underlying rule has been remarkably consistent:

Slack is failure.

Capital should constantly move toward its highest return.

Labor should constantly move toward its highest productivity.

Companies should constantly face the possibility of replacement.

And if technology can eliminate a million jobs, the assumption is that the economy will eventually produce something better for those workers to do.

I increasingly wonder whether this regime has a historical endpoint.

Not because capitalism disappears.

Not because AI fails.

And certainly not because American finance becomes weaker.

Quite the opposite.

The system may eventually change because it succeeds.

My previous essay argued that the China shock could force the United States to turn the advanced allied world into something much closer to a single American-centered strategic system. If that succeeds, the next question becomes stranger:

What happens to American capitalism after the external competitor that justified forty years of disruption no longer looks capable of replacing it?

I suspect the answer could be an American economy that remains technologically extraordinary and financially dominant, but becomes more concentrated, more politically managed, more protective of employment and significantly less tolerant of creative destruction.

The transition might look surprisingly Japanese.

But at American scale.

The Moment of Victory Probably Comes in the 2030s

China does not have to collapse.

That is the wrong threshold.

The relevant moment comes when corporations, financial markets and governments stop organizing themselves around the expectation that China will eventually replace the American-centered system.

That could happen substantially before China becomes weak in absolute terms.

And there are already reasons to think the early-to-mid 2030s could be the critical window.

China's official population fell by 3.39 million people in 2025.

Births fell to 7.92 million against 11.31 million deaths.

Twenty-three percent of the population is already 60 or older. (National Bureau of Statistics of China)

The IMF is not forecasting a sudden Japanese-style crash.

It is forecasting something more useful for this argument: gradual structural convergence toward much slower growth.

Its February 2026 baseline has Chinese real GDP growth falling from 5.0% in 2025 to:

YearIMF China real-GDP growth projection20264.5%20274.0%20283.9%20293.7%20303.4%

The IMF attributes that deceleration to a shrinking labor force, decreasing returns to investment and slower productivity growth. (IMF)

There is another number in the same report that receives less attention.

The IMF's broad measure of China's augmented government debt—including local-government financing vehicles and related off-budget liabilities—rises from an estimated 126.6% of GDP in 2025 to 153.7% by 2030 under its baseline. (IMF eLibrary)

That does not mean China is about to become insolvent.

It means the old combination of:

labor-force expansion + property + infrastructure + debt-financed investment + export-led convergence

gets progressively harder to repeat.

So I would think about the next fifteen years roughly like this:

PeriodPossible political-economic phase2026–2030Maximum China shock: AI, defense, energy, manufacturing and capital mobilization2030–2035Tipping period: structural Chinese slowdown becomes harder to dismiss2035–2045If the American bloc holds together, the expectation of Chinese replacement could break2040s onwardPost-competition political economy begins to emerge

The exact date is unknowable.

But I would put 2035 near the center of the range where the narrative could change.

Not:

China is finished.

Rather:

China is still enormously powerful, but it probably isn't going to become the operating system of the advanced world.

That psychological transition is sufficient.

The China Competition May Make America More Disruptive First

Nothing in this thesis implies a sleepy America over the next ten years.

The opposite seems more likely.

As long as policymakers genuinely believe technological leadership is existential, creative destruction remains strategically valuable.

America needs:

• frontier AI;
• autonomous weapons;
• robotics;
• nuclear power;
• biotechnology;
• semiconductor capacity;
• space systems;
• new materials.

Government therefore has a reason to tolerate extreme experimentation.

If an unknown company can create something that Boeing, Lockheed, Pfizer or Microsoft failed to create, Washington wants that company to exist.

If an AI company needs tens of billions of dollars immediately, capital gets mobilized.

If automation eliminates a category of work but increases national technological capacity, government has reason to tolerate the disruption.

During the competition phase, the answer to almost every objection is:

China.

Why tolerate extreme founder wealth?

China.

Why subsidize semiconductor companies?

China.

Why allow AI companies to accumulate extraordinary power?

China.

Why rebuild defense production?

China.

Why accept large amounts of creative destruction?

Because losing the technological race is perceived as worse.

This is mobilization capitalism.

But Today's Venture Boom Is Already Becoming Something Different

The venture-capital numbers are particularly interesting because they show that:

more innovation capital does not necessarily mean broader entrepreneurial dynamism.

U.S. venture investment reached $320 billion in 2025, the second-highest annual total ever.

That sounds like an extraordinary startup boom.

Look underneath it.

Just 487 mega-deals of $100 million or more accounted for 67% of all venture dollars.

Those deals were only 3.2% of deal count.

Remove them and the remaining roughly 14,865 deals received about $105 billion—much closer to a normal late-2010s venture market.

The five largest companies alone—OpenAI, CoreWeave, xAI, Anthropic and Databricks—raised nearly $60 billion. (NVCA)

AI accounted for 65.4% of U.S. venture deal value.

And the concentration also exists one layer above the startups.

Only 101 first-time venture funds formed in 2025, down almost 78% from 457 in 2021.

Meanwhile the ten largest funds captured 32.9% of all VC fundraising, compared with 13% four years earlier. (NVCA)

That is a very peculiar form of dynamism.

Capital deployment is enormous.

Technological ambition is enormous.

Yet progressively larger fractions of the money are being allocated by fewer institutions into fewer organizations.

The system is becoming:

more technologically aggressive and more financially concentrated at the same time.

That may be exactly the bridge toward the next regime.

America Was Already Becoming Less Entrepreneurial Before AI

The popular image of the United States is still:

startups continually replace old companies.

The long-run data are less romantic.

Census Bureau research finds that approximately 90% of U.S. workers were employed by mature firms by 2019.

The share of workers employed by firms with at least 100 employees increased from about 41% in 1978 to 48% in 2019.

The Census Bureau's conclusion is straightforward: employment has become progressively concentrated in older firms. (Census.gov)

High-growth firms have also become less common.

The share of U.S. firms classified as high-growth fell from just under 20% in 1978 to below 13% in 2020.

Among continuing firms, the share classified as high-growth fell from 4.8% to 2%. (Census.gov)

And by 2023, an NBER analysis found that America had only about half as many publicly listed firms per capita as other developed economies. (National Bureau of Economic Research)

In other words:

the post-1980 era celebrated creative destruction culturally while gradually producing a more incumbent-heavy corporate structure economically.

