The AI boom has created extraordinary wealth for a select group of people: founders, investors and highly skilled employees of AI companies. Consider Anthropic. If the frontier AI lab manages to hit an IPO valuation of $2 trillion, then each of its seven founders could be worth between $27 billion and $29 billion, at least on paper. In comparison, John D. Rockefeller was worth $1.4 billion when he died in 1937 (though the US economy was much smaller then, with nominal GDP of just $93 billion).

For every ‘AI have’ the new economy is minting, there are hundreds of have-nots—even in the US. Since OpenAI’s ChatGPT launched in 2022, AI stocks in the US are estimated to have created as much as $18 trillion in new wealth. But according to US Fed data on stock ownership, the bottom 50% of the population captured just 1% of this wealth. The broader economic gains are equally skewed. An IMF working paper published in July 2026 estimated the annual labour-cost equivalent of time saved through AI at $2.7 trillion, or 3.4% of the combined GDP of the surveyed countries. Using data from 86 countries, researchers found that 96% of this estimated value was concentrated in high-income countries. Within countries, the benefits were also disproportionately concentrated in higher-paid occupations.

Of course, the $2.7 trillion is not income actually received by workers or profits realised by companies. It is an estimate of the value of time saved. But it offers an early indication of who is benefitting from AI-driven productivity gains.

A New Divide
For much of human history, wealth has been concentrated among those who controlled the most valuable resources. Kings and their armies were supreme when brute power mattered. The industrial revolution tilted the scales in favour of people who owned capital, factories and machines. The internet revolution produced another set of winners: founders and owners of digital platforms such as Google and Meta, along with their knowledge workers.

The AI revolution could concentrate wealth further. The divide will not simply be between those who use AI and those who don’t. It will be between those who own and control the technology and those who depend on it. Countries that own the models, intellectual property, computing infrastructure and platforms will capture a larger share of the economic returns. Others may become more productive by using AI, but they will capture a smaller share of the benefits than the ‘owners’ of artificial intelligence.

At the heart of this divide is human labour and intelligence. Earlier technological revolutions increased productivity, but did not remove people from the equation as a source of labour or cognition. AI threatens to do both. If companies can keep increasing output and profits with fewer people, what happens to those whose skills are no longer in demand? And what happens to countries—like India—whose economic advantage lies in labour arbitrage in services?

The latest Periodic Labour Force Survey puts unemployment among 15–29-year-olds at 9.9% in 2025. Among young graduates, it was 22.3%; among postgraduates, 24.7%. For India’s 12 million youngsters entering the job market every year, according to World Bank estimates, AI could make things worse.

China, a much bigger AI power than India, offers another warning. Its official unemployment rate among urban residents aged 16–24, excluding students, was 18% in September 2026. Some economists argue that the broader extent of youth joblessness is considerably higher if those who have stopped looking for work—including the tang ping, or lying flat, youngsters who have opted for limited work—are included.

The Governability Test
Recent youth-led protests in Europe, Latin America, and South Asia have been driven by grievances ranging from unemployment and unaffordable housing to failing education systems, corruption and a sense that governments are doing nothing about these problems. These demonstrations are not a consequence of AI—at least not yet. But they reveal a growing frustration with the gap between what young people expect from their governments and what they actually get. AI could amplify that frustration.

Imagine an economy in which companies report surging profits, stock markets dazzle and productivity improves, but a growing number of young people struggle to find meaningful work. Governments will find it difficult to explain how an economy doing so well is failing to provide livelihoods for so many.

This is where the problem becomes one of governability. Governments are elected on the promise of improving people’s lives. But what happens when the forces shaping the economy are increasingly beyond their control? A government cannot compel a company to hire more people when technology allows it to do the same work with fewer employees. Nor can it retrain millions of workers overnight for jobs that may not yet exist.

Related to this is the fiscal problem. In OECD countries, personal income taxes and social security contributions account for about half of tax revenue, while corporate income taxes contribute less than 4%. If AI allows businesses to generate more value with less labour, then governments may have to rethink how they raise revenue and fund social protection. This does not mean corporate tax receipts will necessarily fall. But it raises the question of whether tax systems built around employment income can adequately capture the gains accruing to corporate entities.

To act effectively, governments need to know which jobs are disappearing, which are emerging and whether employers are actually hiring people with the right skills. They must also worry about entry-level jobs. If AI replaces graduate hires, then how will they develop the experience required for middle or senior roles? Worse, taking up a suboptimal job (say, a gig) could condemn them to a lifetime of poorly paid jobs and relative poverty.

Governments have competing responsibilities to juggle with. They have no choice but to encourage AI adoption to remain competitive; protect workers’ livelihoods; invest in a social safety net; and prevent powerful tech companies from undermining their own authority. Voters, however, will expect results, not another set of long-range promises. If governments repeatedly fail to deliver, frustration could feed anti-establishment movements and make people more receptive to leaders promising quick solutions, even at the expense of democratic institutions. This would not be unprecedented: the world has already seen Hitler and Mussolini do that in the 20th century.

For governments, slowing down AI development is not necessarily the answer. As US President Donald Trump said recently, “Whoever wins AI wins”. But allowing adoption to race ahead without considering its consequences could be equally damaging. Governments must ensure that the gains are shared, workers have a stake in the new economy, and companies controlling its essential infrastructure remain accountable.

After all, any government that’s of the people, by the people, and for the people will be hard pressed to justify a state of affairs in which the economy is richer and more productive, but its people are impoverished.