Anthropic Study: AI Could Expand Economy 32.4% While Pushing Unemployment to 11.9%

On September 10, 2026, Anthropic released a study simulating AI's potential economic impact from 2026 to 2030, based on a survey of 10,980 U.S. adults. The company framed the scenarios as stress tests rather than predictions.

In the mildest scenario, AI gradually enters businesses, lifting the U.S. economy by 1.6% relative to a no-AI baseline, with unemployment at 3.9%, close to the baseline rate of 3.8%. In the middle scenario, AI can perform about half of knowledge work, but jobs are not immediately eliminated; the economy grows by 8.3% while unemployment rises to 4.6%.

The extreme scenario assumes rapid AI adoption with limited new white-collar roles. It projects economic growth of 32.4%, but unemployment could jump to 11.9%, and knowledge work positions might shrink by 21.5% from mid-2026 levels. Knowledge worker unemployment could reach 17.9%, with programmers, customer service representatives, and similar roles most affected.

The study also estimates that in the extreme case, labor's share of national income could fall from about 60% to 45.2%, while capital's share rises to 54.8%. Total labor income would increase only 0.5% compared to a no-AI world, while capital income could grow 81.4%, suggesting the gains from AI may not flow evenly to workers.

Most respondents' expectations aligned more with the middle scenario, and Anthropic noted the study does not account for robotics breakthroughs, new employment policies, or economic cycles.