Key publications:
Trends in AI Supercomputers — Assesses trends in performance, cost, power, owners, and locations of our dataset of over 500 AI supercomputers. (Summary | Paper)
Increasing Compute Efficiency and the Diffusion of AI Capabilities — Analyzes how better hardware and algorithms broaden access to AI but also provide an advantage to the the largest compute investors. (Paper)
Computing Power and the Governance of Artificial Intelligence — Outlines the role of compute for AI development and deployment, characterizes key properties of compute and discusses policy options these features may enable. (Paper)
AI's Power Requirements Under Exponential Growth — Extrapolates compute trends to assess potential power needs of AI data centers in the United States and discusses geopolitical implications of compute potentially going abroad. (Twitter Summary | Report)
Compute at Scale: A Broad Investigation into the Data Center Industry — Report about key features of data centers, including components, global distribution, and future bottlenecks for their buildout. (Report)