If Robots Do All the Work, Who Gets Paid?

In January 2024, the International Monetary Fund estimated that almost 40% of jobs worldwide are exposed to AI, and about 60% in rich economies such as America and Western Europe. Exposed doesn't mean lost: the IMF expected about half of those jobs to be harmed, and the other half to become more productive.¹

So far, the evidence is mixed. Studies this year have found no clear rise in unemployment among workers most exposed to AI.² But one group is feeling it: young people looking for their first job. Researchers at Stanford, using payroll data for millions of American workers, found that since AI chatbots arrived, employment of 22- to 25-year-olds in the jobs most exposed to AI has fallen by about 16% compared with other jobs, while older workers in the same jobs have held steady.³

Robots are further behind. Factory arms have welded and painted for decades, but robots that walk and handle objects like a person are new. This year, a few hundred are doing real work in factories and warehouses. That's a start, not a revolution.

But the direction is clear. AI is reaching office work first. Robots are starting on the physical work that's left. If both keep improving, the question here stops being hypothetical.

Economies haven't always worked the way ours does. Each system grew up around whatever produced wealth at the time, and whoever owned that thing held the power.

In medieval Europe, wealth was land. Lords owned it. Peasants were allowed a strip of it to feed their families, and in return they paid the lord: by working his fields for free, often a few days every week, or by handing over part of their harvest. That system began to break when labour became scarce: after the Black Death killed perhaps a third of Europe's people in the late 1340s, surviving workers could demand more.

×

Then wealth came from trade. From the 1500s, Europe's rulers believed a nation's wealth was the gold and silver in its treasury. To pile it up, a country had to sell more to other countries than it bought from them. So states taxed imports, and took colonies, which supplied cheap raw materials like cotton and sugar and were made to buy finished goods from home. Monarchs gave favoured merchants the sole right to trade with whole regions of the world: the East India Company, founded in 1600, held England's monopoly on trade with Asia, and later ruled much of India.⁴ Economists call this system mercantilism. The wealth belonged to the crown and the merchants it chose.

Then came the factory. From around 1760, the Industrial Revolution made machines the main source of wealth: capital. Whoever owned the mill owned what it made, and most people sold the one thing they had, their labour, for a wage. In 1776, Adam Smith argued against the mercantilists: a nation's wealth wasn't its gold but what its people could produce, and free markets, with each person pursuing their own interest, would make nations rich. That's capitalism.⁵

Not everyone agreed it was fair. In 1848, Karl Marx and Friedrich Engels argued that workers should own the factories themselves. In 1917, the Russian Revolution tried it, with the state owning industry on the workers' behalf and planners deciding what to make. It built steel and tanks, but it fell behind in making what people wanted, and in 1991 the Soviet Union collapsed.

Most of the West chose a middle road. After 1945, Western Europe built welfare states: public health care, pensions and benefits for everyone, paid for by taxes. Britain created the National Health Service, commonly called the NHS, and took coal, rail and steel into public ownership; in Scandinavia, high taxes paid for some of the most generous welfare states in the world; and America, with a smaller state, kept the pensions and protections of the 1930s. Most business stayed private. Economists call these mixed economies: part market, part state.

From 1979, Margaret Thatcher in Britain and Ronald Reagan in America swung back towards free markets: selling state companies, cutting regulation and taxes. Governments of both parties largely kept that model for the next 30 years. And from 1978, China built something else again: markets and private companies, with a state that keeps control of the banks and the biggest industries.

The pattern is the point. When what makes wealth changes, so does who holds power, and the system around it changes too, usually after a fight. If machines that think and move become the main source of wealth, the question of who owns them comes round again.

Even Marx saw it coming. In notes written in 1858, he imagined production run almost entirely by machines, with workers standing beside them, watching over them, and argued that a system built on paying for people's time couldn't survive it.⁶

But people have predicted the end of work before, and they've been wrong.

Between 1811 and 1816, textile workers in the Midlands and the North of England smashed the machines that were replacing their skills. They were called Luddites. Parliament made machine-breaking punishable by death. The weavers did lose out, but over the century, far more jobs appeared than disappeared.⁷

×

In 1930, the British economist John Maynard Keynes gave the fear a name:

"We are being afflicted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come — namely, technological unemployment." — John Maynard Keynes, Economic Possibilities for our Grandchildren, 1930⁸

He meant machines saving labour faster than we can find new uses for it. But he saw it as a passing phase. He predicted that within a hundred years, people would be so much richer that a fifteen-hour working week might be enough. He was right that we'd be far richer. He was wrong about the hours.⁸

In 1964, a group of academics and activists sent President Johnson a memorandum warning that automation would soon break the link between work and income, and calling for a guaranteed income for every American. Employment kept rising through the rest of the decade.⁹

Why do the warnings keep failing? Because technology creates work as well as destroying it. Economists at MIT found that about 60% of American jobs in 2018 were in kinds of work that didn't exist in 1940, from software developers to fitness trainers.¹⁰

