Bloomberg Technology Special: Nvidia CEO Jensen Huang

Bloomberg Audio Studios, podcasts, radio news. 0:00:08.640 --> 0:00:12.959 From Mahart where innovation, money and power Collie in Silicon 0:00:13.039 --> 0:00:13.920 Vallet NBN. 0:00:14.280 --> 0:00:17.799 This is Bloomberg Technology with Caroline Hyde and Ed. 0:00:17.840 --> 0:00:35.720 Ludlow live from San Francisco to our TV and radio 0:00:35.800 --> 0:00:39.239 audiences around the world. Welcome to a special edition of 0:00:39.240 --> 0:00:42.120 Bloomberg Technology. I'm Ed Ludlow. In just a few moments 0:00:42.440 --> 0:00:45.479 in video, CEO Jensen Wang will join us for a 0:00:45.560 --> 0:00:49.479 live interview following their latest earnings report, the company posting 0:00:49.479 --> 0:00:52.960 a revenue forecast that beat consensus that fell short to 0:00:53.040 --> 0:00:57.200 some of the most optimistic estimates, stoking concern that the 0:00:57.320 --> 0:01:01.120 explosive growth is waning. Let's get right to Bloomberg Semiconductor 0:01:01.160 --> 0:01:04.080 correspondent Ian King, who joins me on set. Let's start 0:01:04.080 --> 0:01:07.000 with the basics, the fiscal third quarter forecasts and what 0:01:07.040 --> 0:01:07.760 we learned through it. 0:01:07.920 --> 0:01:11.680 Yeah, that forecast was fine. If you compare it to consensus, 0:01:11.680 --> 0:01:14.240 it was there or thereabouts, and for most of the 0:01:14.319 --> 0:01:16.720 companies in the world, that would be great and everybody 0:01:16.760 --> 0:01:17.360 would be happy. 0:01:17.400 --> 0:01:18.280 But this is in video. 0:01:18.640 --> 0:01:22.840 This is a company which beats pretty much every quarter 0:01:23.160 --> 0:01:26.160 and by an order of magnitude, and it didn't do that, 0:01:26.240 --> 0:01:28.200 and it didn't indicate it would do that, and that 0:01:28.280 --> 0:01:30.119 raised a lot of questions. As we heard on the. 0:01:30.040 --> 0:01:32.680 Call, there are many storylines. I think the demand from 0:01:32.680 --> 0:01:36.039 the hyperscalas is clearly intact, but Blackwell was everything. I 0:01:36.080 --> 0:01:39.039 want to play a SoundBite of what Jensen Wang said 0:01:39.200 --> 0:01:40.480 about Blackwell. Listened to this. 0:01:41.520 --> 0:01:45.640 The change to the mask is complete. There were no 0:01:45.959 --> 0:01:55.160 functional changes necessary, and so we're sampling functional samples of Blackwell, 0:01:55.360 --> 0:01:59.360 Grace Blackwell in a variety of system configurations as we speak. 0:02:00.960 --> 0:02:03.400 The main point here is that there was not a 0:02:03.480 --> 0:02:07.640 design issue with Blackwell itself as had been reported, but 0:02:07.680 --> 0:02:10.799 based on what Nvidia said, this was about the production mechanism. 0:02:11.000 --> 0:02:13.840 Blue Beazine King, you did a very good job in 0:02:13.880 --> 0:02:17.320 the top Life blog of explaining a GPU mask. Could 0:02:17.320 --> 0:02:19.720 you just try to give a short version of that 0:02:19.760 --> 0:02:20.400 to our audience. 0:02:20.440 --> 0:02:24.079 Yeah, the mask is basically the blueprint, which basically that 0:02:24.560 --> 0:02:27.280 is used to burn in the circuit onto the surface 0:02:27.320 --> 0:02:29.519 of the chip to give it its function. Right, is 0:02:29.560 --> 0:02:34.320 a very important step. That's the fundamental blueprint, and they 0:02:34.360 --> 0:02:36.720 were saying we didn't get it wrong, but when it 0:02:36.720 --> 0:02:39.560 came to manufacturing, it didn't produce as many good chips 0:02:39.560 --> 0:02:41.639 as we wanted, so we made some tweaks to that. 0:02:41.880 --> 0:02:44.400 Didn't have to redesign it, but made some tweaks, and 0:02:44.440 --> 0:02:46.120 that is helping us to get a better yield. 0:02:46.400 --> 0:02:49.480 It's worth noting at this stage the stocks down almost 0:02:49.520 --> 0:02:51.640 seven percent in after hours, had been down more at 0:02:51.680 --> 0:02:54.360 the conclusion of the call. I think basically because we 0:02:54.360 --> 0:02:57.560 didn't learn enough about Blackwell. What they said was in 0:02:57.600 --> 0:03:01.040 the fiscal fourth quarter of this year twenty five, there 0:03:01.040 --> 0:03:05.840 will be several billion dollars of sales through Blackwell. Why 0:03:05.880 --> 0:03:07.960 does the market want more and what does it want? 0:03:08.200 --> 0:03:11.800 They wanted a reassurance from in videos management, and they 0:03:11.840 --> 0:03:15.519 wanted reassurance in the form of details. They wanted, you know, 0:03:15.800 --> 0:03:18.200 gens to Puddy's arm around everybody and say, don't worry, 0:03:18.280 --> 0:03:20.480 it'll be fine. This is how much I'm going to get. 0:03:20.720 --> 0:03:24.040 They asked consistently, We're asked questions of how many billions 0:03:24.160 --> 0:03:27.800 and when will those billions exactly come in and collect. Kretz, 0:03:28.080 --> 0:03:32.200 the CFO and Jensen Wang essentially avoided that question and 0:03:32.360 --> 0:03:34.480 refuse to give that precise reassurance. 