Artificial intelligence could push annual US economic growth above 15 per cent by 2030 while leaving nearly one in five knowledge workers unemployed, according to a new economic model from Anthropic that attempts to put numbers around the upheaval AI may bring to the labour market.
The company behind Claude has modelled three possible paths for the US economy over the next four years, ranging from AI having an impact comparable with earlier technological shifts to an extreme scenario in which rapidly advancing systems transform much of knowledge work.
The findings present a striking contradiction. In Anthropic’s most extreme scenario, the United States becomes substantially richer, but many of the workers most exposed to AI are worse off.
Annual GDP growth reaches 15.4 per cent and the economy is 32.4 per cent larger in 2030 than it would have been without AI, reaching about US$44.4 trillion in 2025 dollars. At that pace, the economy would roughly double every four and a half years.
At the same time, employment among knowledge workers falls 21.5 per cent. Their unemployment rate rises to 17.9 per cent, while unemployment across the entire workforce reaches 11.9 per cent. Wages for knowledge workers are 11.5 per cent lower than under the model’s no-AI path.
Workers outside those occupations fare very differently. Their wages rise 33.6 per cent in the extreme scenario, illustrating how unevenly the economic effects of AI could be distributed.
Perhaps more significant is what happens to the division of the nation’s income.
Labour’s share falls from 60 per cent to 45.2 per cent, while capital’s share rises to 54.8 per cent. Capital income is more than 80 per cent higher than it would have been without AI. In other words, the model produces an economy with far greater output, but a much larger proportion of that prosperity flows to owners of capital rather than workers.
Anthropic is careful not to describe this as a prediction. The researchers state that “the scenarios are not predictions and we attach no probabilities to them”.
Instead, the model asks what could happen under different assumptions about how capable AI becomes, how quickly companies adopt it, how autonomously it can perform tasks, how much it raises productivity and how easily displaced workers can move into other occupations.
The less dramatic scenarios look very different.
Under Anthropic’s modest scenario, AI adds 1.6 per cent to GDP relative to a world without AI by 2030, with annual economic growth reaching 2.4 per cent. Knowledge-work employment falls by just 0.5 per cent and overall unemployment is largely unchanged.
The substantial scenario assumes AI becomes capable of performing about half of knowledge work, although it is not adopted across all of those tasks. Annual GDP growth reaches 5.4 per cent, GDP is 8.3 per cent higher than the no-AI path and knowledge-work employment falls 3.9 per cent. Unemployment among those workers rises to 4.5 per cent.
It is that middle path that comes closest to what Americans themselves expect.
Anthropic surveyed 10,980 US adults in August about AI capabilities, adoption, autonomy, productivity and the difficulty workers might face finding new employment. The typical respondent’s expectations produced an outcome close to the substantial scenario, with GDP around 10 per cent higher by 2030 and overall unemployment at roughly 5 per cent. About 10 per cent of respondents held expectations consistent with the extreme case.
Anthropic chief executive Dario Amodei has previously offered a more severe assessment of what may be coming. In May 2025, he warned that AI could eliminate as many as half of entry-level white-collar jobs within one to five years and push US unemployment to between 10 and 20 per cent.
One of the biggest uncertainties is not simply how capable the technology becomes, but how quickly businesses actually use it.
“If AI can do amazing things but nobody uses it, then it’s not going to have an economic impact,” Anton Korinek, who leads Anthropic’s transformative AI economic studies, said.
That distinction matters. A system may technically be able to automate a task long before companies restructure jobs, change workflows or invest enough to make widespread automation commercially viable.
Anthropic also acknowledges the limits of its model. It does not attempt to capture every force that could shape the economy, including government policy responses, business cycles, financial-market disruptions, changes in aggregate demand or catastrophic AI risks. Nor does it model a future dominated by highly capable physical robots.
The exercise nevertheless puts numbers around one of the central economic questions surrounding AI: whether extraordinary improvements in productivity will translate into broadly shared prosperity.
Anthropic’s most aggressive scenario suggests those two outcomes are far from the same thing. An economy can become dramatically larger while the workers who once performed much of its most valuable professional work face falling wages, fewer jobs and a shrinking share of the wealth being created.
Read also Giannis Antetokounmpo Buys US$13.5 Million Miami Mansion After Heat Move
