Forrester：AI支出增长，资本将转向私有模型与替代芯片

What do you think is driving this massive sell off we're seeing? So let's keep this in perspective. So Forrester believes that tech expenditure will be up to $6,100,000,000,000 next year, and it will be up around 8%. So tech expenditure is not slowing down. It's increasing. And most... And tech, AI will be, 30% of all tech expenditure by the year 2030. So let's calm down, everybody. AI is... This is the most important technology invented in my lifetime.

It is... This is extraordinary. So let's let's calm down. Is there, like, at least at the margins, any expected slowing post election? I mean, you look at Texas, for example, where we're expecting so many data centers to come online. You have a governor that largely ran... Or not largely ran, but partially ran on data centers. All of the local electorate like the house and the senate are also seemingly flipping blue.

Wouldn't that at least at the margin pretend to a little bit of a slowdown? I think what's gonna happen is you're gonna see these data centers get built, or they're gonna get built in different states. You're gonna see a migration. This is too powerful of a force. It's too powerful of a technology. It's too useful of a technology to not happen. So it's going to happen, but just in different states. Definitely not in Texas now, probably not in Ohio, and not in Georgia.

So there... There'll be... Let's let's let's call it a migration of capital at this point. And remember, there's a factor which we haven't, we haven't even looked at yet, which is the building of what we call private models. I'll make a distinction for you. The Geminis and the Anthropics of the world are the public models, the big frontier models. But Forrester believes that 70% of the, the revenue made in AI is gonna be in private models.

Private models are the models built by banks, insurance companies, car companies. We're running event, an event over at Hudson Yards with all of our... Those large companies, and they're all now in the midst of building those private models. So this... It's a bit of a misallocation capital that's going on right now. Miss... Miss... Misallocation capital to the public models, but, actually, we're gonna see a lot of capital now go to the billion private models, and that's where most of the action's gonna be in the future.

I wanna stick with the AI data centers. Can you talk to us about the bull case and the bear case of a slowdown in AI build out? Because yesterday, the hyperscalers were rallying because of that. I mean, today, I guess, it's a different story, probably unrelated, probably related. Talk to us about the two scenarios. Yeah. So, I would say... And I'll give you a nuance here. What's gonna happen is the private models get built out in companies, JPMorgan or Bank of America.

You're gonna see a repopulation actually of the data centers. We're gonna bring computers back into the data centers to run those models. So that's gonna take some pressure off the building of the, of the hyperscaler and and and and the outside data centers. So that... It's... This could be a moderation here. I I wonder if that also squares with the narrative that was largely taking hold yesterday that, like, actually, this is a good environment for the hyperscalers because, one, they already have data centers built, so they don't have to worry about more coming online.

Right. And they've already interacted with these communities in terms of the local pushback. They're already there. They know they know how to deal with them. Do you think that's kind of the right way to look at their business at the moment? Yeah. Well... But remember, I I would say yes, but the... Do remember, we've built out these cloud computing data centers over the last ten years, and you're gonna see a lot of this, a bit of a nuance.

These... The systems within large companies are going to be are gonna be flipped over into what we call AI computing. And when that happens, you're going to see a lowering of the demand from the old systems which are running in those data centers to AI. So I'm just talking about some... There there is gonna be some moderation here in the in the demand for these data centers. So today, again I'd say... Sorry to interrupt, but but I think what's interesting is what chips will be running those data centers.

So powerful companies and powerful CEOs don't like to be held hostage, and they're being held hostage by NVIDIA now. So what Forrester believes is... And you see this training is being built by AWS, Maya by by Microsoft. You're going to see other chips in those data centers other than than NVIDIA. Doesn't mean a crash in NVIDIA. Just means that the growth in NVIDIA may not be as quite as optimistic as we... As we're imagining.

So then what would that mean? Because for a long time, it feels like the NVIDIA NVIDIA was the stock market. I mean, last year, I think we would have NVIDIA watch parties when it will report earnings, and now they report November 17. I think to that You're gonna see moderation here because of those those those new chips that are that are gonna be produced by those other companies. By the way, for those who are just joining us, just wanna take a moment and just, again, show our viewers what these markets are doing.

It is an S and P that is down, quite significantly 3.1%. The, the Nasdaq is down now more than 5%. These individual names getting hit hard. George Colony of Forrester Research is here with us. George, I wonder at what point... Look. You're not a stocks guy. I get that. But I wonder at what point if this sell off becomes something more than a stock story and becomes, like, a capital allocation fundamental story. If you continue to see the cost of capital get more expensive Yep.

Because spreads widen out, the banks, Apollo, all of the financial sectors getting hit hard too. If, like, some part of the lending universe starts to pull back a little bit because it's getting more expensive, like, could this turn into something more fundamental than just a stock market sell off? Well, I think, again, we we were over capitalizing that part of the world. I think I think there was too much capital flowing there.

Again, the private models will need more capital as well. So it's f b rebalancing there. But I think what this is gonna do is it's gonna force the models to not get bigger. We're gonna have to go to some dis... To to distilled distilled models, models which are in the, the billion parameter, mode and not the trillion parameter mode. So I think what's gonna... This is gonna mean is a different type of models are gonna be built, smaller models.

And some of the Chinese models are in fact following this following this trend. So are we entering a new phase of the AI trade, a phase where investors are becoming more selective, a phase that maybe NVIDIA won't be as consequential or significant anymore to the ecosystem? I think so. Abs... I think absolutely. I think I think we're... Again, we're... Too much capital in that direction, a a closer examination of the capital, and, again, some of the capital going over to the private model.

So I'm with you on this one. I I... I'm just... A a story that just properly caught my eye, and I realized you won't have seen this yet. So Yeah. Okay. You know? Alright. Let's get Full disclosure for all of our audience. George was included on... In on this before. There's a story that our team just wrote, Brody Ford, about Microsoft that it's souring on the neo clouds. Part of the issue they say is the delays, that it's too risky due to delays, regulatory snags Yep.

And high costs. And I wonder if we should take another look at these neo clouds. If Microsoft, what might be one of many, saying things are maybe getting too risky relying on some of these some of these companies. I think that... Again, I think you're seeing the breaking... There... There's some shattering in in the in the structure here. I think that's another indicator...