Patel says CPU-to-GPU ratio has shifted from 1:100 megawatts to much closer parity.
“So whereas before you'd have like many GPU servers per CPU server, and so you'd be like, 100 megawatts of GPUs would be served by even like one megawatt or less of CPUs.”
Patel says CPU-to-GPU ratios have tightened from 1:100 megawatts to much closer ratios due to RL and agentic inference.
“Nowadays the ratio is like getting much, much closer, both for RL training and for inference, agentic inference.”