The next regime would not need to create concentration from nothing.

It would be accelerating something already underway.

The Question Is What Happens Once Concentration Becomes Strategically Useful

Today we usually interpret concentration as a failure of competition.

But consider the strategic logic from Washington's perspective.

Imagine a future American AI company:

• employs 300,000 people;
• spends $70 billion annually on R&D;
• supplies U.S. defense systems;
• operates critical NATO infrastructure;
• generates enormous profits from foreign markets;
• supports thousands of domestic suppliers;
• maintains data centers in twenty states.

Now imagine a startup appears with technology capable of destroying that company.

The America of 1998 might say:

Excellent.

The America of 2045 may ask a different question:

Do we actually want this institution destroyed?

Once an incumbent becomes:

• an employer;
• a research laboratory;
• a national-security contractor;
• a tax base;
• a foreign-rent collector;

its survival acquires political value independently of shareholder returns.

That is where the logic of creative destruction begins reversing.

High R&D Does Not Require Startups

This is perhaps the most important economic fact underlying the entire argument.

The United States already conducts most of its research inside corporations.

In 2024, businesses performed $769 billion of U.S. R&D, representing 77% of all American R&D performance.

Businesses also funded 75% of total U.S. R&D. (NCSES)

That means a future decline in startup dynamism does not mechanically imply technological stagnation.

An economy can have:

fewer new corporate institutions

while simultaneously having:

enormous research expenditure.

Microsoft can research.

Nvidia can research.

Aerospace giants can research.

Pharmaceutical companies can research.

Defense firms can research.

The question is not whether innovation disappears.

The question is what kind of innovation remains.

Japan Shows That High R&D and Low Dynamism Can Coexist

Japan is useful here for a different reason than in my previous essay.

The interesting post-1990 fact is not that Japan stopped doing research.

It did not.

The OECD's 2024 assessment still describes Japanese business dynamism as weak, with relatively few startups and insufficient exit of low-productivity firms. (OECD)

Historically, Japan's annual firm-entry rate sat around only 4–5%, far below rates elsewhere in the advanced world. Firms more than ten years old accounted for roughly three-quarters of its small enterprises. (OECD ONE)

Yet Japan remains one of the world's most research-intensive economies.

That is the distinction.

A system can continue producing:

• better batteries;
• better automobile components;
• better robotics;
• better materials;
• better manufacturing processes;

without continually producing:

the next Toyota that destroys Toyota.

The technological frontier can continue moving while the institutional hierarchy barely moves.

That may be the strange destination of American capitalism.

The Startup Becomes a Supplier Rather Than a Successor

Imagine Silicon Valley in 2050.

Startups still exist.

A founder discovers:

• a new AI architecture;
• a robotic-control system;
• a cancer treatment;
• a military sensor;
• a new semiconductor material.

But instead of becoming the next Microsoft, the likely successful exit is:

Microsoft buys it.

Lockheed buys it.

Pfizer buys it.

Nvidia buys it.

The startup becomes an external laboratory for the national champion.

Venture capital increasingly functions as:

outsourced corporate R&D.

The founder can still become extremely rich.

Experimentation can remain intense.

But corporate turnover falls.

That produces:

high technological progress

without:

high institutional disruption.

The distinction is critical.

AI Makes the Employment Question Immediate

This isn't merely a 2045 thought experiment.

We can already see hints of the employment tension.

Stanford researchers using payroll data covering millions of American workers found no evidence as of mid-2026 of economy-wide AI job destruction.

But they found something narrower and potentially important.

Among workers aged 22–25 in highly AI-exposed occupations, employment was approximately 19% below where it would have been if it had kept pace with young workers in less-exposed occupations.

The researchers found that the adjustment was occurring primarily through reduced hiring, rather than mass firing of existing experienced workers. (Stanford Digital Economy Lab)

That is almost a miniature version of the political choice the country could eventually confront.

AI does not have to fire the existing employee.

The company simply stops hiring the next one.

From the shareholder's perspective, that can be excellent.

From the perspective of the state, it creates a different problem.

AI Can Also Make Ordinary Workers More Productive

The employment outcome is not technologically predetermined.

One of the most influential early workplace studies looked at 5,179 customer-support agents.

AI assistance increased productivity by about 14% overall.

But novice and lower-skilled workers improved by 34%.

Highly experienced workers received relatively little benefit. (National Bureau of Economic Research)

That's important because AI can support two very different economic models.

Model One: substitution

100 workers become 60.

The company produces the same output.

Margins rise.

Model Two: augmentation

100 workers remain.

They produce substantially more output.

Or they use part of the productivity gain for:

• better service;
• shorter hours;
• additional customers;
• more experimentation;
• more human oversight.

The technology does not determine which equilibrium society chooses.

Institutions do.

Washington May Eventually Prefer Augmentation

The government's incentives could change sharply as the fiscal pressure gets worse.

CBO's 2026 baseline projects federal debt held by the public rising from 101% of GDP in 2026 to 120% in 2036.

The government is expected to borrow another $26 trillion between the end of 2025 and the end of 2036.

Net interest costs rise from roughly $1 trillion in 2026 to $2.1 trillion in 2036.

By the end of CBO's 30-year projection, debt reaches 175% of GDP. (Congressional Budget Office)

That is a government with powerful reasons to care about employment.

A world in which AI generates gigantic corporate profits while eliminating taxable middle-class income produces an awkward fiscal bargain.

The state receives somewhat greater corporate taxation.

But it also receives:

• fewer payroll taxes;
• less labor income tax;
• greater demand for transfers;
• larger healthcare obligations;
• greater political instability.

At the same time, interest payments are absorbing an ever-larger fraction of government revenue.

The government may therefore prefer:

slightly lower corporate efficiency + broader employment

to:

maximum profit per employee + permanent redistribution to the unemployed.

That would be an enormous break from the economic philosophy of the last forty years.

The Deal Could Become Monopoly in Exchange for Employment

The foundations of this arrangement are already visible.

Consider the CHIPS program.

The federal government did not simply say:

Here is money. Maximize shareholder returns.

Its award to Micron included up to $6.165 billion in direct federal funding supporting a long-term domestic investment plan.

The projects are expected to create roughly 20,000 jobs.

Micron's New York and Idaho projects use project labor agreements and registered apprenticeships.

