Nvidia is probably "wasting" just as much money on R&D to figure out "which ML chips will win out" 5 years from now, so that they can built it first and sell it

Nvidia is probably "wasting" just as much money on R&D to figure out "which ML chips will win out" 5 years from now, so that they can built it first and sell it as a "GPU".