0:03:35.600 --> 0:03:37.800 We're showing some of the after ours reaction, not just 0:03:37.840 --> 0:03:39.920 in in video itself, but some of its peers, both 0:03:39.960 --> 0:03:43.760 on the chip making side the server equipment providers, and 0:03:44.080 --> 0:03:46.960 I think AMD and ARM in particular are very noteworthy. 0:03:47.280 --> 0:03:48.960 Bloombogt and King stay with us. I want to go 0:03:49.000 --> 0:03:51.680 out to Chicago and Bloomberg's Ryan for Lostelka on I 0:03:51.760 --> 0:03:54.840 Equities team and Ryan. That's the broad summary from me 0:03:55.080 --> 0:03:57.680 about the names moving and after Hours there's probably a 0:03:57.680 --> 0:04:00.000 bigger picture after ours movement in the markets as well. 0:04:00.080 --> 0:04:02.680 I'll start with Nvidia and work outward from that. 0:04:03.640 --> 0:04:05.960 Sure, well, one thing I would say about in videos 0:04:06.000 --> 0:04:08.440 after hours decline is that it does come after a 0:04:08.600 --> 0:04:11.000 very strong year to day performance. I think it closed 0:04:11.040 --> 0:04:13.000 up more than one hundred and fifty percent this year, 0:04:13.280 --> 0:04:16.120 So even though the re forecast was maybe a little 0:04:16.160 --> 0:04:18.800 bit shy of some of the most optimistic expectations, it 0:04:18.800 --> 0:04:21.600 did beat expectations. And it is, you know, coming off 0:04:21.640 --> 0:04:24.080 such a huge gain. So it's not necessarily surprising to 0:04:24.120 --> 0:04:26.720 see a little bit of a consolidation now. And I 0:04:26.760 --> 0:04:28.600 think even the decline is a little bit less and 0:04:28.640 --> 0:04:31.960 the options market was anticipating. So just some context there 0:04:31.960 --> 0:04:34.159 for looking at the declient. But you're right that pretty 0:04:34.200 --> 0:04:37.960 much we are seeing widespread weakness following this report. All 0:04:37.960 --> 0:04:41.160 the megacaps are, you know, modestly lower. We are seeing 0:04:41.240 --> 0:04:44.080 much more pronounced weakness in other chip makers and chip 0:04:44.120 --> 0:04:46.960 design companies and so forth. So yeah, certainly it does 0:04:47.000 --> 0:04:49.320 seem like the initial read through here is negative, but 0:04:49.360 --> 0:04:51.640 again it does come after a very strong start to 0:04:51.680 --> 0:04:52.000 the year. 0:04:53.080 --> 0:04:55.680 We made a big deal about this earnings print for 0:04:55.800 --> 0:04:58.640 quite a long time. I look at something like an 0:04:58.640 --> 0:05:02.800 investco QQQ, the ETF that tracks the nast that one hundred, 0:05:03.120 --> 0:05:05.479 and I think it's down around a percentage point, right. 0:05:05.560 --> 0:05:09.400 But in truth, Ryan, in the market's reaction, was this 0:05:09.520 --> 0:05:11.800 the macro level event that we thought it was going 0:05:11.839 --> 0:05:12.039 to be. 0:05:14.200 --> 0:05:16.200 That's a great question. I would say that it's kind 0:05:16.200 --> 0:05:19.039 of close enough in line with expectations, even if it 0:05:19.120 --> 0:05:21.440 is a little bit shy the most optimistic ones that 0:05:21.480 --> 0:05:23.440 I don't think this is going to really cause people 0:05:23.440 --> 0:05:26.280 to really change how they're allocating, change their opinions on 0:05:26.680 --> 0:05:29.760 AIS as kind of fundamental secular driver. But maybe in 0:05:29.800 --> 0:05:31.920 the near term we do see a little bit of weakness. 0:05:31.960 --> 0:05:34.400 I mean, again, some of these stocks have been moving 0:05:34.480 --> 0:05:36.200 up so much there have been sort of kind of 0:05:36.240 --> 0:05:39.440 growing calls about their valuation some concerns about that. I 0:05:39.440 --> 0:05:42.080 don't know if this is the kind of absolute blowout 0:05:42.120 --> 0:05:44.400 that we'll just you know, cause people to continue piling 0:05:44.440 --> 0:05:47.200 into the way that they were doing earlier this year. 0:05:48.120 --> 0:05:50.120 There is a lot in the news cycle inclusive of 0:05:50.200 --> 0:05:51.880 and outside of in video. I think one of the 0:05:51.960 --> 0:05:54.200 names that you note in after ours is super Micro. 0:05:54.640 --> 0:05:59.120 It's down significantly. There's an video relationship to that, and 0:05:59.120 --> 0:06:01.720 then there's the news around that name in and of itself. 