Hacker News Reader

OpenAI uses Nvidia GPUs and Google uses their TPUs.

Everyone except Google uses Nvidia for training. Cerebras, Gaudi, and other custom AI accelerators have unable to surpass Nvidia in performance/$ and performance/watt yet.

You might see more training on Intel XPUs when they come out, since they have such enormous RAM pools. Maybe AMD MI300s and Intel Ponte Vecchio (both 128GB) in the shorter term, though I think they will mostly be in HPC supercomputers instead of cloud instances.