Will AI soon lead to double-digit growth?

AI boosters are getting bolder by the month. Anthropic CEO Dario Amodei has talked about growth of roughly 10% to 15% a year. Leopold Aschenbrenner’s “Situational Awareness” imagines 30% annual growth and beyond. Sholto Douglas went even further, saying people should price in economic doublings in the 2030s, with Elon Musk amplifying it. The authors say these aren’t fringe takes; they’re part of a broader wave of double-digit growth forecasts inside AI circles.

Their argument is not that AI won’t matter. Quite the opposite. They think a capabilities explosion is already underway and will keep going. But they say translating better models into economy-wide GDP growth is a slower, messier business than the hype suggests. The key point is timing: in the next 10-15 years, they think growth above 10% is extremely unlikely, and 4% to 5% would already be huge. Even Anthropic’s own model only gets to double-digit growth in its “extreme” case.

A lot of the article is a warning against reading too much into clean growth theory. In a textbook model, automation can remove labor as a bottleneck, let capital compound faster, and push the economy into explosive territory. The authors show how that can happen with a standard task-based production setup and a Solow-style savings rule. If the share of automatable tasks rises from one-third today to 100% by 2045, their calibration gets GDP growth to about 12.5% by 2035. If saving rises too, growth can reach 15% by 2034 and 20% by 2045.

But theory only gets you so far. The article says those explosive scenarios rely on a stack of assumptions that may not hold together in the real world: fast economy-wide automation, continued high spending on what gets automated, someone to buy all that output, no AI-driven cyber damage, and AI also blowing up R&D. Strip away any one of those and the neat mathematics gets uglier. The authors also argue that GDP may miss some welfare gains anyway, which is another reason the headline numbers can mislead.

Their bigger point is almost embarrassingly simple: people confuse a capability jump with an economy-wide miracle. AI may well be transformative. It just doesn’t follow that the whole economy doubles every year like some sci-fi spreadsheet.