00:00 Speaker A
Dan, great to see you. So, let’s let’s start on Meta what you wrote about my friend. Announced new four new AI chips. So a serious push here by Zuck. Break that that down, Dan. I mean, did you see this is uh, you know, is this really about performance, Dan? Is it saying, hey, I I’d like to get maybe a little less dependent on Nvidia? Is it both? What did you make of it?
00:30 Dan
I can’t answer any questions until I get that $250,000 payday. So um I’m I’m going to
00:36 Speaker A
That’s for a seasoned journalist, Dan, seasoned. which I thought meant old. I don’t know.
00:43 Dan
Yes. Um no, so these are these are brand new chips coming out of meta. Uh they’re four new chips, believe it or not. It’s the uh MTI 300 MTI, uh 400, 450 uh and uh 500. And, you know, the big deal here is that believe it or not, meta is saying that it has uh performance uh and cost savings comparable to current commercial level systems. and that means Nvidia and AMD. I mean who else could it be when they say commercial system.
01:21 Dan
So, this is a chip presumably that could go up against those when it comes to raw performance. Now, you know, meta doesn’t necessarily release these out to other companies. They use them for their own capabilities. There it’s for their uh recommender engines as well as standing up their their own uh AI. but it is interesting to see how they’re advancing uh in this area. Now,
01:50 Dan
it also does raise that uh question that that we’ve been talking about when it comes to these companies building their own chips is what does this mean then for Nvidia and AMD? Does it mean that, you know, this is the end of the road for them? No, not by a long shot. Meta just signed a deal, a multi uh year multigenerational deal with Nvidia as well as AMD to use their chips. Every other company continues to want their chips.
02:22 Dan
They’re using theirs uh their own though as a hedge against being locked in to one or the other. So they have these, uh it’s a way to say, okay, here’s here’s our stuff and then we’ll use Nvidia and we’ll use AMD, but we don’t want to be locked in to to one chip manufacturer. So it it is though, like I said, very interesting to see them kind of throw that shot out there saying commercially available systems.
02:56 Dan
Yo, that’s that’s wild.
02:58 Speaker A
So that you make it a point that I wanted to just drill down with you on because I it was a question I had for you was, you know, bottom, I’m a viewer, let’s say right now and I I I’m invested in Nvidia or AMD and I I wanted your take on, okay, do I see a headline like this and think, you know, wait a minute, is that kind of long-term revenue risk or is your point AI demand broadly right now is just too strong for that.
03:26 Dan
Yeah, I think I think it’s just too strong right right now for that. Uh, you know, when it comes to this data center build out. AI, you know, when it comes to enterprise usage or people’s uh, you know, consumer usage, we still aren’t exactly clear on the demand for that, but there there is the clear demand though for the hardware and and and a lot of uh these these large uh hyperscalers are saying that their capacity constraints. they’re going to continue uh to want to build out. they’re going to spend $650 billion dollars
04:00 Dan
this year between Amazon, Microsoft, Meta, Google uh uh putting together these these data centers. And so it’s not as though this is going to disappear. Uh when you go back and read uh CFO Collet’s uh uh commentary from Nvidia’s most recent earnings, uh she pointed out that slightly more than 50% of data center revenue came from those hyperscalers, but the other 50% came from a a growing number of other companies that aren’t within that hyper scalar category.
04:36 Dan
So that that, you know, leads you to believe that there is growth in other areas. They’ve talked a lot about sovereign AI and how, you know, they they would work with uh different uh uh countries to stand up data centers. But, you know, I think the the the main thing to keep in mind here is that, you know, this is something that is going to continue uh, well, where we’ll see these companies roll out their own chips. and as I said, it’s it’s a hedge against being locked into one ecosystem. Um, but I, you know, I I do think in the back of your your mind, you have to think 650 billion, that’s a lot of clams, right?
05:14 Dan
They’re not going to be able to do this forever. Uh and I I think, you know, that’s something obviously to keep in mind that that the explosion of data center buildouts isn’t a, you know, forever thing, but it will continue for the time being. And also don’t forget, you know, sure they’re building data centers. That’s where, you know, these huge amounts of chips go, but they also have to maintain data centers and so we would still see companies purchase chips as those those uh uh uh servers burn out.