The earlier agreement also included at least $40 million specifically for workforce development. (U.S. Department of Commerce)

TSMC's Arizona agreement similarly combines up to $6.6 billion in federal funding with more than $65 billion in planned private investment and more than 20,000 expected jobs. (U.S. Department of Commerce)

And the government's semiconductor R&D institution, the National Semiconductor Technology Center, received a $6.3 billion, ten-year federal award explicitly covering both R&D and workforce development. (U.S. Department of Commerce)

These programs are not proof that America is already becoming corporatist.

But notice the logic.

Government gives:

• capital;
• loans;
• protection;
• demand;
• infrastructure.

Corporations promise:

• domestic production;
• strategic capability;
• jobs;
• training;
• investment.

That is already a different political relationship from:

the market decides.

Imagine That Bargain After America Wins

Now move forward fifteen years.

An American technology company earns enormous rents from:

• AI exports;
• cloud;
• intellectual property;
• finance;
• defense technology;
• pharmaceuticals.

It operates within a foreign market protected by the American military and trade system.

Government can say:

Your profitability is not purely private achievement.

The national system created part of your franchise.

And therefore:

you owe something to the national system.

Not necessarily through nationalization.

The company remains private.

Its shareholders remain wealthy.

Its executives remain powerful.

But privileges become conditional.

Want access to government procurement?

Maintain domestic employment.

Want access to export-controlled technologies?

Keep strategic R&D in America.

Want subsidies?

Train workers.

Want protected access to allied markets?

Maintain domestic production capacity.

Want favorable regulatory treatment?

Do not eliminate 100,000 employees simply to raise margins another four points.

The firm remains capitalist.

But shareholder value is no longer the sole political objective.

That is what I mean by corporate nationalism.

Slack Becomes a Feature

This would be the clearest discontinuity from post-1980 America.

Imagine a national champion employs 300,000 Americans.

Its next-generation AI systems make it possible to produce the same output with 180,000.

The existing ideology says:

Fire 120,000.

The mature hegemonic system might respond:

Why?

Suppose the company already earns a 30% operating margin.

Suppose it enjoys protected access to an enormous allied market.

Suppose the state values employment.

Then some productivity gains can become:

• shorter hours;
• internal training;
• additional research;
• greater redundancy;
• better human service;
• regional offices;
• longer-term experimentation;
• more generous staffing.

The economist calls it organizational slack.

The politician calls it stable employment.

The employee calls it a career.

And the corporation can afford it because its monopoly rents are being generated partially abroad.

America Becomes a High-Tech Rentier Economy

The export sector does not need to employ everyone directly.

That is one of the more counterintuitive implications of AI.

Imagine that a relatively small number of American workers and machines generate enormous revenues from:

• AI;
• cloud software;
• finance;
• aerospace;
• pharmaceuticals;
• intellectual property;
• defense.

Those revenues become American national income.

They appear as:

• wages;
• profits;
• taxes;
• dividends;
• investment;
• construction;
• domestic purchasing power.

Most Americans can then continue working in completely ordinary sectors:

• healthcare;
• restaurants;
• hospitality;
• home construction;
• logistics;
• education;
• entertainment;
• personal services.

The waiter does not need to build the AI model.

His customers just need enough purchasing power to buy dinner.

That purchasing power can ultimately trace back to American technology rents earned overseas.

In one respect, it resembles an oil economy.

A relatively narrow export sector finances a much larger domestic service economy.

Except America exports:

intelligence, technology, finance and security rather than petroleum.

Wall Street Could Become Bigger While Capitalism Becomes Less Dynamic

This sounds contradictory only if we equate finance with entrepreneurship.

They are not the same.

Wall Street could become substantially more powerful in the post-competition economy.

But finance changes jobs.

The last forty years idealized finance as:

identify opportunity → finance disruptor → replace incumbent.

The mature system increasingly uses finance to:

manage assets → finance national champions → refinance infrastructure → acquire competitors → distribute rents.

Asset management becomes more important.

Private credit becomes more important.

Infrastructure finance becomes more important.

Pensions become more important.

Insurance becomes more important.

M&A becomes more important.

Venture capital survives, but increasingly as part of a concentrated institutional financial system.

This is why today's venture concentration matters.

It may be showing an early transition from:

entrepreneurship financed by capital

toward:

capital allocating gigantic sums among strategically important institutions.

The Meritocratic Tournament Could Soften

Another consequence would be social.

American professional life since the 1980s increasingly became a tournament.

Get into the strongest school.

Join the strongest company.

Move to the strongest city.

Work harder.

Switch employers.

Accumulate equity.

Keep moving.

If you stop improving, somebody younger, cheaper or more talented replaces you.

This system maximizes competitive intensity.

It also produces extraordinary insecurity.

A national-champion economy has less reason to organize labor that way.

Internal labor markets can return.

Companies train employees rather than constantly recruiting externally.

A competent middle-aged engineer becomes an asset rather than excess payroll.

The corporation once again functions partly as a social institution.

The strange endpoint could be:

1955 employment norms running on 2050 technology.

Not literally lifetime employment.

But much less expectation that healthy corporations should continually prove their efficiency through layoffs.

The Middle Class Could Return Through Institutional Rents

This does not require recreating Detroit in 1965.

The important feature of the old industrial middle class was not that every factory worker possessed exceptional individual productivity.

It was that millions of ordinary workers were embedded in extraordinarily productive institutions.

A worker's wage was partly a claim on:

the rent generated by the corporation.

Something similar could happen again.

The corporation earns huge AI or financial rents.

Workers receive part through:

• high wages;
• stable employment;
• benefits;
• training;
• pensions;
• shorter working time.

The institution becomes productive enough to carry a broad middle.

That would directly reverse one of the major assumptions of the post-1980 labor market:

every employee must individually justify their compensation through continuously measurable marginal productivity.

Inequality May Peak During the Competition

I would not expect the transition to begin egalitarian.

The China shock could actually create another decade of spectacular inequality.

The winning assets are likely to be concentrated:

• AI equity;
• semiconductor ownership;
• defense technology;
• energy;
• venture funds;
• intellectual property.

The state tolerates enormous fortunes because technological mobilization matters more than distribution.

But after the external threat recedes, the political justification gets much weaker.

The government can ask:

Why exactly does this protected national champion require another tax cut?

