as nvidia goes into making purely big-business hardware, some other player will arrive to provide the chips to the masses. then, the researchers will use this mass-market hardware for research, supplanting nvidia itself.

nvidia got their strong position in AI now by providing CUDA to researchers who couldn't afford their own supercomputing cluster. and now the entirety of machine learning revolves around CUDA.

once you fixate on the business and lock out the masses, you instantly elect to become a fossil like IBM.

They stopped manufacturing the 5090 even under Jensens control. And no the 5090 die is only 750mm squared big, a B200s die is 1628mm squared. That is more than twice the size of a 5090.