0:06:03.000 --> 0:06:05.960 Yeah. Absolutely so. We saw yesterday a short report came 0:06:06.000 --> 0:06:08.800 out today it's delaying the filing of the ten K. 0:06:09.600 --> 0:06:11.479 You know, both of those caused us a weakness in 0:06:11.520 --> 0:06:13.040 the stock. I think today it was down more than 0:06:13.040 --> 0:06:15.360 double digits, so you know, a lot of reasons to 0:06:15.400 --> 0:06:17.800 be concerned there in general. And it is one of 0:06:17.880 --> 0:06:20.560 Nvidia's biggest customers. I think it's the third largest, so 0:06:21.000 --> 0:06:23.040 you know, this is just you know, another reason for 0:06:23.120 --> 0:06:25.600 people who might have been kind of souring on super 0:06:25.600 --> 0:06:27.799 Micro to you know, maybe be pulling the cell button. 0:06:27.640 --> 0:06:28.040 A little bit. 0:06:29.279 --> 0:06:32.159 Bloomdo's Ryan for Selica with the after hours action out 0:06:32.160 --> 0:06:34.719 of Chicago. Thank you. Let's get the reaction from the 0:06:34.760 --> 0:06:38.400 cell side with one of the stocks relative bears. D A. 0:06:38.560 --> 0:06:41.919 Davidson's GIL Luria had a neutral rating and a street 0:06:41.960 --> 0:06:45.120 low ninety dollars price target on shares of Nvidia. He 0:06:45.200 --> 0:06:49.560 joins US now and GIL. I guess does that change 0:06:49.680 --> 0:06:52.480 your barished perspective on the stock and your main takeaway 0:06:52.720 --> 0:06:56.400 from the analyst school, the quarter. 0:06:56.320 --> 0:07:00.080 Was a very good quarter. There ability to continue to 0:07:00.080 --> 0:07:05.160 grow delivering chips at this rate at this scale is 0:07:05.240 --> 0:07:08.520 fantastic and unprecedented. I don't think there's much of a 0:07:08.600 --> 0:07:11.720 revenue issue here, of a growth issue here. I think 0:07:11.880 --> 0:07:14.720 a little bit of a pushback is probably more around 0:07:15.000 --> 0:07:17.600 the margin situation. You had a good conversation with Ian 0:07:17.680 --> 0:07:20.680 about Blackwell and some of the moving pieces part of 0:07:20.720 --> 0:07:24.640 what's happening with Blackwell. It's a new product, it's got 0:07:24.720 --> 0:07:27.200 a few hitches along the way that put pressure on 0:07:27.240 --> 0:07:31.560 gross margins, and then operating margins grew more than anticipated, 0:07:31.600 --> 0:07:35.120 so less of the top line upside flowed to the 0:07:35.120 --> 0:07:38.320 bottom line. And that's probably where investors are getting a 0:07:38.360 --> 0:07:41.480 little bit more cautious than they were before, because in 0:07:41.520 --> 0:07:45.880 the last several quarters, the huge upside to revenue flowed 0:07:45.960 --> 0:07:48.600 all the way to the bottom line, creating huge upside 0:07:48.640 --> 0:07:52.040 there to believe in and nvideo right now is to 0:07:52.080 --> 0:07:54.880 believe twenty twenty six calendar is going to be about 0:07:54.880 --> 0:07:58.040 two hundred billion of revenue and about four and a 0:07:58.120 --> 0:08:01.640 half to five dollars of earnings. That's only possible if 0:08:01.640 --> 0:08:04.600 they continue to get this type of growth on the 0:08:04.600 --> 0:08:08.280 top line and more leverage. And the second part of 0:08:08.280 --> 0:08:10.720 that equation looks a little less certain right now. 0:08:11.280 --> 0:08:11.559 Gil. 0:08:11.600 --> 0:08:14.120 When you said that two hundred billion dollar figure, Bloombergsy 0:08:14.160 --> 0:08:17.120 and King, who's covered semi conductors at this company since 0:08:17.200 --> 0:08:19.360 nineteen ninety eight, sat next to me still for what 0:08:19.400 --> 0:08:22.560 it's worth, gave a little smile. I mean, it's an 0:08:22.600 --> 0:08:25.720 astonishing figure. I think there's a lot of emphasis here 0:08:25.760 --> 0:08:28.520 on Blackwell. How did you react to the news of 0:08:28.600 --> 0:08:32.080 several billion dollars in fiscal fourth with Blackwell and the 0:08:32.160 --> 0:08:36.360 explanation that this was a production issue impacting the ramp, 0:08:36.440 --> 0:08:39.920 not a sort of core design issue around the product itself. 0:08:41.160 --> 0:08:42.960 So, first of all, our estimate is far less than 0:08:43.000 --> 0:08:45.520 two hundred billion for twenty twenty six. 0:08:45.440 --> 0:08:47.800 But that's you're a relative bear, Gill, and I got 0:08:47.800 --> 0:08:48.360 that bit right. 