Productivity Research Notes (PRN Series), No. 2610. Artificial intelligence is increasingly discussed as a driver of future productivity growth, yet its expansion also depends on electricity supply and physical infrastructure. This note argues that AI computation can increasingly be traded internationally as digital services, while electricity supply remains geographically constrained. Using Japan as a case study, it examines how electricity costs, transmission limitations, and industrial linkages shape the domestic location of AI-related computation. The note further highlights the uncertainty surrounding future AI-related electricity demand, emphasizing both the evolving electricity productivity of computation and the uncertain long-run sustainability of rapidly expanding digital consumption patterns. These observations suggest that AI competitiveness may increasingly depend on stable, scalable, and economically sustainable electricity systems as much as on advances in computation itself.
Koji Nomura (Fri,) studied this question.