Drew Baglino

speaker
1,210 appearances 2 recordings 1 series first heard Aug 2026 last heard 6 Aug

Drew Baglino’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months
2 · Aug OctJan 26AprJulnow

Recordings per month over the last 12 months — 2 in all, peaking in Aug 2026 with 2.

Appearances

newest first · ▶ plays the moment
And when you look at the compute, it's happening too.
I mean, Jensen goes up on stage and talks about how many tokens per watt
you know, how that curve of tokens per watt is going up and to the right with, like, every generation.
That's making it easier to... That's making it possible for intelligence to be more and more affordable, right?
Because as you bring the tokens per watt down...
I mean, as you produce more tokens per watt, every query becomes more affordable, more accessible to more people.
And you can do more in a particular data center footprint, right?
Like you can return on capital.
And that's because they've been dramatically improving like the elegance of the architecture inside of the compute.
And I think we have the same opportunity on the power side.
Sure.
I think it's actually just straight up power density.
So to deliver those tokens per watt improvements, and you're bringing more and more GPUs together on the copper plane, because that's where the latency is the lowest, that's where you can get the most efficient compute, that's a power density problem.
So that's why these racks are going up in power.
It's
It's a performance-driven optimization, like to deliver the best possible compute, this is the way to go.
But as you go up in power, you need to go up in voltage.
Otherwise, you're just like, the copper consumption is nuts.
You can't fit that amount of copper in the rack volume.
It just doesn't work.
Showing 141–160 of 1,210 · page 8 of 61 ← Previous Next →