GPUs, TPUs, & The Economics of AI Explained | Gavin Baker Interview
Source
youtube.com
Author
Invest Like the Best
Date
Context
Buying a large collection of AI chips is only part of building a competitive research operation. Investor Gavin Baker argues that keeping the machines working reliably together determines how much useful computing a lab gets from them. Baker pairs that operational challenge with another: choosing which experiments deserve scarce time on a large cluster.
An experiment that occupies substantial hardware for days prevents other experiments from using it. Baker's argument is that research judgment and infrastructure operation jointly determine what a lab achieves with its spending. Hardware purchases alone therefore give an incomplete picture of a lab's capacity to make progress. Listen to the discussion.