Yes, but this just makes a frictionless runtime for inference even more important (which is something that does not exist in a comparable form for AMD).