How AI Could Change the U.S. Economy by 2030

Economic models of AI are useful not as precise forecasts, but as a way to test several scenarios. In Anthropic’s study, professions are viewed as sets of individual tasks: AI can speed up some of them, automate others, while some still require a human.

Three development scenarios

In the moderate scenario, the impact is compared to the advent of the internet: the economy gets an additional boost, but no dramatic shift. The estimate for U.S. GDP by 2030 is 34,1 trillion dollars, compared with 33,5 trillion without AI’s impact.

In the more serious scenario, AI performs about half of all intellectual work, and GDP reaches 36,3 trillion dollars. The radical scenario assumes that intellectual tasks are almost fully automated. In that scenario, annual growth could reach 15%, and GDP could reach 44,4 trillion dollars.

Economic growth does not mean that all workers’ incomes grow

The most important finding concerns how the gains are distributed. In the extreme scenario, unemployment among intellectual workers could rise to 17,9%, and overall unemployment to 11,9%. At the same time, the model estimates that wages for intellectual professionals fall by about 11,5%, while other workers’ incomes rise by 33,6%.

As a result, the economy becomes wealthier, but the share of the value created shifts from workers to capital owners: in one scenario, the ratio changes from 60/40 to 45,2/54,8.

Why this is not a ready-made forecast

The model cannot know what decisions governments will make, how quickly businesses will adopt AI, whether general-purpose robots will emerge, or what will happen to financial markets. The figures therefore show a range of outcomes under specific assumptions; they do not promise a particular future.

What this means for companies and professionals

Organizations should assess not entire professions, but maps of tasks: where AI speeds up preparation, where it reduces the amount of manual work, and where mistakes are too costly. Professionals would benefit from developing skills in defining tasks, checking results, and working with physical or organizational context—precisely the kinds of elements that are harder to replace with a single model.