While artificial intelligence (AI) is delivering noticeable efficiency gains in many workplaces, economists remain divided over how much this will translate into economy-wide growth. Two recent Nobel Prize winners in economics — Daron Acemoğlu (2024 laureate) and Philippe Aghion (the most recent laureate) — have offered markedly different assessments of AI’s potential contribution to productivity and GDP.
What experiments and surveys show
Generative AI applications, such as conversational chatbots including ChatGPT, have produced measurable performance improvements in specific tasks. Several recent experimental studies report productivity or speed gains between 14% and 56% in areas like programming, copywriting, and customer support for users of AI compared with non-users.
However, these AI-boostable activities represent only a small share of total economic activity. AI adoption is most feasible in knowledge-intensive services — notably finance, IT, telecommunications, and other office-based sectors — and widespread business adoption depends on cost-effectiveness, adequate digital infrastructure, skills, and often organizational restructuring and retraining. Those requirements and data needs create substantial uncertainty about when and how task-level productivity gains will aggregate into macroeconomic growth.
Two contrasting forecasts
Daron Acemoğlu takes a cautious view: he estimates that AI might add at most around 0.1 percentage point to annual productivity growth (output per worker) over the next ten years. By contrast, Philippe Aghion is considerably more optimistic and projects roughly a 1 percentage point annual uplift. Other economists consider even faster, potentially explosive growth possible. Small differences in annual growth can lead to very different long-term outcomes because of cumulative effects.
Adoption pace and OECD scenarios
The speed of AI diffusion is a major source of uncertainty. For prior technologies, ten years after introduction roughly 23% of firms used electricity regularly, while about 40% used the internet and PCs. For the current generation of AI, in the third year after the user-friendly chatbot breakthrough, only about 10% of firms use AI regularly in their main activities.
The OECD ran scenarios based on diffusion paths similar to electricity or to the internet and PCs, combined with AI capability assumptions. Under an electricity-like diffusion, AI could raise macro productivity by about 0.4 percentage points per year; under an internet/PC-like diffusion, the boost could be about 0.9 percentage points per year. The more optimistic scenario is comparable to the additional growth economists attributed to the internet boom in the United States between 1995 and 2005.
Importantly, AI’s macro effects are not expected to materialize as an immediate surge: widespread adoption and successful implementation take time and resources, so impacts are likely to accumulate gradually as the technology spreads across different sectors.
Regional gaps and policy implications
Europe’s experience offers a cautionary tale: the continent captured a relatively small share of the internet/PC-driven tech upswing because adoption was slower and narrower across sectors while production concentrated in the United States. The current generation of AI again delivers its strongest gains in knowledge-intensive services, which have a smaller weight in many European economies compared with the United States. Data also show that a larger fraction of U.S. firms are already using AI in core activities.
The OECD authors and the Defacto analysis emphasize that the magnitude of AI’s macroeconomic benefits will depend heavily on diffusion speed and institutional conditions. For Europe the central question is not whether AI can raise growth, but whether Europe can secure a meaningful share of that growth. Policies that promote wider corporate AI adoption and increase Europe’s role in AI development — together with more integrated digital and financial markets and faster energy infrastructure expansion — would improve Europe’s chances.
Conclusion
Given current uncertainties, expert projections span a wide range: Acemoğlu expects a very modest effect, Aghion foresees a much larger boost, and OECD scenarios provide intermediate outcomes depending on diffusion. The key determinants are how quickly AI spreads, the costs and complexity of firm-level adoption, and regional structural differences. AI could deliver significant long-term growth, but how much of that potential is realized will depend largely on policy choices and business responses in the coming years.
Photo: Heather Diehl / Getty Images — an Artisan advertisement for an AI sales agent named Ava on a San Francisco bus, June 30, 2026.