Or:

Why should a company earning enormous foreign rents eliminate 80,000 taxpayers merely to raise its operating margin?

Housing could matter enormously too.

We have spent decades treating rising house prices as household wealth creation.

A mature rentier America may eventually prefer twenty years in which:

wages rise faster than land.

Housing prices do not need to collapse.

They can simply stop outperforming income.

That alone would transfer enormous relative economic power toward younger workers.

So the post-victory society could still contain billionaires and enormous corporations while becoming less unequal because the middle begins catching up.

The State Becomes Stronger Without Becoming an Administrative Monster

This is why I would not expect the endpoint simply to be European-style bureaucracy.

AI itself can automate administration.

The more plausible American mechanism is governance through conditional corporate privileges.

Washington does not have to employ another million civil servants.

It can tell companies:

Want the contract? Maintain domestic production.

Want the subsidy? Train workers.

Want export privileges? Keep strategic research here.

Want protection from foreign competitors? Carry domestic capacity.

The state determines outcomes.

Private institutions implement them.

That's a stronger state without necessarily being a larger clerical state.

The CEO Changes Too

The heroic executive of the 2010s says:

We move fast.

We disrupt ourselves.

We cut unnecessary labor.

Nobody has a right to a job.

The prestigious CEO of 2050 could sound completely different:

We employ 350,000 Americans.

We invest tens of billions in research.

We operate facilities in twenty states.

We train thousands of engineers and technicians.

We provide infrastructure critical to the alliance.

We are an American institution.

The corporation becomes proud of permanence.

Scale becomes legitimacy.

Employment becomes legitimacy.

R&D becomes legitimacy.

National usefulness becomes legitimacy.

That is far closer to the self-image of the old Bell System or General Electric than to the self-image of a 2015 SaaS startup.

The Darker Side Is Obvious

There is a reason America historically celebrated creative destruction.

Protected incumbents become complacent.

Government contracts become political favors.

Corporate executives and regulators become indistinguishable.

Startups get bought before they can become competitors.

Lobbying replaces invention at the margin.

The system develops enormous capacity to improve existing technology but less ability to imagine entirely new institutional forms.

This is where the comparison with post-1990 Japan becomes uncomfortable.

The future America could still produce extraordinary:

• AI;
• pharmaceuticals;
• aircraft;
• robots;
• materials;
• weapons.

But the companies producing them might be largely the same companies decade after decade.

The technological frontier keeps moving.

The ownership hierarchy barely does.

The Historical Discontinuity

The last forty years have roughly said:

1980–2035

• competition;
• globalization;
• financial discipline;
• outsourcing;
• immigration;
• labor mobility;
• startups;
• venture capital;
• layoffs;
• shareholder value;
• creative destruction.

The next equilibrium could increasingly say:

2035–2060

• national champions;
• protected allied markets;
• high corporate R&D;
• stable employment;
• strategic production;
• corporate obligations;
• AI augmentation;
• patient finance;
• mergers instead of replacement;
• state-business coordination;
• rent distribution;
• managed dominance.

The first system asks:

What deserves to survive?

The second asks:

What must be preserved?

The China Shock Could Create the System That Ends the China Shock

There is an irony here.

China may force America into one final extraordinary period of disruptive capitalism.

The threat produces:

• AI laboratories;
• defense startups;
• semiconductor factories;
• nuclear companies;
• robotics;
• biotech;
• giant venture rounds;
• enormous fortunes.

The new companies help America win the competition.

And then the political argument for continuing to destroy institutions begins disappearing.

The startups become national champions.

The founders become establishment figures.

The venture funds become giant financial institutions.

Temporary subsidies become permanent industrial architecture.

Emergency supply chains become permanent supply chains.

The American system gradually shifts from:

How do we move faster?

toward:

How do we preserve what we have built?

That could be the real end of the China shock.

Not China's collapse.

Not capitalism's collapse.

And not a return to 1955.

Something stranger:

a technologically advanced,

financially dominant,

highly automated,

corporately concentrated America

that still spends enormous amounts on R&D,

but increasingly values stable employment and institutional continuity over the constant destruction of yesterday's winners.

The most disruptive phase of modern American capitalism may end not because disruption stopped working.

It may end because there is finally nobody left America urgently needs to outrun.

A few days ago, I thought the emerging American strategy was relatively simple.

China had become the only country capable of challenging the United States across industry, technology and military power.

So America would narrow its perimeter.

It would stop trying to manage the entire postwar world and concentrate on the countries most useful for competing with China.

Japan would move inward.

Australia would move inward.

Mexico would become the low-cost production floor.

Canada would supply energy and resources.

Europe, meanwhile, would become increasingly secondary.

That was the argument I made in my previous essay. I imagined a harder American core centered on the United States, Canada, Mexico, Japan, Australia and Britain, with continental Europe sitting farther outside the deepest layer.

I increasingly think that is wrong.

Not because Europe has suddenly become more useful militarily against China than Japan.

It hasn't.

The mistake was thinking America can afford to compete with China as America.

It probably cannot.

The real competition is increasingly one of scale.

And once you look at the numbers that way, Europe becomes almost impossible for Washington to abandon.

The more plausible American objective is not:

Shrink the empire until only the most strategically useful allies remain.

It may instead be:

Turn the entire advanced allied world into something capable of acting like one economic system—while making sure the United States remains its command center.

That produces a much larger bloc than I previously imagined.

But it also produces a much more hierarchical one.

And strangely, it may mean America squeezes Europe harder while allowing Japan to become stronger.

America Has a Scale Problem

For most of the postwar era, the United States did not really have to think this way.

The Soviet Union could threaten America militarily.

It could build missiles, tanks, satellites and nuclear weapons.

But it never created a civilian economic system remotely comparable to the American one.

China is different.

China is simultaneously:

• a continental-scale economy;
• the world's largest manufacturing system;
• a huge consumer market;
• an increasingly sophisticated technological power;
• and now one of the two largest scientific systems on Earth.

The last part may be the most important.

In 2024, U.S. research and development expenditure was approximately $1.01 trillion in purchasing-power-adjusted terms.

China was approximately $1.03 trillion.

By that measure, the two countries have essentially reached parity. The OECD itself now says China's R&D expenditure caught up with and slightly surpassed America's in 2024, although the exact comparison varies depending on the purchasing-power methodology used.

