Anthropic Scenarios for the U.S. Economy: Strong Growth with Rising Income Inequality

Anthropic released the "AI Economy Explorer" on September 9, 2026, an interactive model examining three scenarios for the impact of artificial intelligence on the U.S. economy through 2030. The central finding: even with strong economic growth, capital owners could benefit significantly more than workers—particularly knowledge workers.

Anthropic's economics team developed the model together with economists Anton Korinek and Chad Jones. The approach breaks down occupations into individual tasks and estimates whether AI replaces, improves, or creates new work for these tasks. The result shows in all scenarios with substantial AI adoption a widening income gap between capital and labor.

Three scenarios with different economic strength

The most conservative scenario compares AI's impact to that of the internet. By 2030, gross domestic product reaches 1.6 percent above the level without AI; unemployment remains within historical ranges, with no major labor market disruptions.

In the middle scenario, AI takes over about half of all knowledge-based tasks largely autonomously but is not deployed across all possible applications. The consequences through 2030: U.S. GDP rises to $36 trillion, 8.3 percent above the baseline scenario. Annual economic growth reaches 5.4 percent. Unemployment remains relatively low at around 4.6 percent, with knowledge worker employment declining by about 4 percent.

Wage development reveals a stark divide: while average wages rise by 2.1 percent, opposing trends underlie this figure. Knowledge workers see a decline of 0.3 percent, while other occupations gain 5.9 percent. Labor's share of national income falls from 60 to 56.1 percent, with capital's share rising accordingly.

Extreme scenario: doubling roughly every five years

The third scenario models rapid self-improvement of AI systems. AI becomes more productive than humans in nearly all knowledge-based tasks, operates with recursive self-optimization, is deployed almost fully autonomously, and generates hardly any new human knowledge work. The consequences through 2030:

• Annual economic growth accelerates to 15 percent

• GDP reaches $44.4 trillion, roughly one-third above the baseline scenario; reporting cites 32 to 33 percent

• At this pace, economic output would theoretically double roughly every five years

• Knowledge worker wages fall by more than 10 percent; reported figures show 11.5 percent

• The unemployment rate rises to recession-like 17.9 percent

• Labor's share of national income drops to 45.2 percent, with capital's share rising accordingly to 54.8 percent

Distribution at the center

The key insight of the model lies not in the observation of rising economic output, but in its distribution. In the fastest growth scenario, a significantly smaller share of income flows to workers. Anthropic emphasizes: this is "not simply more of the same"—higher GDP growth does not automatically benefit all workers equally.

For capital owners and investors, the model suggests that even broad-based AI growth could favor business owners and shareholders more than employees. The greatest burden falls on knowledge workers, whose wages stagnate in the middle scenario and fall by over 10 percent in the extreme case.

Interactive model rather than fixed forecast

The AI Economy Explorer presents occupations as "bundles of tasks." For each task, AI can leave it unchanged, assist in its completion, automate it, or create new work. Users can vary assumptions about AI capability, adoption rates, autonomy levels, productivity gains, and task-level effects.

Anthropic makes clear that the company does not forecast a single outcome. The tool enables examination of different assumptions and their consequences. The scenarios presented are conditional models under different performance and adoption assumptions, not predictions of actual developments.

Survey: Americans expect productivity surge and disruptions

Alongside the model, Anthropic conducted a survey of roughly 10,000 Americans about their AI expectations. On average, respondents expect a significant productivity and growth surge, but also considerable disruptions for workers in AI-exposed sectors—a "mixed outcome" of real productivity gains and burdens for workers in AI-sensitive jobs.

Observers noted that Anthropic, as an organization developing AI systems, invests substantial resources in modeling negative side effects and planning for various economic outcomes—including scenarios with significant job losses and rising inequality.