Comment by ACCount37

Model inference compute over model lifetime is ~10x of model training compute now for major providers. Expected to climb as demand for AI inference rises.