Anthropic's economic model projects AI's impact on US GDP, wages and unemployment by 2030

A scenario explorer for the AI economy

Anthropic's economics team has released an interactive model that projects how AI could affect jobs, growth and unemployment in the United States over the next several years. The tool builds on a technical report, Economic Scenarios for Transformative AI (Korinek et al., 2026), and lets users enter their own expectations about future AI capabilities, how widely it will be adopted, how autonomous it will be, how much it boosts productivity, and how quickly displaced workers can find new jobs. It then shows what the economy might look like in 2030 if those assumptions hold. According to Anthropic, the aim is to give society better visibility into possible futures so the transition can be managed to everyone's benefit. The company already measures current AI usage through its Economic Index; this project looks ahead.

Growth in every scenario, on very different scales

The model's headline finding is that AI raises GDP in all of its scenarios, but the magnitude varies enormously. Measured in 2025 dollars for 2030, compared with an economy without AI:

• Modest scenario: GDP up 1.6%, reaching $34.1 trillion.
• Substantial scenario: up 8.3%, reaching $36.3 trillion.
• Extreme scenario: up 32.4%, reaching $44.4 trillion.

The extreme case corresponds to growth faster than anything in recorded economic history, and it comes with serious downsides for workers.

Knowledge workers bear the displacement

In most scenarios, job churn and unemployment stay within ranges the US has seen before. The exception is the extreme scenario: if AI reaches recursive self-improvement and is adopted rapidly, Anthropic's model shows unemployment spiking to historic levels, with affected workers jobless for prolonged periods.

The mechanism is occupational reallocation. As AI automates knowledge work, jobs in affected occupations shrink as a share of employment between 2026 and 2030, while less AI-exposed occupations grow. At the individual level, Anthropic gives the example of coders and call-center agents moving into roles such as electrician or nurse. But switching occupations is hard: people may not want to change, retraining takes time, and landing a new job is not guaranteed. The more switching a scenario requires, the more people end up between jobs. Unemployment rises in knowledge occupations while falling in others.

Wages rise on average, but not for everyone

Across all three scenarios average wages increase, yet the gains concentrate outside knowledge work. In the substantial scenario, knowledge-worker wages are essentially flat by 2030; in the extreme scenario they fall by more than 10%. Anthropic attributes the divergence to demand: as AI improves, demand for human knowledge work drops, while AI-driven productivity raises demand elsewhere. Its example: faster design and permitting for physical infrastructure could mean more construction projects and therefore rising construction wages.

The model also tracks how the economic pie is divided. Today roughly 60 cents of every dollar produced goes to workers and 40 cents to capital. In the substantial and extreme scenarios the labor share falls noticeably, meaning more of the gains flow to the owners of capital — even as society as a whole becomes far wealthier and average wages rise.

What the public expects

To ground the explorer, Anthropic surveyed more than 10,000 Americans in August about AI capabilities, adoption and the ease of changing occupations. The typical respondent's answers imply outcomes close to the substantial scenario: GDP roughly 10% higher by 2030 than it would be without AI, and an overall unemployment rate around 5%. About one in ten respondents hold views in line with the extreme scenario.

Why it matters

Much of the AI-and-jobs debate runs on anecdote; this is a concrete, quantified model from a lab with a direct stake in the outcome, and it sharpens the real question. Growth appears in every scenario — the contested issues are distributional: whether knowledge-worker wages hold up, whether labor's share of income erodes toward capital, and whether retraining can keep pace with displacement. Anthropic itself cautions that the model is not a complete map of reality, and it is a projection tool rather than a forecast. Even so, it gives policymakers, employers and workers specific numbers to argue over and, over the next few years, to check reality against.