This Working Paper develops a methodology to estimate and track the global physical distribution of public cloud compute availability for artificial intelligence (AI). The methodology counts cloud regions operated by major providers that hold a significant share of the global public cloud market. Cloud regions – physical hubs hosting specialised hardware designed to efficiently run AI workloads – are identified through publicly available data, and their AI compute capabilities are aggregated by geographic location. The resulting indicators categorise economies based on their availability of public cloud AI compute. This work supports efforts to monitor the global AI compute landscape and contributes data to the OECD.AI Policy Observatory and the forthcoming OECD.AI Observatory Index.
