The Anthropic Institute published Economic Scenarios for Transformative AI, a framework for 2026-2030 that turns assumptions about how capable AI becomes, how widely it is adopted, how much it raises productivity and how people search for jobs into figures for U.S. gross domestic product (GDP), wages, workers' share of income and unemployment. The authors say the scenarios are not predictions and attach no probabilities to them. In their most extreme path, very fast growth and sharp job losses among knowledge workers happen at the same time, though the paper does not say which path the country will take.
In the model's three highlighted paths for 2030, GDP ends 1.6%, 8.3% and 32.4% above where it would be without AI in the modest, substantial and extreme cases, with yearly growth of 2.4%, 5.4% and 15.4%. These are the model's results given the paper's assumptions, not a single forecast.
Job outcomes differ by type of work and by measure. In the substantial case, employment in knowledge-work jobs, which the paper calls cognitive occupations, is 3.9% below its mid-2026 level, while unemployment across all workers is 4.6%. In the extreme case, knowledge-work employment is 21.5% lower, unemployment among those workers reaches 17.9% and unemployment across all workers reaches 11.9%. The drop in jobs and the unemployment rate measure different things.
Pay splits too. In the extreme case, average wages are 9.7% above the no-AI path, but wages in knowledge-work jobs are 11.5% below it, while wages in other jobs are 33.6% above it. Workers' share of national income falls from 60.0% to 45.2%, even as total pay to workers stays near its no-AI level. In a what-if calculation, the paper estimates that bringing knowledge workers' total pay back to its no-AI level would take payments worth about 9% of GDP; it is not a policy proposal.
When the median answers from an August Anthropic survey of 10,980 U.S. adults are run through the model, the result lands near the substantial path: GDP about 8.6% above the no-AI baseline and unemployment across all workers around 4.6% by 2030. The model leaves out policy responses, business cycles, shocks to demand or financial markets, catastrophic risks and highly capable robots. It is not known which path the U.S. will follow, or which institutions would share out the gains if AI pushes more knowledge workers out of their jobs.