GPU

“Every supply shortage does ultimately result in a glut. But people have been calling for this glut now for you know, 12 months anyway. And they're calling for it again this year. We're not going to see it again this year.”

“You know, we're kind of at, You know, six million GPUs for next year. The bearish people are at like four and a half million. Like the numbers will ultimately tell if they do six million next year, the stock's going higher. If they do four and a half million, …”

“If you just did that math and you compound it with, you add, you compound that with four X per year, On model size and computing size. And then on the other hand, demand continues to grow in usage. Do we think that we need millions of GPUs? No doubt. Yeah, tha…”

“A 100,000 GPUs, that's, you know, easily the fastest supercomputer on the planet. That's one cluster.”

“Somebody who's really, really good at building GPUs have no clue how to be an accelerated computing company.”

“So I think headcount will, in fact, one of the ways I look at it and say is our total people costs will go down, our cost per head will go up, And my GPU per researcher will go up.”

“What I think it was more a assessment on the real bottleneck, which is data centers and power as opposed to GPUs because GPUs have come online.”

“To stand up two billion or two million GPUs in 2026, you need about 2.7 gigs, gigawatts of power, right?”

“When you talk about two years for GPUs, they're going to, cutting edge GPUs are going to be used for cutting edge training for the, you know, frontier models in that first two year period. But all these things are going to continue to get used for inference.”

“We are in a GPU crunch, so we'll see how, you know, how long that goes.”

“When your customers deploy a gigawatt, that's a, you know, 400,500 thousand GPUs. Getting 500,000 GPUs to work together is a miracle.”

“Search was pretty magical in terms of its ad unit and its cost economics, because there was the index, which was a fixed cost that you could then amortize in a much more efficient way. Whereas this one, you know, each chat to your point, you have to burn a lot…”

“It's useful to start with the most zero sum trade off when you do your planning. So I think starting working backwards from GPUs is usually pretty, pretty good idea.”

“When you just look at token consumption per user especially in the enterprise too, which is, you know a massive opportunity. You see a lot of very GPU hungry, hungry workflows, and yes, demand keeps going up even as prices go down.”