Credit: Anthropic Anthropic has built a model of how AI might reshape the US economy by 2030. Its most extreme case describes an economy roughly a third larger than it would otherwise be. That same case carries knowledge-work unemployment near 18 percent. The company published the model on 10 September as an interactive scenario explorer, based on a technical report, “Economic Scenarios for Transformative AI” (Korinek et al., 2026).
It does not predict which outcome will happen. The scenario explorer treats every job as a bundle of tasks. AI can leave a task alone, augment it, automate it, or add new ones. The model then scales those effects across the economy under different assumptions about how capable AI becomes and how fast it is adopted.
The result is three AI economic scenarios, which Anthropic labels modest, substantial and extreme. Three futures In the modest scenario, AI has roughly the impact the internet did. US GDP in 2030 reaches $34.1 trillion, about 1.6 percent above a no-AI path. The effect is hard to see in the macroeconomic data.
In the substantial scenario, AI does half of all knowledge work by 2030, most of it autonomously, though it is not adopted for all of that work. The economy grows at twice its normal rate to $36.3 trillion. Wages for knowledge workers stay roughly flat, and other workers see gains. This is the scenario that echoes findings that AI has been hitting paychecks rather than payrolls.
The extreme scenario is the outlier. In it, AI is more productive than humans at most knowledge-work tasks, and does nearly all of them autonomously. Anthropic says that path would likely require recursively self-improving systems, adopted quickly. Annual GDP growth reaches 15 percent, which would double the economy every four and a half years.
That takes 2030 GDP to $44.4 trillion. Anthropic says this would leave society far richer, but with many fewer knowledge-work jobs. Where the pain lands The model’s harder findings sit in that extreme case. Unemployment among knowledge workers rises to about 18 percent by 2030, against under 4 percent for other workers.
Anthropic says overall unemployment would climb beyond typical recessionary levels. Displaced coders and call-centre agents would have to move into occupations less exposed to AI, such as electrician or nurse. The model notes that switching occupation is slow, and often leaves people out of work for a long time. It follows warnings, such as one from Cognizant, about how much work AI could expose.
Wages split the same way. In the substantial scenario, pay for knowledge workers is essentially flat. In the extreme scenario, it falls by more than 10 percent by 2030. Pay for other workers rises by about a third, and the average rises by roughly 10 percent.
Anthropic also finds the share of income going to workers falls as the share going to capital rises. Of each dollar the economy produces, about 60 cents goes to labour today. In the extreme scenario, that drops to about 45 cents, with the rest going to the owners of capital. Anthropic surveyed more than 10,000 Americans in August about their expectations.
The typical respondent’s answers implied something close to the substantial scenario. That meant GDP about 10 percent higher by 2030, and unemployment around 5 percent. About 10 percent of respondents gave answers in line with the extreme case. What the model leaves out Anthropic is explicit that the model is a simplification.
It leaves out policy responses, business cycles, and the aggregate demand effects of the data-centre buildout. The company says it also leaves out possible catastrophic risks, and scenarios in which humanity builds hyper-capable robots. The technology outlet Gizmodo noted the timing. The model was released a day after Anthropic’s own alignment lead, Evan Hubinger, said there was a greater than 10 percent chance AI could kill all humans within a decade.
The economic model does not attempt to price that outcome. The authors say they are not forecasting. In general, they note, AI experts expect the technology to spread faster than economists do. Anthropic co-founder Jack Clark placed himself in the middle.
He told NPR the technology would keep improving fast, but would spread through the economy more slowly than many expect. If growth did reach the extreme range, he added, the extra tax revenue would give policymakers options that are unimaginable today. Anton Korinek, who leads Anthropic’s transformative-AI economics work, said the impact depends on adoption, because a capable AI that nobody uses has no economic effect. Economists push back The model landed in a live argument about whether AI can drive double-digit growth at all.
Anthropic chief executive Dario Amodei has said AI could push growth to something like 10 to 15 percent a year, a figure that sits alongside the company’s pitch of a $30tn market. The economists Ben Moll, of the London School of Economics, and Alex Imas, of the University of Chicago, argued that such rates in the next 10 to 15 years are extremely unlikely. They put a more reasonable baseline at 4 to 5 percent, which they note would still be very large. They have placed a public bet that US per-capita real GDP growth will stay below 15 percent every year through 2033.
Their case is that a jump in AI capability does not automatically become growth. Most of the economy is physical rather than cognitive work. Spending shifts toward whatever stays scarce and human. Machines and data centres take years to build.
Anthropic’s own model reaches double digits only in its extreme scenario; its modest and substantial cases stay in the single digits. The reviewers Anthropic credited include the economists Daron Acemoglu and David Autor, and Moll himself, who Anthropic says pushed it to add the channel by which more of the gains flow to capital. Anthropic says it will keep revising the model as the evidence develops.
