Ep. 039 - [EMERGENCY] Our Numbers Are "Insane." Here's the Proof. (Datacenters) | Jeremie Ontiveros, Reyk Knuhtsen, Jordan Nanos

Jordan Nanos (0:01)

It looks like if you lie down, you're going to kick your feet through the wall. Like, dude.

Reyk Knuhtsen (0:05)

Yeah, like if I have a night tear, my foot goes right through the fucking wall. Yeah, take a look at the pole in the middle of the room.

Jordan Nanos (0:23)

Right there. I was really dreaming of a Brad Gerstner and I just like fight in the wall and kick in the walls.

Jeremie Ontiveros (0:28)

Yeah, put a specter in there.

My room is fine. It's just that I got two stucco beds, which was kind of weird, but besides that, it was okay.

Reyk Knuhtsen (0:36)

That room is, look at that carpet. You know that room is nice, yeah. They just said, yeah, I can't break the bucket shitty one.

Jeremie Ontiveros (0:44)

You're just kidding. Oh, right there.

I know. Talk to Michelle.

Reyk Knuhtsen (0:51)

Yeah, I got to talk to someone.

Jordan Nanos (0:53)

Okay. You guys ready to rip into it? You guys ready to start?

Reyk Knuhtsen (0:56)

Yeah.

Jordan Nanos (0:56)

You know. Everybody get in bed for the cold open.

Reyk Knuhtsen (1:00)

Let me jump in.

Jordan Nanos (1:08)

Hello, everyone. Welcome back to another episode of SemiAnalysis Weekly. In this episode, I'm joined by Jeremie and Reyk, our experts from the data center model. We're going to talk about their article, Data Center's 2027, the Year of Acceleration, 50 gigawatts in a single year.

This one comes from a bit of a backdrop. We just the other week, Brad Gerstner from Altimeter was showing our data on the All-In podcast and then his BG2 interview with Jensen, and they were talking about how our forecast of 43 gigawatts in 27 is too aggressive. Well, we're going to dig into why that's not the case. We're actually increasing the forecast. When people think that only about 25 gigawatts are going to get built, we're going to go through exactly what the FUD is about. They talk about moratoriums and other political restrictions. They talk about power generation. They talk about all sorts of reasons why this might not be the case. But guys, I don't want to walk through it. Let's talk through it with you guys. Welcome. Thank you.

Jeremie Ontiveros (2:07)

Thanks for having us.

Reyk Knuhtsen (2:08)

Yeah. We got a lot to talk about today. This is a good one.

Jordan Nanos (2:12)

Okay. Before we get into the actual content of the article, how about we do lay the land and you guys give me a little bit of history on your track record with the data center model. Jeremie, can you tell me a little bit about the history of this model when you launched it and just talk about some of the forecasts that you've made in the past and how that's played out in terms of fidelity of the model?

Jeremie Ontiveros (2:32)

Sure. The model launched nearly three years ago now.

We published an article actually in March of 2024 to discuss publicly some of these results that we got from building the first iteration of that. So that article is still available today. It's AI Energy Dilemma, Rates for Data Service Base, whatever. So we started this article by showing other market forecasts out there, and then putting our own, which was materially ahead of many other firms, like McKeezy or whoever else was making forecasts these days.

And so that has largely, of course, been proven right. We've increased the forecast a bit since then, but overall, the orders of magnitude out there, we were expecting back in early 2024, again, a material acceleration in 2025, and then in 2026 The way we do this is always by combining several data points. So we use data from the CHEP side, we use data from the data center side, from the energy side, we always triangulate as much as we can through everything we can follow. Another thing I could point you to is Oracle. A few months ago, Oracle stopped saying half of US data centers are delayed, where we show this chart of our 2027 number and how it has been revised over the different versions of our data center model. It has barely moved. We're always expecting.

Reyk Knuhtsen (4:04)

Sorry, but it was the 2026 one, though.

Jeremie Ontiveros (4:06)

Yeah, whatever. 2026, 2027 People can go look at it. Same thing.

Jordan Nanos (4:11)

It's all right, Jeremie. It doesn't matter which year it is.

Reyk Knuhtsen (4:14)

Yeah, you can pick any year. It doesn't matter.

Jeremie Ontiveros (4:17)

All right. It's 2026 Sure.

Yeah, exactly. It's been pretty extremely close. Again, we've been doing this for nearly three years now. So the track record has been there now. The methodology is actually not that complicated. It just got to be thorough and comprehensive and make sure you got everything. So we obviously spent a lot of time and built a whole team to ensure that. But that's what got us there is now, we have extremely strong submittals pointing to 27, being a gigantic year, many different subcomponents that we can talk about in depth. And so there we are after Dear Brad called us out publicly and cut our forecast in half.