Will You Escape the Permanent Underclass?
AI will destroy the value of most labour, creating a vast, permanent underclass. There's no solution in sight.
We’re heading towards an economic transformation that will fundamentally reshape social relations. Within a few decades, most human work will become unnecessary as software and robots systematically replace workers across nearly all industries. This will create a society of three distinct classes: a class of people who own capital, a servant class providing low-wage personal services, and a massive unemployed population with no economic role. The last two groups together form what we might call the “permanent underclass”—people who are structurally blocked from building any wealth and thus locked into poverty.
This forecast lines up with economist Daniel Susskind’s analysis in his 2020 book A World Without Work. Susskind argues that technological unemployment—once dismissed as a myth—is becoming real as AI systems learn to do more and more tasks. Unlike past technological revolutions that destroyed old jobs but created new ones, AI threatens to be different. It’s getting good at everything humans can do, and it’s improving faster than humans can retrain.
Basic Economics: Too Many Workers, Not Enough Jobs
The mechanism here is simple supply and demand. As AI systems get better at doing cognitive and physical work, companies need fewer human workers. But the number of people looking for work stays the same or grows. When you have way more supply than demand, prices drop. In labour markets, that price is your wage.
This is already happening in obvious places. Software eliminated millions of jobs in data entry, basic accounting, travel booking, and customer service. Now we’re seeing the second wave hit knowledge workers: legal document review, medical diagnosis, financial analysis, writing, and even programming. The third wave—robots powered by AI—will hit warehousing, manufacturing, driving, construction, farming and more.
Susskind points out that this is different from normal unemployment. Workers can’t just “retrain” when AI learns faster than humans do. A truck driver learning to code in 2025 might find AI writes better code by 2027. A doctor spending ten years training might discover that computer vision systems beat humans before she’s done. AI is now improving faster than humans can learn new skills.
Three Classes
As this plays out, society will split into three classes with completely different relationships to wealth:
The Capital Owners will be people who own significant capital, including AI systems, robots, natural resources and real estate. Their wealth will grow exponentially because their assets produce value without needing to pay workers. The owner of an AI-powered logistics company can scale up without hiring more people. The owners of AI that writes software or provides medical advice capture almost all the money these systems make.
The Servant Class will provide personal services that are hard to automate: home healthcare, childcare, elderly care, personal training, house cleaning, and other face-to-face work. These jobs stick around not because they’re impossible to automate, but because it’s not worth automating them or because people prefer human interaction. But with millions of unemployed people competing for these jobs, wages will crash to subsistence levels. We already see this with gig economy workers—too many drivers and delivery people means terrible pay. Some people argue that enough new jobs will emerge for everyone, but demand for labour is actually not infinite, and such demand will be largely met by AI-powered supply.
The Unemployed will be the third and possibly the largest class—people for whom no economically useful work exists. Machines can do everything they used to do. They’ll depend entirely on whatever help they can get: government payments, charity, or family. With no income and no savings, they’re stuck in permanent poverty.
The servant class and the unemployed together make up the permanent underclass. One group technically has jobs, but both share something important: they can barely build wealth, if any. When you barely earn enough to survive, there’s nothing left to save or invest. The gap between this underclass and the capital owners will be enormous. Capital owners will command armies of tireless AI workers while the underclass fights over scraps.
Most Economic Activity Will Happen Between Capital, Not People
Maybe the biggest change—one people often miss—is that most valuable economic activity won’t involve human work at all. Instead, the bulk of transactions will happen between different pieces of capital owned by the wealthy.
Think about how an AI-powered economy works. An AI system owned by Company A produces software. Company B buys this software, and its AI systems use it to run robots in a factory. Company C’s self-driving trucks transport what the factory makes. Company D’s AI platform sells it online. Company E’s drones deliver it. Bank F’s AI processes the payments. Fund G’s algorithms invest the money in other automated businesses.
Value gets created and money changes hands at every step—but almost no humans are involved. The software engineers who used to write code are gone. The factory workers are obsolete. The truck drivers, store clerks, bank tellers—all replaced. What’s left is capital doing business with capital, and the profits go to the humans who own that capital, who also are the ones buying the resulting goods and services.
