In every industrial revolution, cities were the winners. Steam built Manchester. Electricity built Detroit. The Internet built San Francisco.

But this time, something different is happening: AI doesn’t scale by adding people. It scales by deleting them.

And that turns every major metropolis into a ticking bomb.


1. The End of the 20-Million-Person City

A city of 10 million needs an ecosystem of redundancy—accountants for accountants, managers for managers, lawyers reviewing lawyers. These aren’t parasites; they’re the circulatory system of complexity.

Generative AI just discovered how to compress that entire layer into compute cycles.

Roughly half of all urban service employment—the invisible bureaucracy that keeps skyscrapers lit—is pure coordination. Replace that, and the modern metropolis loses its blood pressure.


2. The False Hope of “New AI Jobs”

The comforting story is: “Sure, some jobs die, but new ones appear.”

That logic worked when new jobs required more people—factory lines, call centers, data labeling.

But the post-LLM world doesn’t work that way. And it doesn't take many to scale down cities enormously.

One prompt engineer with an AI can coordinate what a 2-person marketing agency did. That's already 50% cut in employment. Imagine that across all service sectors.

A 50-person law firm becomes a 25-person oversight cell plus an API.

Imagine there're 1000 dominant companies with 9,000 specialized vertical in service industry from legal to finance to marketing. Think of it like divorce law, or French B2B marketing. Each has 5 companies for natural competition and each company employs 100 people with highly leveraged AI workflow.

In total, that makes it 5 million. There'll be 5 million well paying jobs in the AI age —not enough to sustain the current service-based urban economy.

Those jobs will be concentrated in a few metropolis even accelearting the metropolis wealth concentration in the future.

To support a metropolis of 10 million, you need roughly 4–5 million active workers generating taxes, demand, and momentum.


3. The Political Physics of Collapse

Now layer in fiscal reality.

Metros like New York, London, Paris, San Francisco are built atop countries of high leverage—mortgages, municipal bonds, consumption taxes. When 30–50 percent of white-collar wages vanish, these cities don’t just shrink; they default in slow motion.

The irony? The countries housing these megacities are already over-leveraged.

They can’t absorb ten years of unemployment without social fracture. That'll trigger enormous political violence.

And when politics breaks, capital flees—including the AI capital.

So the same cities that invent the tools of automation could be the first casualties of their success.

Country / Region Net Investment Position Sign / Implication
United States – US$ 26.23 trillion (end-2024) (Bureau of Economic Analysis) Very large negative position (net debtor)
Japan + US$ 3.48 trillion (Dec 2024) (CEIC Data) Net creditor position
China + US$ 3.18 trillion (as of Sep 2024) (CEIC Data) Net creditor position
Singapore + US$ 0.87 trillion (Sep 2024) (CEIC Data) Modest positive NIIP
France – US$ 0.797 trillion (Sep 2024) (CEIC Data) Net debtor
United Kingdom – US$ 1.121 trillion (Sep 2024) (CEIC Data) Net debtor

4. The Great Divergence: Stability Becomes the New Silicon Valley

Where does the capital go?

To balance-sheet nations that can endure a decade of dislocation without rioting.

Tokyo — immense domestic savings, cultural patience, zero populism.

Shenzhen / Hong Kong — command-economy coordination and manufacturing synergy.

Singapore — technocratic neutrality, fiscal reserves the size of GDP.

During the transition with mass job replacement over the decades, these cities/countries can let half their service workforce disappear and keep the lights on using their financial capital buffer.

And that stability itself becomes a magnet for AI infrastructure, funds, and talent away from the debtor countries boosting the AI related  employment and taxable income.

It’s not about who built the first model anymore; it’s about who can survive the transition without burning down their institutions.

Those cities will be the receiving cluster of well-paying jobs until AI first, AI integrated small cities will be built from the ground up (close to the size of 100,000). To maintain fully functioning capacity, service job automation is not sufficient. The full scale robotics and self-enclosed energy source must be available by then. Those cities will be estimated to appear from 2050s.


5. 🧳 The Myth of the Digital-Nomad Future

Most tech commentary still dreams in micro-romanticism: one person, one laptop, sipping matcha in a green city park, working “from anywhere.”

That vision mistakes individual autonomy for civilizational structure.

Yes, a single creator can build billion-dollar leverage from a laptop—but the global economy still runs on interlocking high-trust, high-bandwidth coordination between humans. That is the source of creativity.

When 99 percent of low-value work disappears, the remaining value becomes even more hierarchical and concentrated.

We’re not heading toward a flat world of ten million digital nomads.
We’re heading toward three to five hyper-metropolises—each housing perhaps 100,000 to 200,000 high-earning global professionals who operate the world’s AI infrastructure.

While there are 500 cities in the world (irony of Fortune 500) with the population size of 1 million plus, 90% of them will lose competitiveness following the path of Detroit 1950s. They'll be left with large pools of ghetto from lack of taxable income.

The rest of the planet becomes the residential hinterland of those cores: pleasant, remote, but economically peripheral.

The next century’s “cosmopolitan elite” won’t live anywhere.

They’ll live only where civilization itself is still operational.


6. The Math of Concentration

Let’s do the math bluntly.

Global white-collar employment: ~1.8 billion

Replace 50 % → 900 million jobs gone

Of the remainder, < 100 million are true “AI-leveraged” roles

Those jobs cluster in perhaps three to five politically stable cities

If those ten metros house 5 % of humanity yet generate 70 % of global GDP by 2045, you’ve effectively turned the planet into a network of compute monasteries surrounded by suburbs of history.


7. What Thriving Looks Like After the Purge

Forget the idea of empty towers and ghost offices. That’s for the mediocre metropolis that failed in competitive edge.

The survivors — Tokyo, Shenzhen, Singapore — won’t shrink; they’ll accelerate.

When global productivity collapses into a few stable hubs, these cities become gravitational cores.

More AI means more data centers, energy grids, and transit arteries.

Skylines rise higher, metros run deeper, and every square meter hums with computation.

Their per-capita GDP will look like entire nations compressed into 40 square kilometers.

While weaker cities de-urbanize, the winners go vertical — a fusion of finance, energy, and cognition.

They won’t be post-industrial ruins; they’ll be hyper-industrial intelligences.

The city doesn’t fade. It concentrates — until it becomes the operating system of civilization.

8. The Real Lesson

AI won’t just destroy jobs; it’ll force Darwinism onto global metropolis.

The ones that survive will not be the most innovative—they’ll be the most self-controlled.

Fiscal sobriety will matter more than startup density.

Political composure will matter more than valuation.

If the 20th century crowned the cities that created jobs, the 21st will crown the few that can lose them gracefully.


TL;DR

AI isn’t liberating humanity—it’s compressing civilization.
The next global capitals won’t be digital-nomad paradises,
but fortified oases of political stability where a few hundred thousand people
run the world’s machines.