Anthropic has published a model of the potential economic consequences of artificial intelligence development in the United States through 2030. The analysis matters because its most extreme scenario combines annual GDP growth of as much as 15% with historically high unemployment and worsening conditions for knowledge workers.

The company’s economics team created the Scenario Explorer based on its technical report, Economic Scenarios for Transformative AI. The model treats the economy as a collection of tasks performed by people, software and machines, then assesses whether AI could augment, automate, leave unchanged or create those tasks.

Three scenarios for AI’s economic impact

Under Anthropic’s modest scenario, AI would have an effect roughly comparable to the emergence of the internet. It would provide an additional productivity boost, but economic growth would remain within historical ranges.

In the substantial scenario, AI could perform half of all intellectual work by 2030, mostly autonomously, although companies would not use it for every eligible task. Anthropic projects that the economy would grow at about twice its usual pace.

The extreme scenario assumes that AI becomes more productive than people across the vast majority of intellectual tasks and performs almost all such work autonomously. Anthropic said this outcome would probably require systems capable of recursive self-improvement and rapid AI deployment.

In the substantial scenario, U.S. GDP could reach about $36.3 trillion in 2025 dollars by 2030, compared with $33.5 trillion without AI’s impact. Under the extreme scenario, GDP could rise to $44.4 trillion.

The model estimates that the combined annual value of tasks currently performed by people, machines and software in the U.S. economy exceeds $30 trillion. Changes to individual tasks would therefore affect GDP, employment, wages and the division of income between workers and capital owners.

Anthropic illustrated the model with nursing. AI could help draft discharge instructions, monitor patients remotely or plan care schedules, while physical tasks such as bathing a patient could not be performed by the technology alone. AI could also create responsibilities such as checking automated patient triage or reviewing treatment plans proposed by a model.

Employment and income risks

Anthropic’s model projects that AI would contribute to GDP growth in all three scenarios, but with different labor-market effects. Labor reallocation and unemployment would generally remain within historical ranges under the less disruptive scenarios, while the extreme trajectory could push unemployment to historically high levels.

Knowledge-based occupations face the greatest risk because AI could automate some of their tasks. Programmers, call-center operators and other workers may need to retrain and move into occupations less susceptible to automation, and finding new jobs could take considerable time.

Average wages would rise under all three scenarios, according to Anthropic, but the gains would be uneven. Knowledge workers’ incomes would effectively remain unchanged in the substantial scenario and could fall by more than 10% by 2030 in the extreme scenario, even as demand and wages increase in other occupations.

Large-scale automation could also direct a greater share of income to capital owners. Workers currently receive about 60 cents of every dollar of economic value, compared with 40 cents for capital. In the extreme scenario, the economy would expand significantly, but workers’ share would decline and total labor income would change little by 2030.

Economist Anton Lyubich said Anthropic’s extreme scenario could reduce labor’s income share from 60% to 45%, while unemployment among knowledge workers could reach 17.9%. He also said that achieving the scenario would require a substantial expansion of capital, energy and physical infrastructure, factors that Anthropic’s model does not fully incorporate.

Current evidence remains mixed

Financial Times economics columnist John Burn-Murdoch said current evidence on AI’s employment effects remains mixed. Some studies indicate reduced hiring among young professionals, while companies that are actively adopting AI may be adding more employees.

Anthropic co-founder Jack Clark said major labor-market changes may take time to emerge because businesses typically reconsider their employment structures during large macroeconomic shocks, particularly recessions.

An X user using the pseudonym “Middle Class Dad” highlighted the risk of pressure on office workers’ incomes and a shift in the benefits of growth toward capital owners, AI models and computing infrastructure. Other X users, Jersey Shore Investor and Shikshan Nivesh, said the economy could grow under any scenario, but asset owners could benefit most while workers face uncertainty over their professions.

In an August survey of more than 10,000 Americans, respondents’ typical expectations were closest to Anthropic’s substantial scenario: GDP would be 10% higher by 2030 than without AI, while unemployment would rise to about 5%.

Anthropic cautioned that its model is a simplification and does not account for government responses, economic cycles or financial shocks. The company plans to update the model as new data becomes available.

Source: Incrypted