00:00Vijay, a tepid response from the market. We're only up half a percent this morning. What was your take on
00:04the earnings set?
00:06Yeah, thanks, Danny. So I think NVIDIA overall was a pretty solid, clean beat and guide.
00:11I think if you look at their data center revenues, it is accelerating from up 75 percent last quarter to
00:18up 85 percent in the next quarter year on year.
00:21Networking had the strongest quarter in their history, up almost 260 percent year on year.
00:26So massive growth across verticals. If you look at 2026, data center revenues this year will be $340 billion.
00:34So you're seeing some massive numbers come through. And it's very tough not to own this.
00:40They're growing 70 percent this year, 30 percent plus next year. Those numbers will probably go up.
00:46And the stock is trading at 30 times forward. So it's very hard not to own this thing.
00:51And so we think this will continue to move higher. But everyone does.
00:56Go ahead. Everyone does own it. I mean, we're talking about a stock that's, what, 10 percent of the S
01:04&P, 8 percent of the S&P.
01:06And, you know, we've been buying it for three years now. So maybe the marginal buyer just isn't there for
01:14these shares.
01:15Yeah, I think, Matt, that's a good point. I think over the last two quarters, we think NVIDIA is probably
01:22a little less owned.
01:24It is probably a little under-owned in the last two quarters, given how the stock has tracked.
01:29So I think there's an opportunity to now come back because all the metrics on the data center side and
01:35networking side seem to be accelerating.
01:37They've talked about token generation. That seems to be starting to pick up as well.
01:41They've talked about the cost per token going down 10x, meaning the cost of inferencing cost per token would be
01:48down to a tenth going out.
01:51And that would be key for all the hyperscalers to continue spending.
01:54So we have seen the stock trade side because of some of the hyperscaler CapEx concerns.
02:00But now as the costs start to come down, you're seeing more token generation, more agentic AI, software getting disrupted.
02:07You will start to see more money flowing back to these names.
02:12There is also significant catalysts ahead from GPC coming up.
02:17They are presenting at conferences.
02:19There's a huge Rubin launch in the back half of the year.
02:22So there's plenty of reasons to step back in.
02:25Costs are coming down, but maybe the competitive dynamic is also changing within that because you have custom chips coming
02:33from the hyperscalers.
02:35You have emerging technologies and chip makers also trying to come into.
02:39Vijay, where does the state of competition stand for NVIDIA?
02:44Absolutely. That's a great point.
02:45So I think we do see internally hyperscalers drive some ASIC platforms, custom ASIC platforms.
02:52Because that's because they can control that ASIC platform internally.
02:56They can control the supply.
02:58They can control the cost there.
02:59But NVIDIA still leads the globe in training and in frenzy.
03:03They are the golden standard.
03:05That's what the ASIC guys are chasing in terms of performance.
03:08But it's a very risky trend as well, right?
03:12Because, you know, NVIDIA on every year, they keep bringing inference costs down, training costs down.
03:19So the ASIC hyperscalers have to continuously be on their toes to evolve the ASIC roadmap as well,
03:26so that the gap between the GPU and the ASIC does not become too far wide.
03:31And so that's always a tricky proposition.
03:33But the reason you have these ASIC platforms coming on for the hyperscalers is because they want to control that
03:40roadmap.
03:40They want to have a lower cost option.
03:43But when you look at training and frenzy, no alternative to NVIDIA in terms of where they are in the
03:48lead.
03:49I wonder, Rakesh, if they're constrained by, you know, capacity in terms of production, right?
03:56They can't sell as much as – I mean, we know they are, but maybe that's part of the problem,
04:00right?
04:01And also, what about competition?
04:04AMD seems to be making headway here.
04:07Is it a viable competitor, something that could worry shareholders?
04:11Yeah, no, I think in terms of capacity, they have very good line of sight on their capacity.
04:17They have, obviously, a very strong partner with TSMC, and they are adding capacity in Taiwan, in the U.S.,
04:23in Japan.
04:24So that continues to be, you know, a nice tailwind for them.
04:27I think they have a pretty good handle on that.
04:29In terms of AMD, it's a good competitor.
04:32But, you know, I think as you look at 2026, AMD's data center revenues are somewhere in the 17, 18
04:38billion-ish, 18 billion-plus biners.
04:41If you look at NVIDIA's data center revenues this year, 340 billion.
04:46So that tells you the math of where these things are going.
04:51Plus, also, you know, as you look at some of the new ramps of NVIDIA, the performance in terms of
04:56Blackwell being 50 times the prior platform,
04:58Rubin driving 10x lower influence costs, and CUDA, right?
05:03We cannot forget CUDA because that's a huge software stack that's been a massive moat for them.
05:09Continues to drive a big value proposition for NVIDIA.
05:11Vijay, when are we going to get management assuming some China data center revenue in their outlook?
05:16That's a great question.
05:18I think, you know, this 340 billion obviously does not include any China.
05:21And China, as Jensen has himself noted, is almost 50 billion a year in incremental data center revenues.
05:28And that's not even in the guide or, at this point, in the numbers.
05:33So I think we should start to see that pick up in the back half because, if anything, China does
05:41need the performance and the trading platforms from NVIDIA.
05:46So I think we should start to see that as well.
05:46So I think we should start to see that as well.
05:46So I think we should start to see that as well.
Comments