Google's AI Infrastructure Chief, Amin Vahdat, on the Physics & Economics of Frontier AI
Source
youtube.com
Author
Sequoia Capital
Date
Why it matters
Explains why failure-adjusted goodput matters more than peak FLOPS at 100,000-accelerator scale, and how long-horizon agents shift demand toward CPUs and storage. This informs how to think about future inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition → capacity and cost.
Terms in this piece · Glossary
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
videoWhen to Build Your Own Agent Harness | Harrison Chase, LangChain
videoWhy Hardware-Software Co-Design Is AI's Real 100x: Dylan Patel of SemiAnalysis
videoContext Engineering Our Way to Long-Horizon Agents: LangChain’s Harrison Chase
videoDeal Velocity, Not Billable Hours: How Crosby Uses AI to Redefine Legal Contracting