0:08:48.400 --> 0:08:52.800 I think in terms of Blackwell, what happened to in 0:08:52.840 --> 0:08:54.319 n videos. They used to be on a two year 0:08:54.360 --> 0:08:56.719 cycle for the data center product, and they decided to 0:08:56.760 --> 0:09:00.240 aggressively move to a one year cycle and then put 0:09:00.240 --> 0:09:03.120 a lot of new feature functionality and bundle into the 0:09:03.120 --> 0:09:07.160 Blackwell platform. So that was a very aggressive agenda, and 0:09:07.200 --> 0:09:10.520 it's not that surprising that there's some challenges that they 0:09:10.559 --> 0:09:14.480 have to encounter. If they can still ship in their 0:09:14.559 --> 0:09:18.439 fiscal fourth quarter, that's a very good sign. And by 0:09:18.480 --> 0:09:21.800 the way, their customers don't really care that much. Their 0:09:21.840 --> 0:09:24.200 customers are just going to continue to buy the latest 0:09:24.240 --> 0:09:26.640 and greatest, and if that's the Age two hundred, they'll 0:09:26.679 --> 0:09:29.120 just continue to buy h two hundred. They have made 0:09:29.120 --> 0:09:34.160 that abundantly clear on their conference calls, Microsoft, Meta, Amazon, Google, Tesla. 0:09:34.440 --> 0:09:37.400 Elon Musk made it abundantly clear they'll buy as much 0:09:37.440 --> 0:09:40.640 GPUs as in Video can produce, at least this year, 0:09:41.120 --> 0:09:43.679 because they're saying the demand signals are there. As long 0:09:43.679 --> 0:09:46.000 as the demand signals are there, they'll buy whatever and 0:09:46.120 --> 0:09:48.480 video is selling, and towards the end of the year, 0:09:48.760 --> 0:09:50.319 a piece of that will be Blackwell. 0:09:51.520 --> 0:09:54.560 Gil Jensen Wong was it pains and gave a detailed 0:09:54.600 --> 0:09:58.440 answer that basically summarizing video as a systems vendor. The 0:09:58.480 --> 0:10:02.920 discussion around Envy link, CPUs, everything beyond GPU. How much 0:10:03.000 --> 0:10:05.719 credit do you assign the company for that, the sort 0:10:05.760 --> 0:10:07.760 of complete systems that they sell. 0:10:08.880 --> 0:10:12.160 A tremendous amount of credit. Let's start with the beginning. 0:10:12.840 --> 0:10:17.040 When history is written, we'll talk about how Jensen Wangen 0:10:17.160 --> 0:10:21.199 Nvidia created the capabilities that make generative AI possible today, 0:10:21.200 --> 0:10:23.640 in a similar way that Elon Musk is responsible for 0:10:24.000 --> 0:10:27.240 the electric vehicle. And part of the genius wasn't just 0:10:27.320 --> 0:10:30.600 seeing that this is coming, that accelerate computers coming, AI 0:10:30.720 --> 0:10:34.360 capabilities is coming well before anybody else did. It's also 0:10:34.480 --> 0:10:37.920 realizing that in order to create a moat around the chip, 0:10:37.960 --> 0:10:40.679 they need to do a couple of things. One is 0:10:41.080 --> 0:10:44.120 to own the proprietary software around that KUDA, but the 0:10:44.200 --> 0:10:47.920 other thing is to enhance the bundle as much as possible. 0:10:48.320 --> 0:10:52.240 These GPUs work in a raise when they work together, 0:10:52.520 --> 0:10:55.679 ten thousands or sometimes tens of thousands of GPUs together. 0:10:56.200 --> 0:10:59.720 When you box those up in a box with multiple 0:10:59.800 --> 0:11:05.760 g CPUs, wiring, cabling, and firmware, you make it harder 0:11:05.800 --> 0:11:07.760 for your competitors to catch up. And they've done a 0:11:07.800 --> 0:11:09.040 remarkable job of that. 0:11:10.120 --> 0:11:12.520 I also wonder about all the rest. The story thus 0:11:12.559 --> 0:11:16.040 far has been about the hyperscale is a meta five names, 0:11:16.160 --> 0:11:19.720 essentially being the core customer of Nvidia. We got some 0:11:19.920 --> 0:11:23.120 detail on low double digit billions sales this year in 0:11:23.200 --> 0:11:26.560 Sovereign AI. Do you see evidence that the market is 0:11:26.640 --> 0:11:29.119 broadening out beyond just the cloud providers? 0:11:30.320 --> 0:11:30.959 Not really. 0:11:31.040 --> 0:11:34.200 It seems like those five are still a substantial amount 0:11:34.240 --> 0:11:38.080 of the revenue. And those five have in their most 0:11:38.080 --> 0:11:41.000 recent calls said some interesting things. One of them is 0:11:41.000 --> 0:11:44.400 that they consistently said they are over investing. So they 0:11:44.400 --> 0:11:46.880 said they definitely want to invest more. We're seeing that 0:11:46.920 --> 0:11:50.000 in results today. But they also use the term over investing. 0:11:50.400 --> 0:11:52.960 Over investing is something you do before you do the 0:11:53.000 --> 0:11:55.920 other thing, and they made it clear that they're only 0:11:56.000 --> 0:11:58.760 going to do so when they have the demand signals. 