Anthropic Put Numbers on the Scenario Where Knowledge Work Stops Paying
The Anthropic Economics team published an interactive model of what AI does to the US economy, and unlike most such exercises it commits to arithmetic. Three scenarios, each with a GDP number for 2030 and an unemployment number attached.
Modest: AI lands like the internet did, growth stays on trend, GDP up 1.6 percent to $34.1 trillion. Substantial: AI handles half of knowledge work by 2030, GDP up 8.3 percent to $36.3 trillion, unemployment around 5 percent, and knowledge-worker wages flat. Extreme: AI beats humans at most knowledge tasks and adoption is fast, GDP up 32.4 percent to $44.4 trillion, unemployment at recession highs, knowledge-worker wages down more than 10 percent. Labor's share of GDP falls from about 60 percent today to 45.2 percent, with the difference going to capital.
Read that last line twice, because it is the only one that matters. The scenario with the most growth is the scenario where workers do worst. Anthropic says it out loud: in the extreme case the challenge is not achieving growth, it is making sure the benefits are broadly shared. A frontier lab publishing a model in which its own best-case technical outcome produces historic unemployment is not a normal corporate blog post.
Two details give it teeth. They surveyed over 10,000 Americans and found public expectations line up with the substantial scenario, meaning ordinary people are already pricing in half of knowledge work going away and mostly not panicking. And the model treats jobs as bundles of tasks that get augmented, automated, replaced, or newly created, which is the only framing that survives contact with what agents actually do. Put it beside Anthropic's own IPO revenue math (https://clauday.com/article/5c3ad89d-2337-471f-ba32-b1eb4ba55ecb) and you get a company that has now published both the number it needs investors to believe and the number that says the growth may not reach wages: https://www.anthropic.com/institute/econ-scenarios
← Back to all articles
Modest: AI lands like the internet did, growth stays on trend, GDP up 1.6 percent to $34.1 trillion. Substantial: AI handles half of knowledge work by 2030, GDP up 8.3 percent to $36.3 trillion, unemployment around 5 percent, and knowledge-worker wages flat. Extreme: AI beats humans at most knowledge tasks and adoption is fast, GDP up 32.4 percent to $44.4 trillion, unemployment at recession highs, knowledge-worker wages down more than 10 percent. Labor's share of GDP falls from about 60 percent today to 45.2 percent, with the difference going to capital.
Read that last line twice, because it is the only one that matters. The scenario with the most growth is the scenario where workers do worst. Anthropic says it out loud: in the extreme case the challenge is not achieving growth, it is making sure the benefits are broadly shared. A frontier lab publishing a model in which its own best-case technical outcome produces historic unemployment is not a normal corporate blog post.
Two details give it teeth. They surveyed over 10,000 Americans and found public expectations line up with the substantial scenario, meaning ordinary people are already pricing in half of knowledge work going away and mostly not panicking. And the model treats jobs as bundles of tasks that get augmented, automated, replaced, or newly created, which is the only framing that survives contact with what agents actually do. Put it beside Anthropic's own IPO revenue math (https://clauday.com/article/5c3ad89d-2337-471f-ba32-b1eb4ba55ecb) and you get a company that has now published both the number it needs investors to believe and the number that says the growth may not reach wages: https://www.anthropic.com/institute/econ-scenarios
Comments