00:00Gil Luria is standing by right now. He's managing director and head of technology research
00:04at D.A. Davidson. I don't know if you've had a chance to catch your breath here, Gil.
00:08A lot of numbers crossing the wire, and I don't know, maybe you can kind of make sense of it
00:11here. Are we looking at a good thing or a bad thing right now when it comes to these four
00:15major hyperscalers? It's a good thing, and the difference in stock reactions really just has
00:21to do with how high the expectations were. It looks like in the three critical data center
00:26businesses, the hyperscaler businesses exceeded expectations. They're all still accelerating.
00:32AWS, Microsoft Azure, Google Cloud accelerating, still growing faster and faster, and that's
00:38a very good sign in terms of the progress of AI. That's a key takeaway. The stocks are reacting
00:43differently because expectations from AWS were for even higher, and for Microsoft, investors
00:48were a little more skeptical, but both look good, as does Google Cloud. From an initial
00:54look, it looks like advertising for Google and Meta is also strong. That has to do with
00:59the broader strength of the economy. Amazon, the retail side, looked good. It has to do
01:04with the broader strength of the economy. So I'd say, yeah, overall, first look is a positive
01:08look from all these companies.
01:10A positive look for all these companies, and you, of course, know it was a couple of days
01:13ago where everybody was kind of wringing their hands over that Wall Street Journal report
01:17about OpenAI and its sales and internal metrics that, at least according to that report, were
01:22falling short. And of course, everybody started looking at this broader AI ecosystem, particularly
01:26the hyperscalers. The idea, let's face it, all these names we're talking about right now,
01:30they have basically been the main financiers, if you will, of this buildout. Do you get the
01:35sense here that these companies will remain committed to some of these CapEx forecasts? And more
01:40importantly, do you think they feel confident enough to articulate some sort of return on
01:46investment, or at least being able to communicate that to investors?
01:49Yeah, we'll find out soon enough. A lot of it's going to be in the commentary on the calls. But
01:53yes,
01:54they are seeing very strong demand. That article about OpenAI was really a backward-looking article,
01:59and it was a little out of context, because one of the things that happened to OpenAI is that
02:03Google Gemini did so well, is that Anthropic is doing so well. Overall, if you look at across
02:10Anthropic and OpenAI are at a $50 billion revenue run rate, they were at zero two years ago.
02:15That is the ROI. By the time you add what Google's getting in revenue and the direct revenue Amazon,
02:21Microsoft, and Google are getting from selling these services, the returns are showing up.
02:26The demand is off the charts. We know that because Anthropic is compute constrained. So we're likely to
02:32hear very positive commentary about demand. The subtlety will be the supply side. Can we build
02:39data centers fast enough to supply that demand, given all the bottlenecks, CPUs, and memory,
02:45and advanced packaging, and optical, and turbines? That's where a lot of the conversation is going
02:51to be. But I would expect the commentary over the next couple of hours to be very positive
02:54about the demand environment in AI.
02:57Yeah. And Katie, I just want to point out, too, I mean, they actually did talk about those Gemini
03:01enterprise numbers, 40 percent quarter on quarter growth. So basically talking about the sequential
03:07growth here. So pretty substantial growth. Yeah, absolutely. Well, Gil, I do want to talk about
03:11CapEx specifically in the context of Meta, because you take a look at shares right now down more than
03:175 percent after hours. They did boost their full year CapEx guide. And I mean, this is something that
03:24Meta has gotten dinged for in the past. We know that they're trying to offset those costs when it comes
03:28to
03:29maybe shrinking the workforce. But when you think about, you know, the year ahead, all of this CapEx
03:34that's coming through, does it make as much sense for a Meta to be spending these types of numbers?
03:41It does not, because Microsoft, Amazon, and Google are selling most of that capacity. Meta, it's all for
03:49internal use. These are very, very high levels of CapEx for internal use. Unless they start selling
03:57external capacity at some point, this is coming out of their shareholders' pockets. And that's why
04:04shareholders are reacting strongly. Whenever the CapEx and the OpEx for Meta grow faster than the
04:11revenue growth, which is substantial, investors react badly to that, because it's all for internal
04:16use, where, again, Microsoft, Amazon, and Google are going to talk about how the demand from the
04:22economy, from consumers, from enterprises is rising. Meta is only using it to train its own
04:28models and deliver its own models.
04:30So, Gil, where does the story go for Meta, then? Are we just setting up for another year of
04:35efficiency? We know that we ended up in that situation a couple of years ago. And to some,
04:40I mean, we're looking at a very similar sort of backdrop right now.
04:44Yeah, I think that part of the answer is going to be that the OpEx is going to have to
04:49decline
04:50going forward if they're going to increase CapEx that much. And then at some point, either
04:54Meta turns around and starts selling that capacity or slows down this CapEx and at the very least
05:01delivers a model that's competitive. Right now, the frontier models that are coming out of Meta
05:07aren't even in the top five. We have OpenA, Anthropic, Google. Then we have some Chinese.
05:12We have Grok. Meta's not even in the top 10, I would say, in terms of its frontier model. Unless
05:18they show they can be competitive there, investors are going to expect them to either calm down on the
05:23CapEx or reduce enough OpEx to allow margins not to be degraded. The good news from Meta is the
05:29top line is still growing very fast. Again, the advertising business is very strong. And they
05:33make the argument that because we're so good at AI, that's why we're selling so many more ads
05:38at higher prices. That only goes so far.
Comments