0:11:59.200 --> 0:12:04.120 If or whatever reason, demand for AI capacity goes down, 0:12:04.679 --> 0:12:10.120 for instance, because compute doesn't continue to drive performance of 0:12:10.200 --> 0:12:15.320 foundational models, you're gonna see those same five actors pull 0:12:15.440 --> 0:12:17.960 back and say, you know, maybe we have enough data 0:12:18.000 --> 0:12:20.680 center capacity, or maybe what we have in the pipeline 0:12:20.720 --> 0:12:24.000 is enough. And that's the concern for twenty twenty five 0:12:24.040 --> 0:12:27.920 and twenty six is that those very companies could do that. 0:12:28.800 --> 0:12:32.359 Beyond that, these are the same companies that are developing 0:12:32.360 --> 0:12:35.560 their own chips. They're at various stages of this, but 0:12:35.679 --> 0:12:38.720 Google's TPUs are probably as good as in videos. So 0:12:38.960 --> 0:12:41.479 in terms of their internal use and selling to customers 0:12:41.600 --> 0:12:45.720 such as Apple, Google's already caught up, Amazon's catching up, 0:12:46.280 --> 0:12:50.680 Meta and Microsoft are earlier stages. But these same companies 0:12:51.040 --> 0:12:54.400 that are buying every GPU from video they can will 0:12:54.400 --> 0:12:56.760 be the ones that will buy possibly less next year 0:12:56.760 --> 0:12:57.480 in the year after that. 0:12:58.240 --> 0:13:00.680 I just want to reflect for our TV audience there 0:13:00.720 --> 0:13:03.079 are people tuned in around the world. My understanding is 0:13:03.120 --> 0:13:05.959 there are watch parties in cities from New York to London. 0:13:06.440 --> 0:13:09.840 For this earnings, We've hyped it up for weeks. Probably 0:13:09.920 --> 0:13:12.559 this is the most unique earnings print that I've covered 0:13:12.559 --> 0:13:16.079 in my career. What about for you in the semiconductor space, 0:13:16.360 --> 0:13:17.720 Why is this so different? 0:13:18.720 --> 0:13:21.600 Well, so, first of all, I tried to participate by 0:13:21.640 --> 0:13:26.439 wearing in video green today. But importantly and video is 0:13:26.480 --> 0:13:29.719 important because of how much they've added to the cumulative 0:13:29.720 --> 0:13:32.880 earnings of this in p. Five hundred and the cumulative performance. 0:13:33.360 --> 0:13:38.559 But I wouldn't necessarily extrapolate their success or sometimes setbacks 0:13:38.800 --> 0:13:41.320 to the rest of the market because what they do 0:13:41.440 --> 0:13:45.480 is very specific. The growth and data center construction and 0:13:45.880 --> 0:13:52.200 data center a GPU fulfillment into those data centers is 0:13:52.240 --> 0:13:55.720 a very small set of companies. I'd argue and Video 0:13:55.800 --> 0:13:58.280 is sucking the oxygen for the room from a lot 0:13:58.320 --> 0:14:01.880 of other technology companies. Well, we celebrate with Nvideo because 0:14:01.880 --> 0:14:04.800 they've added so much value. I wouldn't necessarily read through 0:14:04.840 --> 0:14:06.040 to the rest of technology. 0:14:07.080 --> 0:14:11.880 Gilerya of DA Davidson, the relatively most bearish or perhaps 0:14:11.960 --> 0:14:14.800 the least bullish name on the street, Thank you so 0:14:14.880 --> 0:14:18.600 welcome to this special edition of Bloomberg Technology for our 0:14:18.600 --> 0:14:21.360 TV and radio audience around the world. Joining me now 0:14:21.840 --> 0:14:25.920 is Jensen One, CEO of Nvidia, straight from the analyst's call, 0:14:25.960 --> 0:14:29.200 and Jensen, good evening, good afternoon to you. I think 0:14:29.280 --> 0:14:33.040 the market wanted more on Blackwell, they wanted more specifics, 0:14:33.080 --> 0:14:34.840 and I'm trying to go through all of the call 0:14:34.880 --> 0:14:38.560 and the transcript. It seems like a very clearly this 0:14:38.680 --> 0:14:41.480 was a production issue and not a fundamental design issue 0:14:41.480 --> 0:14:45.080 with Blackwell, but the deployment in the real world, what 0:14:45.160 --> 0:14:47.440 does that look like? Tangibly and is there a sort 0:14:47.440 --> 0:14:50.560 of delay in the timeline of that deployment and thus 0:14:50.640 --> 0:14:51.840 revenue from that product. 0:14:54.320 --> 0:15:00.360 I let's see, that's just the fact that I was 0:15:00.400 --> 0:15:03.800 so clear and it wasn't clear enough kind of tripped 0:15:03.840 --> 0:15:04.840 me up there right away. 0:15:05.160 --> 0:15:08.520 And so let's see, we made a mass change to 0:15:08.560 --> 0:15:12.480 improve the yield. Functionality of Blackwell is wonderful. We're sampling 0:15:12.520 --> 0:15:18.240 Blackwell all over the world today. We show people giving 0:15:18.280 --> 0:15:21.120 tours to people of the Blackwall systems that we have 0:15:21.200 --> 0:15:24.960 up and running. You could find pictures of Blackwall systems 0:15:25.480 --> 0:15:30.160 all over the web. We have started volume production. Volume 0:15:30.200 --> 0:15:34.440 production will ship in Q four. Q four, we will 0:15:34.480 --> 0:15:38.960 have billions of dollars of Blackwell revenues and. 0:15:40.320 --> 0:15:43.160 We will ramp from there. We will ramp from there. 