Apple simply does not have the talent pool to take on either nvidia or the big LLM providers anywhere on the stack except for edge inference.
If you're saying Apple is going to 'take on nvidia' in edge inference, then I don't disagree but I would hardly even count that as taking on nvidia.