And research output is moving in the same direction.

In the 2026 Nature Index rankings, based on 2025 research output in the journals the index tracks, China ranked first overall with roughly twice the fractional research share of the United States. China also ranked first in natural sciences, chemistry, physical sciences and applied sciences.

This does not mean China has already surpassed America in every meaningful dimension of science.

It clearly hasn't.

America still dominates many elite institutions, commercial research networks, biotechnology ecosystems, AI companies and highly influential journals.

But the scale shift is unmistakable.

America is no longer competing with a country that possesses:

30% of American research capacity.

It is increasingly competing with another civilization possessing roughly comparable aggregate research resources.

That changes the strategic problem completely.

America Alone Is No Longer the Relevant Unit

Now add Europe.

The European Union spent about $612 billion PPP on R&D in 2024.

Japan spent about $234 billion.

South Korea spent about $162 billion.

Taiwan spent another $77 billion.

Put those together and the picture changes dramatically.

R&D system, 2024Approx. PPP-adjusted R&DChina$1.03TUnited States$1.01TEU-27$612BJapan$234BSouth Korea$162BTaiwan$77B

America against China is roughly:

$1 trillion versus $1 trillion.

But an integrated American system incorporating Europe and the major advanced East Asian allies approaches:

$2 trillion versus $1 trillion.

Suddenly the competition looks very different.

The same logic applies elsewhere.

China has population scale.

America has alliance scale.

China has a continental industrial ecosystem.

America has the ability—at least potentially—to combine:

• American software and capital;
• European science and wealth;
• Japanese machinery and materials;
• Korean semiconductors;
• Taiwanese fabrication;
• Australian and Canadian resources;
• Mexican manufacturing.

That may be America's single largest remaining structural advantage.

Not that America itself is larger than China.

It isn't.

But America sits at the center of a network that can be larger than China if the network behaves coherently.

And that creates the central political problem of the next era:

Having allies is not enough.

America increasingly needs those allies to behave like parts of the same system.

The Difference Between an Alliance and a Political-Economic Unit

This does not necessarily mean a literal United States of America and Europe.

There does not need to be one parliament.

There does not need to be one flag.

Europe does not need to become American territory.

The relevant integration can occur one layer above ordinary domestic politics.

Imagine Europe retaining:

• its own welfare states;
• its own tax systems;
• its own elections;
• its own languages;
• its own local industrial policies.

But on the questions determining global power, the system increasingly behaves as one.

One technology perimeter toward China.

One semiconductor export-control architecture.

One military command framework.

One broadly interoperable defense-industrial base.

One favored capital and investment zone.

One AI and cloud ecosystem.

One sanctions system.

One privileged internal market.

One broad foreign-policy orientation.

That would amount to something approaching political unity at the strategic layer even if constitutional sovereignty survives below it.

And if America is the military, financial and technological center of that system, Washington does not need to formally annex anything.

It needs to ensure that the most important decisions converge.

Europe Is Already Deep Inside the American Capital System

This is where my previous thesis becomes much harder to sustain.

I had imagined America potentially deciding Europe was no longer sufficiently useful and narrowing the strategic perimeter toward the Pacific.

But American capital is already extraordinarily concentrated in Europe.

At the end of 2025, the entire U.S. direct-investment position abroad was approximately $7.14 trillion.

The four largest European destinations alone were:

DestinationU.S. direct-investment positionUnited Kingdom$1.115TNetherlands$1.044TLuxembourg$645BIreland$512B

Together:

about $3.32 trillion.

That is roughly 46% of all U.S. direct investment abroad, sitting in only four European economies.

And that doesn't include Germany, France, Switzerland, Belgium, Spain, Italy or the rest of Europe.

There is an important accounting qualification.

Almost 46% of total American outward direct investment is booked through holding companies, so a trillion dollars booked in the Netherlands or Luxembourg does not mean there are literally a trillion dollars of American factories physically sitting there.

But that actually makes the point more interesting.

Europe is not merely an American manufacturing partner.

It is deeply embedded in the ownership architecture of American capitalism.

The region contains:

• corporate subsidiaries;
• holding structures;
• intellectual-property entities;
• financial operations;
• acquisition targets;
• pharmaceutical operations;
• manufacturing affiliates;
• enormous pools of consumers and corporate cash flow.

And the relationship is still deepening.

Of the $438 billion increase in America's outward direct-investment position during 2025, $350 billion was in Europe.

Meanwhile European companies are enormous investors in America too.

Europe accounted for $116.6 billion, or just over half, of all new foreign direct-investment expenditures in the United States in 2025.

So this is not a loose alliance sitting on top of separate economies.

It is already an unusually dense transatlantic ownership system.

Why Would America Abandon an Asset Base Like That?

This changes how I think about European military disengagement.

Suppose Washington simply says:

Europe no longer matters. Defend yourselves. We're going to Asia.

Europe then has an obvious response.

Build autonomous defense.

Build autonomous capital markets.

Build European cloud infrastructure.

Build European AI champions.

Build independent semiconductor capacity.

Create a stronger euro.

Keep European savings in Europe.

Develop a foreign policy independent of Washington.

Trade with China wherever European interests dictate.

From a European perspective, that would be rational.

But look at the result from America's perspective.

Washington would have voluntarily helped create:

another autonomous advanced economic pole.

And this happens precisely while America is struggling to match Chinese scale.

That makes less sense the more serious the China competition becomes.

If America needs aggregate scale, Europe is not something it can casually discard.

Europe contains too much:

• capital;
• income;
• scientific capacity;
• high-end consumption;
• industrial capability;
• accumulated wealth;
• human capital.

More importantly, American corporations already own enormous claims on that system.

Why would Washington voluntarily convert a deeply integrated capital zone into an independent rival financial and technological pole?

The China Shock Is Different From the Japan Shock

America has experienced something like this before.

But it is important to remember what the Japan shock actually was.

It was not simply:

Japanese companies are taking market share from American companies.

By the late 1970s and 1980s, Japan was producing a much more uncomfortable possibility:

What if a non-Western society had built a better-performing version of modern industrial civilization?

That was the real shock.

Consider life expectancy.

In 1960, Americans still lived longer on average: about 69.8 years in the United States versus 67.7 in Japan.

