The paper "Computing Power and the Governance of Artificial Intelligence" explores the role of computing power (compute) in the development and deployment of artificial intelligence (AI). Compute is crucial for training and deploying advanced AI systems, and its concentration and detectability make it an attractive target for governance. The authors argue that compute can be leveraged to enhance three key areas of AI governance: visibility, allocation, and enforcement. They discuss the potential benefits of compute governance, such as increasing regulatory visibility into AI capabilities, steering AI progress towards beneficial outcomes, and enforcing restrictions on irresponsible or malicious AI development. However, they also highlight the risks associated with compute governance, including privacy concerns, economic impacts, and power centralization. The paper provides a range of policy options and considerations for different governing entities and suggests guardrails to minimize these risks, such as focusing on industrial-scale compute and incorporating privacy-preserving practices. The authors conclude by emphasizing the need for a holistic theory and appraisal of compute governance to address its promises and perils.The paper "Computing Power and the Governance of Artificial Intelligence" explores the role of computing power (compute) in the development and deployment of artificial intelligence (AI). Compute is crucial for training and deploying advanced AI systems, and its concentration and detectability make it an attractive target for governance. The authors argue that compute can be leveraged to enhance three key areas of AI governance: visibility, allocation, and enforcement. They discuss the potential benefits of compute governance, such as increasing regulatory visibility into AI capabilities, steering AI progress towards beneficial outcomes, and enforcing restrictions on irresponsible or malicious AI development. However, they also highlight the risks associated with compute governance, including privacy concerns, economic impacts, and power centralization. The paper provides a range of policy options and considerations for different governing entities and suggests guardrails to minimize these risks, such as focusing on industrial-scale compute and incorporating privacy-preserving practices. The authors conclude by emphasizing the need for a holistic theory and appraisal of compute governance to address its promises and perils.