Anthropic economists sketch three AI futures for 2030
Anthropic’s Economics team has modeled three ways AI could change US growth, jobs and unemployment by 2030, and in an October 3 post on Threads the company invited readers to map their own predictions against them. Whatever answers you give the explorer at anthropic.com/institute/econ-scenarios, it places them in one of those futures, which run from a case Anthropic likens to the impact of the internet to one where annual growth reaches 15.4% by 2030 and economy-wide unemployment 11.9%. And the paper behind the model says “the scenarios are not predictions and we attach no probabilities to them.”
“Economic Scenarios for Transformative AI”, the paper by Anton Korinek, Charles I. Jones, Szymon Sacher, Tess Cotter and Peter McCrory that the explorer is built on, maps paths of AI capability into paths for GDP, the labor share, wages, labor reallocation and unemployment between 2026 and 2030. Each job in it is a bundle of tasks from the US Department of Labor’s O*NET taxonomy, and for each task AI can leave it untouched, augment it, automate it entirely or create new tasks. The economy splits into cognitive occupations (management, professional, sales and office work) and everything else; as AI automates a growing share of cognitive work, productivity rises and displaced workers have to search for jobs in other occupations. The explorer asks, among other things, “Out of every 100 instances of a task AI can do in 2030, how many will AI actually be doing?” and “How many will be fully automated?”
In the modest scenario AI affects about 4% of tasks across the economy by 2030, GDP ends 1.6% higher than without AI, growth reaches 2.4% a year and unemployment sits at 3.9% against the model’s 3.8% baseline. In the substantial one AI becomes capable of half of all knowledge work by 2030 while most knowledge-work tasks are still done without it; growth climbs to 5.4%, GDP lands 8.3% higher and economy-wide unemployment rises only to 4.6%, though cognitive employment falls 3.9% from mid-2026 levels. Cognitive wages there end 0.3% below the no-AI path, while wages in other occupations come out 5.9% higher.
The extreme scenario assumes AI affects roughly half of the tasks performed by cognitive workers in 2025, automates 90% of those affected instances and creates essentially no replacement cognitive tasks; Anthropic wrote that it is “likely driven by recursively self-improving AI systems and a faster rate of AI adoption.” Cognitive employment there falls 21.5% from mid-2026 levels, and workers who began in cognitive occupations face 17.9% unemployment and wages 11.5% below the no-AI path.
Version 1.0 of the explorer leaves out policy responses, business cycles, aggregate demand and financial market disruptions, catastrophic risks and robots that could take over physical work, and it cannot track individual workers through displacement.
About 60 cents of every dollar the US economy produces now goes to workers and 40 cents to capital. In the model labor’s share slips to 59.4% in the modest scenario, 56.1% in the substantial one and 45.2% in the extreme one. In that last case capital income rises more than 80%, because capital becomes useful for more things and its price goes up. GDP finishes 32.4% above the no-AI path, and total labor income barely changes by 2030.
Up next
OpenAI’s agent audit costs over $500,000 a day
More than $500,000 per day is what OpenAI says it spends reviewing 50 petabytes of its agents’ records. The review isn’t finished, so more organizations may learn they were targeted.
Related stories
- Junior technical hires at Anthropic fell from 27% to 9%
- Anthropic bought a biotech with one full-time employee
- A perfect refusal score isn’t a safer model, Anthropic says
- Anthropic touts 94.2% true positives; curl rejected 19 of 34
- Anthropic’s IPO filing says a key safety threshold is met
- 1,000 Claude agents redid the Fermat proof in 28.7 hours
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