But Japan closed the gap astonishingly quickly. By roughly 1964–65, Japanese life expectancy had crossed above America's for the first time. By 1980, Japan was at about 76.0 years versus 73.6 in the United States. (countryeconomy.com)

A country that had emerged devastated from war only thirty-five years earlier was now keeping its citizens alive several years longer than the richest country in the world.

Education produced an equally uncomfortable comparison.

In the Second International Mathematics Study around 1981–82, Japanese lower-secondary students averaged about 63.5% correct, compared with 46.0% for American students.

The gaps appeared across essentially every major area:

Among advanced secondary students, Japan also substantially outperformed the United States in algebra and calculus. (ERIC)

This was not a marginal difference.

The median American teenager was living inside a country with vastly more accumulated wealth, famous universities and a much longer industrial history—and yet Japanese mass schooling appeared to be producing considerably stronger mathematical competence.

That fact went directly into America's national anxiety. The 1983 A Nation at Risk report opened by warning that America's “once unchallenged preeminence” in commerce, industry, science and technological innovation was being overtaken, and international mathematics and science comparisons were among its central pieces of evidence. (National Center for Education Statistics)

Then there was research.

Japan was not merely assembling technology invented elsewhere.

Its R&D intensity rose from about 2.0% of GDP in 1980 to roughly 2.9% by 1990.

The United States went from about 2.3% to 2.7% over the same period.

By the end of the decade, Japan was devoting a larger share of its national economy to R&D than America. (ERIC)

And the industrial results were becoming impossible to dismiss.

In semiconductors, American firms held about 57% of the world market in 1980, compared with 27% for Japanese firms.

By 1989 the positions had nearly reversed:

• Japan: 52%
• United States: 35%

Japan had become the world's largest semiconductor producer. (GAO)

The same broad story appeared in automobiles, consumer electronics, machine tools, robotics and manufacturing quality.

Japanese factories became global models for lean production, statistical quality control and continuous improvement.

American companies began importing not merely Japanese products, but Japanese methods of organizing production.

And this is what made the moment psychologically unusual.

Japan was not simply saying:

We can manufacture your inventions cheaply.

It was increasingly saying, through demonstrated outcomes:

Our factories may work better.
Our children may learn more mathematics.
Our people may live longer.
Our infrastructure may function better.
Our firms may invest more patiently.
Our national institutions may be better adapted to the next stage of industrial society.

That is why Ezra Vogel could publish a book in 1979 literally called Japan as Number One: Lessons for America.

The intellectual question was no longer simply:

How can Detroit beat Toyota?

It was:

What does Japan know about running a modern society that America does not?

That was a genuine civilizational-performance shock.

And America reacted accordingly.

It created SEMATECH.

It pressured Japan intensely over semiconductors and market access.

American firms imported Japanese production methods.

Education reform became explicitly linked to national power.

Industrial policy, technological competitiveness and the quality of the American workforce became national-security questions.

But there was one thing Japan could not do.

It could not scale its superior performance into a replacement world system.

Japan had roughly half America's population.

It depended heavily on American military protection.

It had no comparable alliance network.

The yen could not replace the dollar globally.

Tokyo's capital markets could not organize the world's financial system.

Japan could demonstrate that a non-Western society might outperform America on surprisingly fundamental measures of modernity.

But it could not plausibly organize the world around itself.

China changes that.

The Japan shock asked:

What if another society is doing modernity better than we are?

The China shock adds a second question:

What if that society is also large enough to build an alternative world system?

That is the discontinuity.

Japan forced America to become more competitive.

China may force America to become larger than America.

And That Changes Europe's Role

This is where I now depart most sharply from my previous argument.

I previously thought Europe's distance from China made Europe less important.

That remains true militarily.

Germany does not sit beside Taiwan.

France does not control the First Island Chain.

European factories cannot substitute for Japanese geography.

But the conclusion I drew was wrong.

Europe being strategically behind the front line does not necessarily mean America abandons Europe.

It may mean America treats Europe differently from the frontier.

America needs European scale.

But it may not need European strategic autonomy.

In fact, autonomous European capability can directly compete with American power.

Consider several hypothetical European successes.

A European Nvidia strengthens the West.

But it weakens Nvidia.

A European AWS strengthens Western cloud resilience.

But it weakens Amazon.

A deep European capital market strengthens Western finance collectively.

But it reduces New York's centrality.

A powerful euro makes the Atlantic world financially more diversified.

But it weakens the dollar.

An autonomous European defense command makes Europe militarily stronger.

But it reduces American control of the alliance.

A European technology policy independent of Washington gives Europe more options.

But it makes containing Chinese technology harder.

That produces a tension that did not matter nearly as much in the 1990s.

America wants Europe strong enough to add to the system.

But perhaps not so autonomous that Europe becomes another center of the system.

The Ideal American Outcome Is Not a Weak Europe

This distinction matters.

America does not benefit from Europe becoming poor.

A poor Europe:

• buys fewer American products;
• generates less investment income;
• produces fewer scientists;
• contributes less to defense;
• accumulates less savings;
• becomes politically unstable;
• requires more American resources.

The ideal Europe is wealthy.

Possibly extremely wealthy.

But it is wealthy in sectors that do not challenge the American command layer.

Europe could remain world-class in:

• tourism;
• luxury goods;
• specialized machinery;
• cultural industries;
• professional services;
• high-value property;

Its households could remain affluent.

Its historic cities could become even more valuable as global wealth increases.

Its real estate could attract enormous pools of foreign capital.

Its tourism economy could earn substantial external income.

Its governments and companies could borrow from Japanese and other global savings pools.

Europe could remain one of the world's most desirable places to live.

But the highest-rent scalable platforms increasingly sit in America:

• AI;
• software;
• cloud;
• frontier biotechnology;
• capital markets;
• venture capital;
• defense architecture;
• reserve currency;
• strategic intellectual property.

In that arrangement, Europe is not poor.

It is rich but increasingly subordinate at the commanding heights.

America Wants the European R&D System Too

This is another reason Europe cannot simply be discarded.

Remember the research numbers.

Europe's R&D system is roughly 60% the size of America's in purchasing-power terms.

That is enormous.

If Europe becomes strategically independent, America competes with China using approximately one trillion dollars of domestic R&D.

Europe's $600 billion becomes an independent pool.

Some European researchers cooperate with America.

