Nouriel Roubini has made a prediction about how societies will pay for themselves once machines do most of the work, and it is worth examining both for its internal logic and for how much weight it deserves.
Speaking to Fortune, the economist argued that "a large chunk of the population will be replaced by AI and robots in the next 20-25 years", and that conventional responses such as raising the retirement age will not be adequate to the scale of it.
The two paths he describes
Roubini frames the response as a choice between two forms of redistribution.
The first is what he calls ex-post distribution: "universal basic income for everybody while they work and once they retire." The state lets ownership stay where it is, taxes the returns, and pays a cash income to everyone.
The second is ex-ante, which he describes as "some form of socialism" in which "the government is going to take over some fraction of the big tech firms." Rather than taxing profits after the fact, the state holds equity and receives the returns directly as an owner. He suggested AI companies have already shown some willingness to share stakes.
Both routes, in his account, arrive at a similar destination: the government would "tax the 'winners' and redistribute the money to everyone else."
The growth assumption doing the work
The part of the argument that deserves the most scrutiny is the funding.
Roubini's projections have growth accelerating from the current 2 to 4 percent to 6 percent by 2040 and 10 percent by 2050. He characterizes his own outlook as optimistic, on the assumption of "machines doing all the work."
That assumption is load-bearing. Universal basic income is arithmetically straightforward but fiscally enormous: a payment large enough to substitute for wages, paid to everyone, is a commitment on the scale of a country's largest existing programs combined. It is affordable in Roubini's telling only because the economy is assumed to be several times more productive. Remove the growth assumption and the policy does not finance itself.
Sustained 10 percent growth in a mature economy would also be historically extraordinary. It has been achieved by economies catching up from a low base, not by frontier economies already at the technological edge.
Universal basic income, briefly explained
A universal basic income is a regular cash payment made to everyone, without a work requirement and without means testing.
The design differs from most existing welfare in two ways. It is unconditional, so it does not taper away as recipients earn more, which removes the effective high marginal tax rates that means-tested benefits can create. And it is cash rather than in-kind support such as food or housing assistance, leaving spending decisions to recipients.
Pilots have been run in several countries, and their results are frequently over-claimed in both directions. They are typically small, time-limited and paid to a subset of a population, which means they cannot capture the economy-wide effects, on labour supply, prices and the tax base, that would determine whether a national scheme worked. They are evidence about recipients, not about macroeconomics.
The counter-argument
Roubini's forecast is one view, and the current data does not yet show the displacement he describes.
The OECD's June 2026 economic outlook found no signs of widespread labour displacement from AI, and noted that job vacancies in AI-exposed industries had risen more than in other sectors. The IMF's work emphasizes that the effects depend on the composition of tasks within occupations rather than on whole jobs disappearing, which implies uneven wage and employment outcomes rather than uniform elimination.
Among economists working directly on this, Daron Acemoglu of MIT has published productivity estimates far below the boom scenarios, in the order of roughly 1 percent added to US GDP over a decade. That is a very different world from 10 percent annual growth, and the gap matters: the fiscal case for a large basic income rests on the optimistic figure.
It is also relevant that Roubini has a long record of pessimistic macroeconomic calls. He is best known for warning about the housing and credit crisis before 2008, a call that proved correct and made his reputation. A forecaster known for warning of disruption should be read carefully when forecasting disruption, in both directions.
What is actually at stake
Strip out the twenty-five-year horizon and a narrower question remains, which is who owns the returns when capital substitutes for labour at scale.
That question does not require artificial general intelligence to become pressing. It is already visible in the concentration of AI infrastructure spending among a handful of firms, and in the political argument over how that gets taxed. Roubini's contribution is less the timeline, which is unfalsifiable in the near term, than the framing: if the returns to labour fall relative to the returns to capital, governments end up choosing between taxing those returns or owning a share of them.