0:15:43.640 --> 0:15:48.920 The demand for Blackwell far exceeds its supply, of course 0:15:49.000 --> 0:15:52.800 in the beginning, because the demand is so great, But 0:15:53.040 --> 0:15:55.760 we're going to have lots and lots of supply and 0:15:56.120 --> 0:15:59.240 we will be able to ramp. Starting in Q four, 0:16:00.040 --> 0:16:03.000 we have billions of dollars of revenues and we'll ramp 0:16:03.000 --> 0:16:05.040 from there into Q one, into Q two and two 0:16:05.080 --> 0:16:07.000 next year. We're going to have a great next year 0:16:07.000 --> 0:16:07.280 as well. 0:16:08.560 --> 0:16:11.800 Jensen, what is the demand for accelerated computing beyond the 0:16:11.880 --> 0:16:13.640 hyperscalers and meta. 0:16:15.200 --> 0:16:20.000 Hyperscalers represent about forty five percent of our total data 0:16:20.000 --> 0:16:25.400 center business. We're relatively diversified today. We have hyperscalers, we 0:16:25.440 --> 0:16:31.720 have Internet service providers, we have sovereign AIS, we have 0:16:33.040 --> 0:16:39.680 industries enterprises, so it's fairly fairly diversified. A site outside 0:16:39.720 --> 0:16:45.760 of hyperscalers is the other fifty five percent. Now, the 0:16:45.840 --> 0:16:49.680 application use across all of that, all of that data 0:16:49.680 --> 0:16:55.560 center starts with accelerated computing. Accelerated computing does everything, of course, 0:16:56.040 --> 0:17:00.320 from well the models the things that we know about, 0:17:00.360 --> 0:17:04.400 which is generative AI, and that gets most of the attention. 0:17:05.119 --> 0:17:09.920 But at the core we also do database processing, pre 0:17:10.080 --> 0:17:15.119 and post processing of data before you use it for 0:17:15.280 --> 0:17:21.560 generative AI, trans coding, scientific simulations, computer graphics of course, 0:17:21.680 --> 0:17:25.760 image processing of course. And so there's tons of applications 0:17:25.800 --> 0:17:30.199 that people use are accelerated computing for, and one of 0:17:30.200 --> 0:17:33.880 them is generative AI. And so let's see what else 0:17:33.880 --> 0:17:34.480 can I say? 0:17:35.280 --> 0:17:36.480 I think that's the. 0:17:36.640 --> 0:17:39.639 Lever jump in Jensen. Please on sovereign AI. You and 0:17:39.640 --> 0:17:42.400 I've talked about that before and it was so interesting 0:17:42.440 --> 0:17:45.800 to hear something behind it that in this fiscal year 0:17:46.200 --> 0:17:48.000 there will be low double digit I think you said 0:17:48.040 --> 0:17:51.119 billions of dollars in sovereign AI sales. But to the 0:17:51.200 --> 0:17:53.960 lay person, what does that mean? It means deals with 0:17:54.040 --> 0:17:55.880 specific governments, if so, where. 0:17:57.480 --> 0:18:04.640 It's not necessarily Sometimes it's deals with particular regional service 0:18:04.680 --> 0:18:08.320 provider that's been funded by the government. And oftentimes that's 0:18:08.359 --> 0:18:11.000 the case in a case of in the case of Japan, 0:18:11.080 --> 0:18:15.880 for example, the the Japanese government came out and UH 0:18:16.440 --> 0:18:21.480 offered UH subsidies of a couple of billion dollars. I 0:18:21.520 --> 0:18:26.639 think for several different internet companies and telcos to be 0:18:26.680 --> 0:18:32.240 able to fund their AI infrastructure. UH India has a 0:18:32.280 --> 0:18:37.880 sovereign AI initiative going and they're building their AI infrastructure. Canada, 0:18:39.119 --> 0:18:47.920 the UK, France, Italy, missing somebody, Singapore, Malaysia, UH. You know, 0:18:48.040 --> 0:18:53.200 a large number of countries are subsidizing their regional data 0:18:53.200 --> 0:18:57.160 centers so that they could become able to build out 0:18:57.160 --> 0:19:03.880 their AI infrastructure. They recognize that their countri's knowledge, their 0:19:03.880 --> 0:19:08.960 countries data digital data is also their natural resource, not 0:19:09.080 --> 0:19:11.000 just the land they're sitting on, not just the air 0:19:11.040 --> 0:19:15.480 above them. But they realize now that their digital knowledge 0:19:15.560 --> 0:19:19.960 is part of their natural and national resource, and they 0:19:19.960 --> 0:19:23.879 are to harvest that and process that and transform it 0:19:23.920 --> 0:19:29.320 into their national digital intelligence. And so this is what 0:19:29.359 --> 0:19:31.720 we call sovereign AI. You could imagine almost every single 0:19:31.760 --> 0:19:36.600 country in the world will eventually recognize this and build 0:19:36.600 --> 0:19:37.840 out their AI infrastructure. 