But there's another side to that record. Of everything an economy earns, part is paid out as wages and part goes to owners, as profits, rent and interest. Even as jobs kept coming, the workers' part has shrunk in most rich countries since around 1980, and the IMF linked about half of that fall to technology. More of what the economy makes has gone to owners. Machines haven't ended work. But they have shifted who gets paid.¹¹¹²

And what owners have, they keep. In America, the richest 1% now hold about a third of all household wealth, the highest share since the Federal Reserve's records began in 1989, up from less than a quarter in 1990.¹³ Worldwide, the richest one and a half percent of adults, the millionaires, hold nearly half of all wealth.¹⁴ In Britain, the richest tenth of households hold 43%, while the poorer half hold less than a tenth.¹⁵

Here's why some think this time is different. Earlier machines replaced muscle and routine, and people moved into work that needed thinking: managing, designing, caring, selling. AI is aimed at thinking itself. And robots are aimed at the physical work that was left. If both succeed, there may be nowhere left to move to.

There's a second difference. Capitalism is supposed to correct itself through competition: when a company gets too comfortable, a newcomer with a better idea can take it on. But competing takes capital. To start a factory in 1800, you needed a building and some machines. To build the most powerful AI today, you need data centres and chips on a vast scale. This year, Microsoft, Google's owner Alphabet, Amazon and Meta are expected to invest around 700 billion dollars between them, most of it in AI.¹⁶ If the machines that do the work cost that much to build, very few can afford to own them, and very few can challenge those who do.

At the same time, many of those companies are shrinking their workforce. On the 26th of October, Microsoft offered voluntary redundancy to about 60,000 of its staff, around a quarter of its workforce. It gave no reason, but reports have linked it to the cost of building its data centres.¹⁷ Across the tech industry, more than 650,000 job cuts have been recorded since the start of 2023, according to the tracker Layoffs.fyi.¹⁸ Not all of that is down to AI: many companies hired heavily during the pandemic and have been cutting back since. But the direction of the money is clear: more into machines, less into people.

Against that: smaller companies keep releasing capable AI models at a fraction of the cost, some free for anyone to use and build on, and today's giants were themselves upstarts not long ago. Whether AI ends up owned by a few or by many is still open.

There's a third argument, and it goes to the heart of how capitalism works. Companies need customers. If one company replaces its workers with machines, it saves money. But if every company does it, the people who used to be workers have no wages to spend. So who buys what the machines make?

That isn't just theory. In the Great Depression, a version of it happened: as people lost their jobs, they stopped spending, so companies sold less and laid off more, and the spiral fed itself. By 1933, about a quarter of American workers were out of work.¹⁹ Keynes's answer was that when people can't spend, governments should, to get the economy moving again.

×

Some analysts point to a way out for the owners: sell to each other. In America, the richest tenth of households already account for about half of all consumer spending, up from about a third in the 1990s, according to Moody's.²⁰ An economy can keep running on fewer, richer customers: more luxury goods, fewer ordinary ones. But it's a narrower, more fragile economy, and it widens the gap between rich and poor further.

And a widening wealth gap tends to bring pressure for change. The Depression gave America the New Deal, including its first national old-age pensions and unemployment insurance, in 1935.²¹ In Britain, the hardship of the 1930s and the war fed the vote for the welfare state in 1945. It's one reason some of the people building AI say a basic income will be necessary. As Elon Musk put it in 2017: "I don't think we're going to have a choice."²²

Free-market thinkers answer all three arguments the same way: the fear that machines will leave people with no work, and no wages, is as old as the Luddites, and it has been wrong every time. Marc Andreessen, a Silicon Valley investor who backs technology start-ups, calls it the "lump of labour fallacy": the mistaken idea that there's a fixed amount of work to go round. In his view, when machines make things cheaper, people spend what they save on something else, and that creates new jobs we can't yet imagine.²³

They add that machines make things cheaper to produce, and competition between companies passes that on in lower prices, so whatever money people have goes further. Over the long run, that is what has happened. Prices have risen, but pay has risen faster, so most things take far fewer hours of work to afford. In the 1950s, families across the West typically spent between a fifth and a third of their budget on food. Today it's more like a tenth.²⁴ The last few years have gone the other way: since 2021, food and energy prices rose faster than pay, which is why so many people feel worse off. And that's the test for the free-market answer: cheaper goods only help if prices fall faster than people's incomes do.

Others sit in between. Daron Acemoglu, an MIT economist who shared the 2024 Nobel Prize, estimates that over the next ten years AI will add only a little to productivity, how much each worker produces. He argues that what it does to jobs depends on how companies use it: to replace people, or to make them better at their work.²⁵

Nobody knows yet who's right. But suppose the machines really can do almost everything. Then we're back to the question this piece started with: who owns them?

There are, broadly, four answers, plus a fifth that says no answer is needed. Here's each, in its supporters' words, and its weakness.

1. Platform lords

The first is that things carry on as they're going. The Greek economist and former finance minister Yanis Varoufakis argues we're already past capitalism, into what he calls techno-feudalism. Tech platforms like Amazon and Apple don't so much compete for profit as charge rent for access, like lords charging peasants to work their land. Apple, for example, takes up to 30% of what app makers earn through its App Store. If the same few companies own the robots, he argues, everyone else becomes a tenant.²⁶ Critics answer that this is still capitalism, and that platforms do fall: MySpace and Nokia once looked unbeatable.