This breaks completely with how economies have always worked. Even in the most mechanised industries, human workers were essential. A car factory in 1970 needed huge amounts of machinery and equipment, but it also needed thousands of workers to run that equipment. Those workers earned wages and spent them on goods, which created demand for more production. Money circulated from capital to labour and back.
In an AI economy, this circulation stops. When capital only needs other capital to make things, the permanent underclass becomes pointless. They’re not needed as workers—AI and robots handle production.
Rich people buy homes (built by robots, designed by AI, sold through automated systems), vacation properties (managed by AI), luxury goods (made in automated factories), and premium services (increasingly delivered by AI rather than human servants). They invest in each other’s companies, trade financial assets, and develop more sophisticated capital. The whole economy can work—actually, it produces more than ever before—with the permanent underclass on the sidelines.
This is why the permanent underclass really is permanent. They’re not temporarily out of work, waiting for the next boom. They’re locked out of an economy that doesn’t need them. There’s no way back in because there’s no role for them. Money flows from capital to capital, and if you don’t own capital, you have nothing to offer.
Redistribution Won’t Work—Governments Are Losing Their Power to Tax
Many progressives—including Susskind—think wealth redistribution can solve this. Tax the rich, fund universal basic income, create massive social programs. Sounds good in theory. But it assumes governments will keep their power to tax rich people and mobile capital.
That assumption is falling apart. In their 1997 book The Sovereign Individual, James Dale Davidson and William Rees-Mogg predicted this. They argued that the information revolution would transform the relationship between citizens and states. States historically got their power from controlling territory. In farming or industrial economies, wealth was tied to physical things: land, factories, buildings. These couldn’t move. A medieval king could tax farmers because farms stay put. A modern government could tax manufacturers because factories are stationary. But in an information economy, wealth becomes digital and portable.
Look at a capital owner in the AI age: much of her wealth is equity in a multitude of companies, AI systems, cryptocurrency, and digital assets. None of this needs to be in any specific place. She can manage her businesses from anywhere with internet. Her assets can be registered in whatever country has the best tax treatment. If her home country tries aggressive redistribution, she just moves herself and her money somewhere friendlier.
The capital-to-capital economy makes this even easier. When economic activity happens between capital assets rather than employers and employees, ties to physical location mostly disappear. An AI trading stocks doesn’t need to be “located” anywhere—it runs on servers that can be in any country. A robotic factory can easily be moved to another jurisdiction because no new workers have to be hired. And when capital is the only source of income, and its owners are more mobile, real estate will become almost worthless in most places while becoming worth a lot in places where the owners of capital move to. The money these capital-to-capital transactions generate can flow through whichever countries treat it best.
Tax Competition: A Race to the Bottom
This creates what economists call “tax competition”—countries competing to attract mobile capital by offering low taxes. We see it accelerating now. Ireland, Singapore, Switzerland, and Caribbean tax havens offer low corporate taxes, minimal capital gains taxes, and friendly treatment of intellectual property. Countries like Italy now offer special tax regimes for the rich. They do this because it works. Rich people and their money really will move, and countries that don’t compete lose out.
AI will make this much more intense. When capital becomes incredibly productive and completely mobile, and becomes the only source of income for its owners, the incentive to shop around for the best tax deal becomes overwhelming. Why would a millionaire stay in a country with 70% wealth taxes when she can move to Dubai, pay zero income tax, and run exactly the same businesses?
The usual response is that governments can just prevent capital flight with controls or exit taxes. But as Davidson and Rees-Mogg explain, these controls get harder to enforce in a digital world. Cryptocurrency makes money flows hard to track. Intellectual property can be transferred to offshore companies. Rich people can give up citizenship (thousands of Americans already do this every year to escape U.S. taxes).
Also, harsh capital controls backfire. If a country makes it really hard for money to leave, money just never comes in the first place. Why would a company set up in a place with strict controls when dozens of other countries offer easy capital movement? The result is economic stagnation—the opposite of what you need to fund big social programs.
The capital-to-capital economy makes tax enforcement even harder. When value was created through visible employment—companies hiring workers, paying wages, making products—tax authorities could track it fairly easily. Payroll taxes, corporate taxes, and sales taxes all created natural enforcement points. But when AI systems owned by offshore entities do business with other AI systems owned by different offshore entities, with money flowing through complex corporate structures spanning multiple countries, enforcement becomes nearly impossible without perfect international cooperation—which will never happen given that countries compete on taxes.
