Google DeepMind’s Alex Imas says AI agents may yield only 4% to 5% GDP growth

Millions of AI agents working around the clock sounds like the setup for an economic explosion. Alex Imas thinks the explosion may look more like a firm push.

Imas, director of AGI economics at Google DeepMind, said in a November 7, 2026 appearance on the MTS podcast that widespread deployment of autonomous AI agents is likely to produce only a modest bump in annual GDP growth. His projection for developed economies: 4% to 5%.

Why cheaper stuff means smaller GDP gains

The core of Imas’s argument is about prices, not productivity. Automation makes goods and services cheaper to produce, and those savings tend to show up as lower prices for buyers.

GDP measures the dollar value of what an economy produces. So when the cost of something collapses, its contribution to GDP can shrink even if people are getting far more of it.

Imas points to agriculture as the historical proof. Farming’s share of GDP fell from roughly 40% to 2%, even as output surged. His point is that AI could follow the same script. Real output and overall welfare could rise substantially, while the headline growth figure stays relatively grounded.

A counterweight to the explosive-growth camp

Imas’s framing pushes back on more aggressive forecasts of AI-driven growth from figures such as Will MacAskill and Samuel Altman. Those predictions tend to assume that demand and prices hold up as AI floods the economy with new capacity.

Imas questions that assumption. If efficiency gains mostly flow into lower prices, the macroeconomic scoreboard may look calmer than the technology itself would suggest.

He also offered a reality check on the present. In remarks on October 14, 2026, Imas said that US economic growth of about 2% already reflects contributions from AI and productivity gains. Without those factors, he suggested, even sustaining that modest pace would likely be difficult.

Who Imas is, and what he has been working on

Imas joined Google DeepMind as director of AGI economics in May 2026. He did so while on leave from the University of Chicago.

His podcast comments fit with his broader research at the lab. On October 29, 2026, he released a paper titled “Organizing Intelligence,” co-authored with Andrew Koh and Shane Legg, a DeepMind co-founder.

The paper examines how coordination among AI systems shapes economic structures. Its central claim is that how AI is organized matters significantly more for the economy than what any single AI system can do on its own.

Imas has emphasized the structural shifts under way in the economy. He argues that economists need models that track overall welfare, not just raw production capacity.