WEBVTT - Instant Reaction: Nvidia Delivers Solid Earnings 0:00:02.520 --> 0:00:08.039 Bloomberg Audio Studios, podcasts, radio news fourth. 0:00:07.840 --> 0:00:11.200 Quarter revenue has beat estimates. Fourth quarter data center revenue 0:00:11.200 --> 0:00:14.080 coming in at thirty five point six billion dollars thirty 0:00:14.080 --> 0:00:17.279 four point oh nine billion dollars. Fourth quarter revenue that 0:00:17.360 --> 0:00:21.160 top line thirty nine point three billion dollars, exceeding estimates 0:00:21.160 --> 0:00:24.680 of thirty eight point twenty five billion dollars. Now coming 0:00:24.720 --> 0:00:27.880 to that revenue estimate for the first quarter. In Video 0:00:27.960 --> 0:00:31.960 C's first quarter revenue plus or minus two percent, uh, 0:00:32.040 --> 0:00:35.440 the estimate was for forty two point twenty six billion dollars. 0:00:35.479 --> 0:00:38.040 Fourth quarter data center revenue coming in at thirty five 0:00:38.080 --> 0:00:41.360 point six percent, the company saying that it ramped up 0:00:41.400 --> 0:00:44.560 massive scale production of Blackwell AI. In the after hour 0:00:44.640 --> 0:00:47.160 shares hire right now by about three point four percent. 0:00:47.320 --> 0:00:51.000 Yeah, actually in Video saying demand for Blackwell is amazing, 0:00:51.120 --> 0:00:53.520 So that's what we talked about in terms of that transition. 0:00:54.120 --> 0:00:56.800 But again the key metrics, as Tim said, stock is 0:00:56.880 --> 0:00:59.639 up almost four percent here in the aftermarket. The first 0:00:59.720 --> 0:01:04.559 quarter revenue, so looking ahead outlook forty three billion dollars 0:01:04.600 --> 0:01:07.720 plus or minus two percent. The estimate was for forty 0:01:07.760 --> 0:01:10.520 two point three billion, and again fourth quarter data center 0:01:10.560 --> 0:01:13.319 so looking at the quarter that they just wrapped up 0:01:13.480 --> 0:01:15.759 data center revenue. That's the bulk of their business, that's 0:01:15.840 --> 0:01:19.520 those hyperscalers and more. Thirty five point six billion. That 0:01:19.680 --> 0:01:22.520 is a beat. Thirty four point zero nine billion was 0:01:22.560 --> 0:01:25.679 what the street was expecting first quarter, justin gross margin 0:01:25.760 --> 0:01:29.640 again looking forward seventy point five percent to seventy one 0:01:29.640 --> 0:01:32.200 point five percent. Street was looking for a little bit more. 0:01:32.240 --> 0:01:35.000 This is a measure of profitability, and the estimate on 0:01:35.000 --> 0:01:37.000 the street was seventy two point one percent. Again, the 0:01:37.040 --> 0:01:39.319 stock now not up as much, In fact, it's down 0:01:39.480 --> 0:01:39.880 just a hair. 0:01:40.000 --> 0:01:41.880 Yeah, shares climbed as much as four percent after this 0:01:41.920 --> 0:01:44.440 first quarter revenue forecast top estimates. We're getting some more 0:01:44.480 --> 0:01:47.800 commentary too, in Vidia saying that Blackwell is achieving billions 0:01:47.840 --> 0:01:50.200 of dollars in sales in the first quarter. The company 0:01:50.240 --> 0:01:53.920 also saying that it's successfully ramped up Blackwell massive production. 0:01:54.400 --> 0:01:57.520 In Vidia also coming out and saying, as Carol mentioned, 0:01:57.560 --> 0:02:01.600 demand for Blackwell is amazing. We're continuing to get more 0:02:01.680 --> 0:02:04.720 numbers here too. Fourth quarter compute revenue coming in above 0:02:04.800 --> 0:02:07.680 estimates at thirty two point five six billion dollars. Fourth 0:02:07.760 --> 0:02:10.720 quarter networking revenue coming in just shy of estimates a 0:02:10.800 --> 0:02:12.320 three point oh two billion dollars. 