Comment by j4hdufd8

GPUs are also used to speed up inference (the math is virtually the same). You think your ChatGPT queries are running on x86 servers?