0:19:38.640 --> 0:19:41.359 Jenson, you use the word resource, and that makes me 0:19:41.400 --> 0:19:44.320 think about the energy requirements here. I think on the 0:19:44.320 --> 0:19:47.200 cool you talk about how the next generation models will 0:19:47.240 --> 0:19:50.639 have many orders of magnitude greater compute needs. But how 0:19:50.640 --> 0:19:53.359 will the energy needs increase and what is the advantage 0:19:53.400 --> 0:19:58.080 you feel in Vidia has in that sense, Well, the. 0:19:58.119 --> 0:20:01.000 Most important thing that we do is is increase the 0:20:01.040 --> 0:20:05.080 performance of and increase the performance and efficiency of our 0:20:05.119 --> 0:20:10.320 next generation. So Blackwell is many times more performance than 0:20:10.480 --> 0:20:14.120 Hopper at the same level of power used, and so 0:20:14.200 --> 0:20:18.639 that's energy efficiency, more performance with the same amount of power, 0:20:18.760 --> 0:20:20.560 or same performance at a lower power. 0:20:21.119 --> 0:20:22.240 And that's number one. 0:20:22.840 --> 0:20:26.600 And the second is using luca cooling. We support air 0:20:26.640 --> 0:20:29.000 cool we support air cooling, we support liqual cooling, but 0:20:29.040 --> 0:20:31.359 liqual cooling is a lot more energy efficient. And so 0:20:32.200 --> 0:20:34.639 the combination of all of that, you're going to get 0:20:34.640 --> 0:20:37.000 a pretty large, pretty large step up. 0:20:37.720 --> 0:20:39.280 But the important thing to also. 0:20:39.119 --> 0:20:41.800 Realize is that AI doesn't really care where it goes 0:20:41.840 --> 0:20:45.359 to school, and so increasingly we're going to see AI 0:20:45.520 --> 0:20:49.000 be trained somewhere else, have that model come back and 0:20:49.040 --> 0:20:52.840 be used near the population, or even running on your 0:20:53.280 --> 0:20:54.480 PC or your phone. 0:20:54.520 --> 0:20:55.120 And so we're going. 0:20:55.040 --> 0:20:57.840 To train large models, but the goal is not to 0:20:57.920 --> 0:21:00.879 run the large models necessarily all the time. You can 0:21:00.920 --> 0:21:03.679 surely do that for some of the premium services and 0:21:04.119 --> 0:21:08.359 the very high value AIS, but it's very likely that 0:21:08.440 --> 0:21:11.320 these large models would then help to train and teach 0:21:11.400 --> 0:21:14.520 smaller models, and what we'll end up doing is have 0:21:14.560 --> 0:21:18.840 one large, few large models that are able to train 0:21:18.960 --> 0:21:21.200 a whole bunch of small models and they run everywhere. 0:21:22.840 --> 0:21:27.880 Jensen, you explain clearly that demand to build generative AI 0:21:27.960 --> 0:21:31.000 product on models or even at the GPU level is 0:21:31.080 --> 0:21:35.080 greater than current supply. In black Vel's case in particular, 0:21:35.320 --> 0:21:38.159 explain the supply dynamics to me for your products and 0:21:38.200 --> 0:21:41.440 whether you see an improvement sequentially quarter on quarter or 0:21:41.480 --> 0:21:43.640 at some point by the end of fiscal year into 0:21:43.680 --> 0:21:44.160 next year. 0:21:45.440 --> 0:21:48.720 Well, the fact that we're growing would suggest that our 0:21:48.840 --> 0:21:53.560 supply is improving and our supply chain is quite large, 0:21:53.960 --> 0:21:56.359 one of the largest supply chains in the world. We 0:21:56.440 --> 0:22:00.440 have incredible partners and they're doing a great job supporting 0:22:00.480 --> 0:22:03.320 us in our growth. As you know, we're one of 0:22:03.320 --> 0:22:07.040 the fastest growing technology companies in history, and none of 0:22:07.040 --> 0:22:10.240 that would have been possible without very strong demand but 0:22:10.320 --> 0:22:13.359 also very strong supply. We're expecting Q three to have 0:22:13.400 --> 0:22:15.720 more supply than Q two, We're expecting Q four to 0:22:15.760 --> 0:22:17.760 have more supply than Q three, and we're expecting Q 0:22:17.800 --> 0:22:19.800 one to have more supply than Q four, And so 0:22:19.840 --> 0:22:22.760 I think our supply, our supply condition going into next 0:22:22.800 --> 0:22:25.480 year will be in it will be a large improvement 0:22:25.520 --> 0:22:31.560 over this last year with respect to demand. Blackwell is 0:22:31.640 --> 0:22:34.840 just such a leap and there's several things that are happening, 0:22:35.440 --> 0:22:38.760 you know, just the Foundation model makers themselves. The size 0:22:38.760 --> 0:22:42.119 of the foundation models are growing from hundreds of billions 0:22:42.240 --> 0:22:46.520 parameters to trillions of parameters. They're also learning more languages. 