Will AI Soon Deliver Double-Digit Economic Growth? The Case for No

Summary

Ben Moll and Alex Imas examine predictions that AI could lift advanced economies to annual GDP growth of 10%, 30%, or even 100%. They argue that such growth is theoretically possible: in a Solow-style model, automation raises the share of tasks performed by machines, removes labor as a bottleneck, and can create feedback between machine accumulation, output, and AI-assisted research. Their baseline view is nevertheless that growth above 10% is extremely unlikely over the next 10–15 years, with roughly 4–5% annual growth already representing a major increase. The explosive scenario requires fast, economy-wide automation, including physical work; continued spending on goods and services even after they become cheap; enough demand and investment to absorb the extra output; no major AI-enabled cyber incidents; and an R&D feedback loop in which AI fully replaces human research bottlenecks. The authors argue that diffusion takes time, many jobs are messy or relational, and energy, construction, land, permitting, and chip-supply constraints limit the speed at which machines can be built. Falling wages could also create political resistance and weaken demand, although the authors say evidence for the demand channel remains limited. Automating R&D may raise GDP’s level without increasing long-run growth if humans remain bottlenecks or ideas face diminishing returns, while new discoveries can take decades to reach measured output. They also note that GDP may understate AI’s welfare gains when prices and spending fall, citing an estimate that consumers receive far more value from chatbots than they pay. Their conclusion is a large, gradual increase in economic output rather than an imminent growth explosion.