Data Center Buildout Going Too Far: Bokeh Capital's Forrest

Bloomberg

Watch Data Center Buildout Going Too Far: Bokeh Capital's Forrest - Bloomberg on Dailymotion

Transcript

00:00You're joining us now on the theme of the day, I think.

00:02It has been all about data centers, whether it's 50 billion coming from Anthropic,

00:06whether we're thinking about Meta continuing its focus in Wisconsin.

00:10But why do you think it's too much too soon from the Wall Street's perspective?

00:15Well, for a lot of reasons.

00:17Living in a virtual world, as most of Silicon Valley does,

00:22it's really hard, whatever they have to, you know, go out into the real world and do physical things.

00:28And that's what we're talking about with data centers.

00:30We need power.

00:32We need really kind of exotic.

00:37They're kind of boring buildings, but they need to be very uniform,

00:41and there's only certain kind of people that can build them.

00:44And then we have the demand.

00:46So I've seen some reports, maybe on Bloomberg,

00:49about how some data centers that are already established are having problems, you know, selling their capacity.

00:56So we have a lot of physical world problems.

01:00But I think my biggest thing is, is I'm not a believer that we really need all these data centers.

01:07I think human beings are incredibly smart and are going to work around what we're doing now,

01:12which is the brute force method of training large language models.

01:17I believe in AI, just not AI as it's presented right now.

01:21So, Kim, there's the rub, right?

01:23Because you're not just saying that we've got utilities issues, power issues, supply chain issues.

01:27You're saying we're going to innovate ourselves out of that.

01:29Are we going to the United States or is it going to be Asia?

01:31Well, right now it looks like Asia was first on track with the whole DeepSeq product that they came out with, right?

01:39We can quibble about if they unfairly used OpenAI to leverage, but they did leverage.

01:47And that's the most important part.

01:49They thought out of the box as opposed to saying, you know,

01:52we need to have the state of Texas be one big data center.

01:55I'm using hyperbole here.

01:57But, you know, they thought about how can we do this differently?

02:01And I think ultimately it's a good race between the U.S. and the West in general and China

02:07about who is going to be able to come up with usable AI that doesn't, you know,

02:13break the bank and cost every last dollar that Microsoft, Meta and the rest of them have.

02:21Kim, you've presented your thesis.

02:23So as an investor, custodian, fiduciary of your clients,

02:28how do you invest to support the thesis you've just outlined?

02:33Sure.

02:34Well, I think we have to have many options, right?

02:38We do own some NVIDIA, but we also own AMD, Micron and believe it or not, Intel.

02:47AI is real and AI is going to have very different needs.

02:50Kim, sorry, may I just, may I just interrupt you?

02:53Like the AMD story today, the whole thing is predicated on the idea that they do participate

02:59in some massive data center build out.

03:01And you just said we don't need the data centers.

03:04Well, there's going to be some data centers, but not as many as are planned currently.

03:09And I think that's the big thing that even, you know, last week's big cha-cha-cha in the world of AI

03:17was the Brad Gessner and Sam Altman interview where, you know, they're saying,

03:24where are you getting the money, Sam?

03:25So, I mean, there's a lot of concern about the whole theory of we need all of these data centers.

03:31But I think over time we need them.

03:34But isn't AMD currently trading in a valuation that is pricing in an enormous data center boob

03:39that will come to light?

03:42Or are you thinking actually AMD at this price is reflective of perhaps not such a significant build out?

03:49Well, I am.

03:50And the other thing is, is whenever we discover that we don't need the brute force method,

03:55these chips will be instrumental in that, right?

03:59So I think we're going to deploy money.

04:02It's just differently than the way we're currently thinking.

04:05How about that?

Category

Transcript

Be the first to comment

Recommended