Deep learning generally doesn't run on AMD GPUs, but it runs on a lot more than only Nvidia GPUs. If we're only counting model training, it runs on CPUs, Googl

Deep learning generally doesn't run on AMD GPUs, but it runs on a lot more than only Nvidia GPUs. If we're only counting model training, it runs on CPUs, Google's TPUs, FPGAs, whatever other secret datacenter ASICs are out there, various DL-specific mobile chips, etc. Way more than 5% of the world's hardware can run inference with deep neural nets, which is the important thing for mass adoption, and definitely more than only Nvidia GPUs can run training.