0:22:46.920 --> 0:22:49.880 Instead of just learning human language, they're learning the language 0:22:49.880 --> 0:22:55.119 of images and sounds and videos, and they're even learning 0:22:55.520 --> 0:22:58.800 the language of three D graphics. And whenever they are 0:22:58.800 --> 0:23:01.760 able to learn these languages, they can understand what they see, 0:23:01.800 --> 0:23:04.560 but they can also generate what they're asked to generate, 0:23:05.040 --> 0:23:08.160 and so they're learning the language of proteins and chemicals 0:23:08.240 --> 0:23:11.480 and physics. You know, it could be fluids and it 0:23:11.520 --> 0:23:14.800 could be particle physics, and so they're learning all kinds 0:23:14.840 --> 0:23:18.840 of different languages, learn the meaning of what we call modalities, 0:23:18.840 --> 0:23:20.320 but basically learning languages. 0:23:20.760 --> 0:23:24.240 And so these models are growing in size. 0:23:24.280 --> 0:23:27.639 They're learning from more data, and there are more model 0:23:27.640 --> 0:23:30.800 makers then there was a year ago. And so the 0:23:30.880 --> 0:23:33.680 number of model makers have grown substantially because of all 0:23:33.680 --> 0:23:36.520 these different modalities. And so that's just one, just the 0:23:36.520 --> 0:23:41.400 frontier model. The foundation model makers themselves haven't really grown tremendously. 0:23:41.840 --> 0:23:46.159 And then the generative AI market has really diversified, you know, 0:23:46.280 --> 0:23:50.719 beyond the Internet service makers to startups and now enterprises 0:23:50.760 --> 0:23:54.960 are jumping in and different countries are jumping in, so 0:23:55.240 --> 0:23:56.359 the demand is really growing. 0:23:57.000 --> 0:23:59.359 Jensen, I'm sorry to cut you off. I will lose 0:23:59.560 --> 0:24:03.840 your time soon. You've also diversified. And when I said 0:24:03.880 --> 0:24:05.639 to our audience you were coming on, I got so 0:24:05.720 --> 0:24:09.320 many questions. Probably the most common one is what is 0:24:09.359 --> 0:24:12.359 in Nvidia. We talked about you as a systems vendor, 0:24:12.600 --> 0:24:16.000 but so many points on in Vidia GPU cloud, and 0:24:16.040 --> 0:24:19.280 I want to ask, finally, do you have plans to 0:24:19.280 --> 0:24:21.600 become literally a cloud compute provider. 0:24:23.640 --> 0:24:23.720 No. 0:24:25.320 --> 0:24:29.800 Our GPU cloud was designed to be the best version 0:24:30.680 --> 0:24:35.320 of Nvidia Cloud that's built within each cloud. Nvidio DGX 0:24:35.320 --> 0:24:42.880 cloud is built inside GCP, inside Azure, inside AWS, inside OCI, 0:24:43.600 --> 0:24:46.520 and so we build our clouds within THEIRS so that 0:24:46.560 --> 0:24:50.000 we can implement our best version of our cloud, work 0:24:50.080 --> 0:24:54.000 with them to make that cloud, that that infrastructure, that 0:24:54.040 --> 0:24:58.800 AI infrastructure, in video infrastructure, as performance, as great TCO 0:24:58.880 --> 0:25:04.800 as possible. And so that strategy has worked incredibly well. 0:25:05.320 --> 0:25:08.760 And of course we are large consumers of it because 0:25:08.800 --> 0:25:11.200 we create a lot of AI ourselves, because our chips 0:25:11.240 --> 0:25:14.000 aren't possible to design without AI, our software is not 0:25:14.040 --> 0:25:16.520 possible to write without AI, and so we use it 0:25:16.560 --> 0:25:19.919 ourselves tremendous, you know, tremendous amount of it self driving cars, 0:25:19.920 --> 0:25:22.480 the general robotics work that we're doing, the omniverse work 0:25:22.480 --> 0:25:25.639 that we're doing. So we're using the DGX cloud for ourselves. 0:25:26.520 --> 0:25:30.360 We also use it for an AI foundry. We make 0:25:30.440 --> 0:25:34.639 AI models for companies that we'd like to have expertise 0:25:34.640 --> 0:25:36.200 in doing so, and so we are AI. 0:25:36.720 --> 0:25:37.320 We're an AI. 0:25:37.440 --> 0:25:40.840 We're a foundry for AI like tsmcs a foundry for 0:25:40.840 --> 0:25:44.600 our chips, and so there are three fundamental reasons why 0:25:44.640 --> 0:25:47.520 we do it. One is to have the best version 0:25:47.640 --> 0:25:50.240 of Nvidia inside all the clouds. Two because we're a 0:25:50.320 --> 0:25:53.240 large consumer to ourselves, and third because we use it 0:25:53.240 --> 0:25:55.440 for AI foundry for to help every other company. 0:25:56.880 --> 0:25:59.240 Jensen one, CEO and Video. I want to thank you 0:25:59.280 --> 0:26:02.360 for your time the extended conversation straight off the earnings call. 0:26:02.720 --> 0:26:06.160 Thank you in videos earnings, Fiscal is good to see 0:26:06.160 --> 0:26:09.000 you too. Fiscal second quarter done. I point out that 0:26:09.080 --> 0:26:11.800 in after hours the stock is down still almost seven percent. 0:26:12.040 --> 0:26:14.720 There'll be a lot of analysis to be done overnight 0:26:14.760 --> 0:26:16.960 by the cell side, by the byside, and we'll bring 0:26:17.000 --> 0:26:20.280 you the best from Bloomberg's reporters and editors from San Francisco. 0:26:20.600 --> 0:26:30.000 This is Bloomberg Technology