2. Pay everyone

The second is to pay everyone an income, whether they work or not: a universal basic income. Some of the people building the machines say it's coming. In September, at the White House, Elon Musk said the likely outcome is "an age of abundance", where "we don't have just universal basic income; we have universal high income", because AI and robots will produce so much.²⁷ Economists have questioned how it would be paid for.

In 2021, Sam Altman, the head of OpenAI, proposed one way to pay for it: an American Equity Fund. Every year, the largest companies would hand over 2.5% of their value to the fund, in shares, and landowners would pay 2.5% of the value of their land. Every adult would receive a yearly payment from the fund.²⁸

Would people stop working? A three-year American study, funded partly by Altman, gave a thousand people $1,000 a month with no conditions. Most kept working: on average they worked about an hour less a week than a comparison group, and spent the money mostly on basics like food, rent and transport.²⁹ The weakness: a country that brings in a tax like this on its own may find that its companies move somewhere that doesn't tax them as much.

3. A citizens' fund

The third is for the public to own a share directly. It already happens. Norway put its oil money into a fund now worth more than two trillion dollars, which owns a small slice of almost every large company in the world, on behalf of every Norwegian, now and in the future.³⁰ Alaska pays every resident a dividend from its oil fund each year: this year, $1,000, plus $200 towards energy bills.³¹ The weakness: in Norway and Alaska, the oil belonged to the public from the start. A country that wanted a stake in AI companies would have to buy it, or take it.

4. Public ownership

The fourth is common ownership: the machines owned by the public, through the state or communities, rather than by private companies. The British writer Aaron Bastani argues for what he calls "fully automated luxury communism": once machines can make almost anything cheaply, the things everyone needs could be provided to all.³²

Critics point to the 20th century. In the Soviet Union, officials in Moscow decided what every factory made, how much of it, and what it cost: millions of decisions, made without prices to show what people actually wanted. They often got it wrong. Shops ran out of basics while warehouses filled with goods nobody bought, and families waited years for a car. In 1945, the economist Friedrich Hayek argued that no planner could ever know enough. Markets work, he said, because prices gather up what millions of people know and want: when something is scarce, its price rises, and makers respond.³³

Supporters answer that the problem was information, and machines are now very good at it. AI already predicts what millions of people will buy, which is how online shops keep their shelves stocked. Critics reply that information was only half the problem: without competition, nobody had a reason to make things better.

Or: nothing needs to change

And the fifth answer is that no new system is needed. Prices fall, new jobs appear, and markets spread the gains, as they have since the Luddites. The record supports the first two: we are far richer than any generation before us, and new kinds of work have kept appearing. It's less clear on the third: since 1980, a growing share of the gains has gone to owners rather than workers.¹¹

So, does capitalism end when the robots arrive? Not on its own. The robots are capital. If they do the work, the system doesn't disappear. It concentrates, unless something changes who owns them.

That's what history suggests. Every time the source of wealth changed, from land to trade to factories, the system changed with it. But it was never automatic, and it was rarely quick, because those who held the old source of wealth didn't give it up willingly.

In 1381, English peasants rose up demanding an end to serfdom, the rules that tied them to their lord's land and his fields. The revolt was crushed, and serfdom faded only slowly over the next two centuries.³⁴ From 1815, Britain's landowners kept taxes on imported grain, the Corn Laws, which protected their profits and raised the price of bread. They held on to them for 31 years, until Parliament repealed them in 1846.³⁵

×

The bloodiest change was Russia's. Russia had been an autocracy: the Tsar ruled with almost unlimited power, and until 1861 millions of peasants were serfs, bound to their lords' land.³⁶ The revolutions of 1917 swept that order away, and the Bolsheviks set out to build communism: a system with no private owners, where the state holds the land and the factories in the people's name. The civil war that followed, until 1922, killed millions, in the fighting and from hunger and disease.³⁷

But change has also come peacefully, sometimes because those in power gave ground to avoid something worse. In the 1880s, Germany's conservative chancellor, Otto von Bismarck, created the world's first national sickness insurance and old-age pensions, partly to win workers away from the socialists.³⁸ And in 1945, Britain voted in a government that built its welfare state, including the NHS.³⁹

×

So, if machines do the work, who gets paid? Today, most people are paid for their work. If machines one day do most of it, that stops being true. What people get will depend instead on what they own, or on the rules: a share in a fund, a payment from the government funded by a tax on the machines, or nothing at all.

The free-market thinkers may be right that new work will come, as it always has. If they're wrong, the pressure will build the way it has before: a wider gap between rich and poor, fewer people with money to spend, and louder calls for change. History suggests that what happens next depends on whether those who own the machines share what they make, or have to be made to.

Those rules are already being written: in how we tax technology companies, in who owns them, and in whether countries take a public stake in what they make. Unlike the peasants of 1381, most of us now get a vote on them.