OECD analyses competition risks in AI infrastructure supply chains

The OECD has published a background paper on competition in artificial intelligence infrastructure, focusing on compute hardware, data centers, cloud, and related supply chains as input to AI.

• Main analysis and findings: The paper maps the AI compute stack (chips, fabrication, OSAT, EDA, data centers, cloud, connectivity, power and cooling) and documents high concentration (e.g. Nvidia in GPUs, TSMC in advanced fabrication, ASML in EUV lithography, hyperscalers in cloud), large R&D outlays, vertical integration, cross‑shareholdings/partnerships, switching barriers, and supply bottlenecks; it then assesses how existing tools—antitrust enforcement, merger control, market studies, advocacy, and potential ex‑ante regulation—can address foreclosure, collusion, killer acquisitions, and discriminatory access to compute.
• Context and policy background: Prepared for the December 2025 OECD Competition Committee session on “Competition and Artificial Intelligence”, the paper situates AI infrastructure within broader industrial policy and state intervention (e.g. US CHIPS Act, EU InvestAI, semiconductor subsidies in South Korea, Japan, Germany, France, UK sovereign compute plans), highlights energy and water intensity of data centers and related grid/land constraints, and summarises recent national competition authority work on cloud and generative AI (France, UK, Japan, Korea, Portugal, Canada) and international co‑ordination (EU–US–UK joint AI competition statements, G7 digital competition communiqué).