Maximizing AI output per megawatt

  • – EDT
  • The Mission Critical Power Channel
Speakers

This episode will be available to stream at 9:00am ET

Tokenomics reframes AI infrastructure performance around productive AI output, not just dollars per watt or facility efficiency. One emerging metric is tokens per sec per MW, which connects useful inference throughput to provisioned utility power and shifts the conversation toward maximizing AI output on the lowest-TCO compute infrastructure. This discussion explores how two levers, improving peak PUE to increase sellable megawatts, and improving IT efficiency to increase compute per megawatt, can influence AI factory performance. We will cover:

  • The practical framework for understanding how power, cooling, processor architectures, and IT optimization work together
  • How to maximize the AI value generated from every contracted megawatt
  • How to optimize tokens per second per MW
  • Real-world strategies for reducing TCO while increasing inference throughput

Related Episodes