Work

The Macroeconomic Impact of Generative AI: Evidence and Projections

Year: 2026
Type: article
Abstract: Generative artificial intelligence (gen-AI) has diffused faster than any previous general-purpose technology, yet its effect on aggregate output remains contested. Forecasts of its ten-year contribution to GDP range from below 1% to around 7%, while early cross-country estimates suggest that measured macroeconomic effects have so far been close to zero. This paper addresses the resulting "micro–macro gap": the disconnect between large, well-identified productivity gains at the task level (commonly 14–56%) and negligible movements in national productivity statistics. We synthesise experimental, firm-level and macroeconomic evidence published between 2023 and 2026, and develop a transparent task-based growth-accounting framework in the tradition of Hulten (1978) and Acemoglu (2024). The framework decomposes aggregate gains into task exposure, adoption, and cost savings, and adds an institutional-readiness parameter that allows projections to differ across advanced, emerging and low-income economies. Calibrated to published parameters, the model reproduces the conservative benchmark of roughly 0.66% cumulative TFP growth for advanced economies, and yields cumulative ten-year GDP gains of 0.99–4.86% for advanced economies, 0.50–2.43% for emerging markets and 0.22–1.05% for low-income countries across three scenarios. The results imply that gen-AI is more likely to widen than to narrow cross-country income gaps unless adoption capacity is deliberately built. We argue that the micro–macro gap is explained mainly by slow diffusion, the need for complementary organisational investment, and mismeasurement of quality improvements, and we propose a cross-country panel design to test these channels as data accumulate.
Author: Devi, Tilka
Language: English
Open Access status: green
OpenAlex ID: W7213578289
FWCI:
Cites:
Cited by:
Related to: 10