The Organisation for Economic Co-operation and Development (OECD) has published a policy brief assessing how governments are securing artificial intelligence (AI) compute, which covers the hardware and software infrastructure needed to develop, scale and access AI systems. It identifies compute as a strategic economic and security asset, while finding that policy approaches differ according to countries’ economic, geographic and technological circumstances and are constrained by limited comparable data on effective policy measures. Among the 46 Global Partnership on AI member countries and the European Union, 44 jurisdictions have adopted national or regional AI strategies, including four that are being updated. Twenty-one treat compute as a dedicated strategic pillar, 10 make it an explicit commitment and 13 refer to it more broadly. Canada and the United Kingdom are the only countries with dedicated national AI compute strategies. Governments are pursuing capacity and access through public infrastructure, procurement, grants and tax incentives, regulatory and permitting measures, workforce programs and cross-border partnerships. The OECD calls on governments to commission domestic AI compute needs assessments and align investment with clear national policy objectives. Strategies should account for capacity, effectiveness and resilience, including ownership, location, supply chain security and efficiency. Cross-border collaboration may provide a sufficient alternative to domestic infrastructure in some countries, while public funding should seek to attract private investment where appropriate.