The Math Doesn’t Work
This creates an impossible problem for redistribution. Supporting a permanent underclass of maybe 60-70% of the population would require extracting huge amounts from the capital-owning class. But capital owners are the most mobile people. They’re globally connected, digitally skilled, and can afford sophisticated tax planning.
The likely outcome is that redistribution defeats itself. High-tax places will drive away their richest people and businesses, leaving a smaller tax base to support more dependent people. This creates a death spiral: higher taxes cause more exits, which requires even higher taxes on whoever’s left, causing more exits. Eventually, the only people left are those too poor to leave—not exactly a tax base that can fund generous programs.
We already see this inside the United States. California, New York, and Illinois have lost wealthy residents to Florida, Texas, and Nevada—states with no income tax. The wealthy call these moves “lifestyle choices,” but the tax savings are massive. When Tesla moved its headquarters from California to Texas, it wasn’t about lifestyle—it was about saving hundreds of millions in taxes.
Scale this to the global level and the problem gets severe. There’s no world government that can enforce uniform taxes. Instead there are nearly 200 competing countries, many actively marketing themselves as havens for capital and rich people. The UAE advertises zero income tax. Portugal and Italy offer special tax deals for wealthy immigrants. Estonia has pioneered digital residency and crypto-friendly rules. These aren’t exceptions—they’re growth industries.
The Sovereign Individual
Davidson and Rees-Mogg predicted the information age would create “sovereign individuals”—people whose skills, wealth, and digital assets make them essentially independent of any particular country. These people will have more in common with medieval merchants or Renaissance bankers than with modern citizens. Their loyalty to countries will be transactional: they’ll pay for services received and leave when better deals appear elsewhere.
AI will massively expand this group. Anyone who successfully builds significant capital becomes potentially sovereign. A software engineer with major equity in an AI company, an entrepreneur who builds an automated business, an investor with a portfolio of stocks—all these people can exit high-tax countries.
Historically, unhappy citizens had two options: “voice” (political participation to change things) or “exit” (leaving). Throughout the 1900s, exit was hard and voice seemed more practical. Emigrating meant learning new languages, abandoning professional networks, and losing cultural connections. Most people stayed and tried to influence policy.
But technology has made exit much easier. English is now used globally. Professional networks exist online. Culture is available everywhere. Remote work eliminates the need to be physically present. For the wealthy and skilled, exit has become easier than voice. Why spend years lobbying for lower taxes when you can just move next month?
This destroys the political foundation for redistribution. In a democracy, redistribution requires that the wealthy stick around long enough to be outvoted by everyone else. But if the wealthy can leave before the vote happens, democracy won’t make a difference. You can’t redistribute wealth that’s already gone.
Worse, the capital-to-capital economy means the wealthy don’t even need local workers or customers. A 1900s factory owner might hesitate to leave because his factory needed local workers and his products needed local buyers. But the AI-era capital owner needs neither. His AI systems can work from anywhere, serve global markets, and mostly do business with other capital. He has no economic ties keeping her anywhere—only tax bills pushing him to leave.
Geographic Splitting
The result won’t be generous redistribution everywhere. Instead, we’ll see radical geographic splitting. Some places will try aggressive redistribution—and become poverty traps as capital flees and economic activity collapses. Other places will embrace low taxes and light regulation—and become havens for the wealthy, gleaming cities surrounded by poverty as the servant class clusters around rich enclaves.
The permanent underclass will concentrate in declining high-tax regions, dependent on shrinking government payments funded by a vanishing tax base. Meanwhile, capital owners will spread across favourable countries, their wealth compounding in tax-efficient structures while their AI systems do global business and mostly transact with each other in the capital-to-capital economy.
This geographic sorting will feed on itself. Once the wealthy cluster in low-tax havens, network effects kick in: the best schools, hospitals, culture, and opportunities concentrate where rich people live. Talented ambitious people will migrate toward these centres, further draining high-tax regions. The permanent underclass will be left behind in places with crumbling infrastructure, failing institutions, and fiscal crisis.
