Artificial intelligence is a rapidly developing technology that has dominated public consciousness among consumers, firms and policymakers in recent years. AI is now being discussed as the next general-purpose technology with the transformative effects permeating throughout society being likened to developments like the steam engine and electricity (Calvino, Haerle and Liu, 2025[1]). 1
Training and deploying generative AI services requires an enormous amount of computing resources (referred to as AI “compute” (OECD, 2023[2])), as well as a strong foundation of digital connectivity infrastructure and reliable energy sources. There has been little discussion to date at the intersection of competition policy and AI compute infrastructure. This paper furthers the discussion by focusing on competition in the physical infrastructure that underpin AI services, markets understood to be both highly complex and often highly concentrated (Gambacorta and Shreeti, 2025[3]). This includes AI accelerator chips and other compute hardware (i.e. the technologies needed to train, finetune and deploy AI models), as well as a more limited discussion on the power and cooling, networking and storage infrastructure. Other important non-physical inputs to AI including data provision, model development, finance and skills are outside the scope of this paper.
Given the economic significance and strong public interest in the AI sector, it is essential that competition policy and enforcement mechanisms are equipped to monitor and respond effectively to emerging risks in the AI supply chain. Ensuring that competition in these markets function well is critical not only for protecting consumer welfare, but also for sustaining innovation and long-term market dynamism. At the same time, governments are increasingly intervening in the AI supply chain to encourage investment and promote broader public interest objectives. This creates a need for competition authorities to engage with government and advocate to ensure such interventions consider competition policy. The focus of this background paper is on exploring the competition issues in AI infrastructure and the supply chains which underpin it. This paper is structured in three parts:
Section 2 examines the supply chain, describing the relevant technologies and key players.
Section 3 draws out some common features and highlights some of the potential implications for competition.
Section 4 highlights current competition policy responses and tools as well as some of the challenges authorities face in responding to this rapidly evolving sector.
This paper is part of the OECD Horizontal Project on ‘Thriving with AI: Empowering Economies, Societies and Citizens.’ It builds on previous OECD policy papers in relation to AI and competition, namely those on AI, data and competition (OECD, 2024[4]) and Competition in the Provision of Cloud Computing services (OECD, 2025[5]). These OECD background papers on competition also complement a broad range of work ongoing in the OECD seeking to better understand the implications of AI. These include: ‘The effects of generative AI on productivity, innovation and entrepreneurship’ (Calvino, Reijerink and Samek, 2025[6]), ‘The macroeconomic productivity gains from Artificial Intelligence in G7 economies’ (Filippucci et al., 2025[7]) as well as papers on the value chain including: ‘Mapping the semiconductor value chain’ (OECD, 2025[8]) and ‘Vulnerabilities in the semiconductor supply chain’ (Haramboure et al., 2023[9]).