Anthropic's economics team released the interactive Economic Scenario Explorer, which translates assumptions about AI development into U.S. GDP, employment, and wage forecasts for 2030. Three built-in scenarios produce a range from +1.6% to +32.4% in GDP, and the faster the growth, the harsher the picture for intellectual professions: in the latter two scenarios, wages for mental workers barely grow, and in the most radical scenario they fall by more than 10%.

image

What happened

In September 2026, Anthropic introduced Economic Scenario Explorer version 1.0 — a web tool where sliders for AI capabilities, autonomy, and adoption speed translate into specific macroeconomic figures for the U.S. The model relies on the U.S. Department of Labor's O*NET task taxonomy and the technical report Economic Scenarios for Transformative AI, prepared by Anton Korinek, Chad Jones, Shimon Zaher, Tess Cotter, and Peter McCrorie. The moderate scenario assumes U.S. GDP of about $34.1 trillion by 2030, the significant scenario — $36.3 trillion with growth roughly twice as fast as usual, and the extreme scenario — $44.4 trillion with 15% annual growth, when the economy doubles every 4.5 years.

Context

Economic Scenario Explorer is not a new neural network or a leap in AI capabilities, but a transparent macroeconomic simulation where the main engineering task is the input-output link: parameters of "AI capabilities, autonomy, adoption speed" are converted through the task level of the O*NET taxonomy into GDP, unemployment, wages by professional groups, and the labor and capital shares of income. The key and most controversial mechanism is built into the extreme scenario: AI autonomously performs almost all intellectual tasks and creates virtually no new tasks for humans, so the labor share falls from about 60% to 45.2%, and the capital share rises to 54.8%. A point of public comparison is a survey of more than 10,000 Americans conducted in August 2026: the expectations of a typical respondent were close to the significant scenario — GDP about 10% higher, unemployment around 5%.

Why this matters for the industry

For the industry, a quantitative bridge from AI parameters to macro indicators has emerged: GDP, unemployment, wages by professional groups, and the labor and capital shares are linked to specific assumptions about capabilities, autonomy, and adoption speed. In fast-growth scenarios, benefits concentrate with capital and non-intellectual labor, while mental workers face wage stagnation and rising unemployment — this directly affects employee reskilling strategies and AI-labor pricing, and also fuels the debate over the distribution of automation income. Founders and corporate teams now have a citable tool for conversations with investors and corporate clients, and the "capabilities, autonomy, adoption" scale may become a common language for public forecasts, simplifying the comparison of claims from different labs. There is no direct impact on deployed products, however: this is a web page without an API or pricing, but in the coming months, expect scenario figures to be cited in industry reports and responsive scenario models from other teams.

Why this matters for users

The tool is already available: any reader can open Economic Scenario Explorer, set their own forecasts for AI capabilities, autonomy, and adoption, and immediately see what the U.S. economy will look like by 2030, then compare their answer with the expectations of more than 10,000 surveyed Americans. For those whose work involves intellectual labor, this is a clear fork in the road: how much their personal career trajectory depends on which scenario materializes in practice. For those who want to check the calculation logic, a technical report in PDF format with model details is available, and the page honestly marks the tool as version 1.0 and a simplification — this helps to soberly assess the figures, not taking them as an exact forecast.

What is still unknown / limitations

The quality of the model as a predictor is unconfirmed: public sources do not describe calibration against historical data, sensitivity analysis, or backtesting. The extreme scenario is an extrapolation of parameters into a domain without empirical anchors: the U.S. economy has never observed such growth rates, so it should be read as a boundary condition, not a probabilistic forecast. The strongest and most controversial assumption is the fall in the labor share without a compensating increase in demand for human tasks. The expectations of surveyed Americans are a useful calibration benchmark, but not a validation standard; if model confirmation does not appear, the figures will remain a framework for discussion, not a forecast.

Sources

Author

Look at AI, editorial team