And this is the only thing their leadership did right in my eyes: they hired the talent² to provide raw GPU power, which they did. This accidentally turned out to be the thing behind crypto, and, after crypto faded, also the thing behind AI. They didn't anticipate any of that, and their competitors didn't either.
If their luck runs out, which isn't a given¹, we'll see what their leadership can in fact do. I don't think it will be much.
¹ https://en.wikipedia.org/wiki/Gambler%27s_fallacy
² their engineers are of the flap norris kind, though
So just getting up on the surfboard is laudable.
Riding the wave, maybe not.
Huh? They released cuDNN in 2014, a full decade ago, and started shipping tensor accelerator hardware with the Volta architecture in 2017. That was all well before the current hype wave, they saw it coming a mile away.