Some cooperate with China.

Some build European competitors.

But suppose the Atlantic technology system becomes much more integrated.

Then European R&D increasingly feeds:

• American-owned companies;
• American cloud infrastructure;
• American AI platforms;
• American capital markets;
• American acquisitions;
• American defense systems.

The scientist does not even need to move to California.

A laboratory in Munich can create value that ultimately accrues through an American corporate ownership structure.

A French biotech company can be acquired by an American pharmaceutical company.

A Swedish startup can scale through U.S. venture capital.

A British AI company can run on American compute and eventually list in New York.

That means America can capture some of Europe's research capacity without physically importing every researcher.

This potentially solves one of America's other emerging problems.

The Old Talent Model May Be Becoming Harder

For several decades, America solved domestic human-capital shortages through an extraordinary mechanism:

recruit from the entire planet.

The best Chinese graduate student could go to Stanford.

The best Indian engineer could move to Silicon Valley.

The best European scientist could join an American laboratory.

This dramatically increased American technological capacity.

But a hard geopolitical division with China eventually makes that model harder.

Chinese students and scientists become security-sensitive.

India may remain strategically autonomous.

Immigration itself becomes politically contested.

The fully global talent market begins breaking apart.

At first glance, that looks disastrous for American science.

But Europe offers an alternative.

Instead of importing every scientist individually, America can integrate a whole allied research ecosystem.

That would still probably reduce the extreme selection intensity of American universities.

America would no longer be drawing as freely from the best people produced by every rival civilization.

But the system-level R&D capacity could remain enormous.

And strangely, that might improve the labor-market position of merely very good Americans.

The American Upper Middle Class Could Actually Benefit

The hyper-global talent system was excellent for maximizing American frontier output.

It was not necessarily excellent for every American professional.

A competent American engineer did not compete merely with other Americans.

He competed with the strongest people willing to migrate from:

• China;
• India;
• Russia;
• Iran;
• Eastern Europe;
• everywhere else.

That raises the quality of the American technological system.

But it also raises the threshold for participating in it.

Now imagine a different arrangement.

America captures more European R&D organizationally.

American companies gain access to an enormous integrated allied market.

Chinese labor-market competition is reduced.

Some Indian competition is reduced.

European researchers increasingly remain in Europe but work inside American-controlled corporate systems.

The quantity of American-controlled high-value activity becomes much larger.

Suddenly there may be more room for the merely excellent American rather than only the globally exceptional one.

The engineer who is clearly above average but not a world-class mathematician.

The technical manager.

The process engineer.

The laboratory operator.

The product engineer.

The defense technologist.

The person capable of participating in sophisticated systems without personally inventing the frontier.

That could regenerate a surprisingly broad high-tech upper-middle class.

Even the Ordinary Worker Might Gain

There is a second domestic effect.

Suppose America not only captures the high-rent technological layer but increasingly insists that strategically important goods sold into the allied market be produced within the trusted system.

Then European demand can support American:

• factories;
• defense plants;
• aerospace production;
• energy exports;
• semiconductor facilities;
• chemical plants;
• data centers;
• machine production.

The mechanism by which an ordinary American worker benefits is not that a high-school diploma magically becomes valuable again.

It is that industrial labor becomes scarce.

If America forces more production into the domestic economy while simultaneously limiting easy replacement through offshoring, companies have to bid for workers.

A completely ordinary worker can earn a strong wage when embedded inside an extremely productive, capital-intensive system.

That was one of the hidden foundations of the old American middle class.

The factory worker did not need to be a genius.

The productive system around him was extraordinary.

So the political economy of a harder American bloc could become surprisingly attractive domestically:

GroupPotential source of gainsCapital ownersPlatform, financial and monopoly rentsScientists / elite engineersLarger integrated research systemOrdinary-good professionalsMore high-value positions inside protected industriesSkilled workersDomestic reindustrialization and labor scarcitySome high-school workersManufacturing, logistics, construction, energy and defense demand

That is a much broader coalition than simply Silicon Valley.

It begins to explain how economic nationalism and continued American hegemony could coexist.

But Why Would Europe Accept This?

This is the obvious problem.

Why would Europeans voluntarily agree to become a rich but subordinate economic zone?

They wouldn't necessarily.

Which brings us to the asset America possesses that Europe still cannot easily reproduce:

military power.

Europe remains deeply dependent on the United States for parts of:

• nuclear deterrence;
• intelligence;
• strategic airlift;
• command architecture;
• missile defense;
• long-range strike;
• satellite infrastructure;
• military logistics.

At the 2025 Hague summit, NATO members committed to moving toward defense and security-related spending equal to 5% of GDP by 2035 while deepening transatlantic defense-industrial cooperation.

Europe is becoming militarily stronger.

But there are two entirely different ways that can happen.

One is:

Europe builds a militarily autonomous third pole.

The other is:

Europe supplies much more military capability inside an American-centered architecture.

Those outcomes are not equivalent.

America should strongly prefer the second.

Because then European taxpayers provide more of the resources while Washington retains much of the strategic leverage.

Military Protection Can Be Converted Into Economic Leverage

This is where the relationship becomes more openly imperial.

Suppose Europe needs continued American deterrence.

America needs Europe to align export controls against China.

Those issues can become connected.

Europe wants access to American AI and semiconductor technology.

America wants European regulation to become friendlier toward U.S. firms.

Those issues can become connected.

Europe wants privileged market access.

America wants more European defense spending.

Connected.

Europe wants U.S. intelligence and military interoperability.

America wants European China policy aligned with Washington.

Connected.

None of this requires a written grand strategy.

It follows naturally from asymmetric bargaining power.

If one side controls something the other side cannot easily replace, that dependency becomes negotiating leverage.

The harsher word is extortion.

Not criminal extortion.

Imperial extortion.

Use a dependency in one domain to extract concessions in another.

And once America needs every marginal advantage it can get against China, why would it voluntarily refuse to use that leverage?

Now We Can Return to the British Empire

This is where the historical analogy becomes useful.

Empires do not treat every subordinate territory equally.

The relationship depends on what the imperial center needs from that territory and what alternatives the subordinate territory possesses.

British Canada and British India illustrate two radically different arrangements.

Not because Canada mattered and India didn't.

India was vastly more important economically.

But Canada sat beside something India did not:

the United States.

That mattered enormously.

