Comment by mattnewton

inference hardware, especially starting with on device ai for the mac. I think they should go as far as making a server chip, but that's less obvious today.