Anthropic Models 2030 AI Scenarios: GDP Up 32.4% in Extreme Case but Wages Fall Over 10%

aigclink · x · 2026-09-10

Anthropic's economics team, led by Korinek and Chad Jones and reviewed by Acemoglu and Autor, released Economic Scenarios for Transformative AI, modeling three futures for the US economy by 2030:

Mild: AI's impact ≈ the internet; GDP +1.6% ($34.1T), barely visible in macro data.
Significant: AI autonomously handles about half of knowledge work but adoption lags; GDP +8.3% ($36.3T), growth rate doubles.
Extreme: AI surpasses humans at nearly all knowledge work, almost fully automated, creating no new human tasks (requires recursive self-improvement); GDP +32.4% ($44.4T), 15% annual growth, economy doubles in 4.5 years.

Key findings: unemployment stays in historical ranges in the mild and significant scenarios; only the extreme case sees mass knowledge-worker displacement. Average wages rise in all three, but the gains go to physical/offline jobs — knowledge-worker wages stagnate in the significant case and fall over 10% by 2030 in the extreme one. Capital's share jumps from 40% to 54.8% in the extreme scenario, with total labor income essentially flat despite explosive growth.

A survey of 10,000+ Americans puts the median expectation near the significant scenario (GDP +10% from AI by 2030, 5% unemployment), with 10% expecting the extreme case. The poster adds four critiques: use tasks rather than jobs as the unit of substitution; data, compute and packaged know-how are the scarcest AI-era assets; the "no new tasks" assumption is the most dubious; and the report ignores robotics, so blue-collar safety only holds if embodied AI doesn't break out.

Original post →

Anthropic Models 2030 AI Scenarios: GDP Up 32.4% in Extreme Case but Wages Fall Over 10%

More from AGI Musings