Canada's Outside Option

British North America had a huge alternative trading system directly across the border.

The United States was growing rapidly.

Trade flowed across the border.

People migrated across it.

The 1854 Reciprocity Treaty substantially deepened commercial integration between British North America and the United States.

When Washington terminated that agreement after the Civil War, the resulting loss of American market access became one of the pressures encouraging Confederation.

Canadian historical accounts explicitly list fear of American expansion, the loss of U.S. reciprocity and Britain's desire to reduce its direct colonial obligations among the external pressures behind Canadian union.

That means Britain faced an unusual problem.

If London extracted too much from Canada, Canada had somewhere else to turn.

If Britain suppressed Canadian economic development too aggressively, the United States remained next door.

If imperial membership became sufficiently unattractive, the neighboring economy offered an enormous alternative.

The map itself created bargaining power.

So Britain gradually discovered that a locally capable, increasingly self-governing Canada could remain inside the British system more securely than a resentful, tightly controlled one.

Race, settler politics and local institutions obviously mattered enormously to the actual history.

But there was also a strategic logic:

The frontier had an outside option.

That increased the cost of direct control.

Japan Has China Next Door

Now look at Japan.

Japan cannot simply become a Chinese satellite.

The political and security barriers are enormous.

But China is permanently next door.

China is:

• a gigantic market;
• a manufacturing superpower;
• an important trading partner;
• a source of supply chains;
• an alternative technological ecosystem.

Japan can also hedge through:

• Southeast Asia;
• India;
• Australia;
• its own foreign investment networks.

The point is not that Japan can effortlessly defect to China.

It cannot.

The point is that Washington cannot eliminate Japan's surrounding Asian economic geography.

And America simultaneously needs Japan to remain extremely capable in exactly the location where a Chinese challenge would occur.

That combination increases Japan's bargaining power.

America Cannot Afford to Hollow Japan Out

This produces a different American policy toward Japan than toward Europe.

Imagine America successfully relocates every valuable Japanese industry into the United States.

That could increase American GDP.

But eventually the strategy becomes self-defeating.

A semiconductor plant in Arizona cannot defend Okinawa.

A Texas shipyard cannot instantly repair Japanese naval forces during a Pacific war.

An American robotics company does not substitute for an entire functioning Japanese supplier ecosystem sitting beside China.

The United States therefore needs real industrial depth to remain physically inside Japan.

Japan needs:

• shipyards;
• missile production;
• precision manufacturing;
• advanced materials;
• semiconductor equipment;
• energy resilience;
• military logistics.

That means America may have to tolerate exactly the kind of allied industrial capability that it once tried to constrain.

During the Japan shock, a powerful Japanese semiconductor sector primarily looked like a competitor.

During the China shock, a powerful Japanese semiconductor and defense-industrial sector can become part of the strategic balance.

The same capability has acquired a second value.

Commercially it may compete with America.

Strategically it strengthens the American system.

The Frontier Gets a Better Deal

This suggests a general rule.

The closer an ally sits to the rival, and the more America physically depends on that ally's productive capability, the more autonomy Washington may have to tolerate.

Japan can become stronger.

Australia can build more military capability.

Possibly Korea and Taiwan can preserve unusually sophisticated industrial systems.

But there is a catch.

America may tolerate more autonomy of means while demanding less autonomy of alignment.

Japan gets more missiles.

More ships.

More defense production.

More industrial capability.

But those capabilities become increasingly interoperable with the American military structure.

Japan is not being made independent of the system.

It is being made more powerful inside the system.

The ideal frontier ally is not weak.

It is:

strong enough to fight, but sufficiently integrated that its strength reinforces the center.

Europe Faces the Reverse Incentive

Europe sits farther from the primary rival.

That means America gets less strategic benefit from European duplication of American commanding industries.

A Japanese shipyard may directly determine the balance near China.

A European cloud company does not.

A Japanese missile factory may strengthen the First Island Chain.

A European alternative to American financial markets mostly reduces American rents.

So Washington's tolerance for allied competition may increasingly depend on the military value of that competition.

That produces the seemingly strange outcome:

America may become more economically permissive toward Japan while becoming more economically demanding toward Europe.

Not because America likes Japan more.

Because Japan is the frontier.

Europe Could Become the Wealthy Rear

Under this system, Europe does not disappear.

It becomes economically indispensable in a different way.

Europe supplies:

• a gigantic affluent market;
• enormous savings;
• scientists;
• specialized industry;
• pharmaceutical capability;
• luxury goods;
• tourism;
• high-value property;
• professional services;
• complementary military capacity.

It may continue attracting foreign capital even while some of its most ambitious scientists and entrepreneurs move into American-controlled institutions.

It could increasingly finance consumption through:

• tourism income;
• foreign investment;
• wealthy immigration;
• property;
• borrowing;
• accumulated assets.

There is no reason such a Europe must become poor.

Many countries already maintain high living standards while owning less of the technological frontier than their consumption levels would suggest.

Europe could remain extraordinarily rich in assets and quality of life.

It could simply become less central to who owns the future.

East Asia Could Become the Creditor

There is another side to this system.

East Asian countries have spent decades accumulating foreign assets.

Japan in particular has one of the world's largest net foreign asset positions.

An aging creditor society eventually gains the ability to transform past industrial surpluses into future investment income.

Instead of every Japanese retiree depending solely on a shrinking Japanese workforce, Japanese institutions can own claims on production elsewhere.

European infrastructure.

American equities.

Foreign railways.

Energy projects.

Factories.

Government bonds.

That creates another possible division of labor.

America specializes increasingly in:

• financial intermediation;
• startups;
• frontier technology;
• platform ownership.

Japan and perhaps other mature East Asian creditor economies specialize more in:

• long-duration capital;
• infrastructure finance;
• foreign-asset income;
• selected strategic high technology.

Europe supplies more:

• labor;
• services;
• tourism;
• property;
• consumption markets.

Again, these are tendencies rather than clean categories.

But the architecture is economically coherent.

The American Empire Could Therefore Become Larger, Not Smaller

This is the largest change in my thinking.

My previous essay argued that America might respond to China by narrowing its empire.

I now think it may instead narrow the number of independent strategic decision-makers inside the empire.

That is different.

Geographically, the system could actually become larger and more integrated.

Economically, Europe remains inside.