The permanent underclass in these declining regions won’t just be economically irrelevant—they’ll be completely disconnected from where actual economic activity happens. The real economy—the capital-to-capital transactions creating most value—will happen in global networks centred on low-tax places. The declining regions will be economic backwaters, kept alive only by whatever minimal aid the international community provides to prevent humanitarian disaster.
The Window Is Closing
This leads to an urgent conclusion: the only escape from the permanent underclass is building capital now, before AI takes over most work and before governments try desperate redistribution that accelerates capital flight. There’s a narrow window—maybe one to three decades—when human work still pays enough that middle-class people can save, and when you can still build capital without triggering confiscatory taxes.
People who can save and invest substantial capital in this window won’t just own shares in the AI-powered economy—they’ll also be able to become sovereign individuals, free to choose their country and optimise their taxes. They’ll own rental properties, shares in tech companies, stakes in automated businesses, or other capital that generates returns independent of their work or location. Critically, they’ll own capital that participates in the capital-to-capital economy—assets that create value by doing business with other capital owned by other members of the capital-owning class.
People who can’t build capital during this period will be triply trapped: permanently locked out of ownership as wages collapse, geographically stuck in declining high-tax places because they can’t afford to move, and completely disconnected from the capital-to-capital economy where most value gets created. They’ll depend on whatever failing social programs these places can afford with their shrinking tax bases, while the real economy functions elsewhere without them.
The race to build capital is already happening, but most people don’t see it. The professional who spends her whole paycheck on a middle-class lifestyle is making a catastrophic mistake—she’s choosing current consumption over future survival and access to the capital-to-capital economy. The worker who thinks democratic politics will protect him through redistribution is unaware of capital mobility and the collapse of state taxing power.
A New Social Reality?
In this new class society, we’ll probably have a new social reality. The few people with capital or well-paid jobs will be very popular. If you’re one of the happy few and almost everyone else is struggling economically, many people will want to be your friend or your lover. In history, the rich had such a position in society not only because of how rich they were, but also how poor others were. Will society move back to such a reality, where money gives access to much more than only goods and services?
A Silver Lining?
AI also promises to radically reduce the cost of goods. Total economic output will be larger than ever before. Perhaps what now would be considered a meagre means of subsistence will afford a good life in the age of AI abundance. In that case, a universal basic income, however modest, might actually give people a good means of subsistence. In any case, it’s not wise to take the risk. Instead, it’s better to accumulate as much capital as possible today. If AI-powered productivity makes this unnecessary in the future, then at least you can live a better life than you’d be able to otherwise. Under AI abundance, that little capital you might save now could buy you a lot.
Conclusion
The AI revolution isn’t mainly about technology—it’s about class formation in a post-national world where most economic activity happens between pieces of capital rather than involving human work. We’re probably watching the creation of a permanent underclass: billions of people who have no economic role, can’t build wealth, and are trapped in declining places as capital and opportunity leave. They’ll be spectators to an economy that works without them, watching from outside as capital does business with capital in networks they can’t access or influence.
The comfortable progressive story—that democratic governments will tax AI profits and redistribute to support displaced workers—ignores how capital mobility works in the digital age. As The Sovereign Individual predicted, we’re entering an era where states lose their ability to tax mobile capital. Tax competition will intensify, redistribution will fail, and places will split between wealthy havens where the capital-to-capital economy thrives and poor welfare states where the permanent underclass survives on meagre handouts.
For individuals, the stakes are extreme. Your career choices, your savings rate, your investment strategy, and maybe your citizenship—these aren’t just financial decisions. They’re survival decisions about which side of history’s greatest economic divide you’ll be on, whether you’ll be free to choose where you live or trapped in declining regions, and most fundamentally, whether you’ll own capital that participates in the capital-to-capital economy or be permanently locked out of where real value gets created.
The window for escaping the permanent underclass is closing. People who see this reality—who build capital aggressively, develop portable skills, position themselves to become sovereign individuals, and acquire stakes in the capital-to-capital economy—will survive the transition with their freedom and prosperity intact. Those who don’t will spend their lives serving those who did, trapped in whichever declining place they’re stuck in, dependent on whatever inadequate programs can be funded by a disappearing tax base, completely disconnected from an economy that doesn’t need them. The future isn’t a question. The only question is whether you’ll build enough capital, and position yourself strategically enough, to escape it.