0:02:12.440 --> 0:02:15.079 Ye, now the stock is down about one percent here, 0:02:15.120 --> 0:02:17.200 it's bouncing around a lot, not surprising a lot of 0:02:17.240 --> 0:02:19.320 investors in this name. So you're going to see a 0:02:19.360 --> 0:02:22.399 lot of movement as we get tick by tick, headline 0:02:22.400 --> 0:02:25.400 by headline. But again, the key ones right now, it 0:02:25.440 --> 0:02:28.800 has to do with the fourth quarter data center revenue. 0:02:28.800 --> 0:02:31.760 Excuse me, thirty five point six billion. Again, that's above 0:02:31.800 --> 0:02:34.440 what the Street was expecting. Street was looking for thirty 0:02:34.440 --> 0:02:37.240 four point zero nine billion. And again key is the 0:02:37.280 --> 0:02:40.840 revenue outlook forty three point zero billion plus or minus 0:02:40.840 --> 0:02:44.280 two percent. Street was at forty two point three billion. 0:02:44.360 --> 0:02:44.880 And again we. 0:02:44.919 --> 0:02:47.760 Got some context context, if you will, in terms of 0:02:47.800 --> 0:02:51.520 demand for Blackwell being amazing, successfully ramping up the Blackwell 0:02:51.520 --> 0:02:54.560 massive production, Blackwell achieving billions of dollars in sales in 0:02:54.560 --> 0:02:57.080 the first quarter and ramping up the massive scale production 0:02:57.200 --> 0:02:58.040 of Blackwell AI. 0:02:58.080 --> 0:02:59.520 If the company else is saying it to lilivered eleven 0:02:59.520 --> 0:03:02.959 billion dollars of Blackwell architecture revenue in the fourth quarter. 0:03:03.400 --> 0:03:05.960 Joining us right now, Red Brown is here in our 0:03:06.000 --> 0:03:09.160 New York studio, Bloomberg News Earnings reporter Caroline Hyde, co 0:03:09.200 --> 0:03:14.280 host of Bloomberg Technology, also here in studio. You know, Red, 0:03:14.280 --> 0:03:16.120 I want to bring you into this conversation. You keep 0:03:16.120 --> 0:03:18.480 a watch on earnings, just you know, your initial thoughts 0:03:18.600 --> 0:03:20.000 and what we're seeing in the trade because the stock 0:03:20.080 --> 0:03:21.040 is bouncing around a lot. 0:03:21.200 --> 0:03:22.560 Yeah, I mean I do feel like we got to 0:03:22.600 --> 0:03:25.120 like kind of take a cyber relief here because they 0:03:25.160 --> 0:03:27.800 did check off all of the boxes kind of what 0:03:27.880 --> 0:03:31.440 Kun John answer the question marks Kun John has kind 0:03:31.440 --> 0:03:34.440 of touched on there. The outlook is solid, the fourth 0:03:34.560 --> 0:03:36.520 quarter was solid, and there seems to be real kind 0:03:36.560 --> 0:03:40.000 of optimism around the progress with Blackwell. So yeah, I 0:03:40.040 --> 0:03:42.320 think overall this kind of was in line with what 0:03:42.360 --> 0:03:45.320 analysts were expecting, and the markets maybe been a little 0:03:45.320 --> 0:03:48.720 bit down going into this court into this report, but yeah, 0:03:48.800 --> 0:03:50.520 kind of a cyber relief I think for everybody. 0:03:50.640 --> 0:03:52.520 The company did give an up eat forecast. It is 0:03:52.560 --> 0:03:55.080 a sign that the AI build out is strong. Carolin Heid, 0:03:55.120 --> 0:03:56.720 I want to bring you in, co host of Bloomberg 0:03:56.760 --> 0:04:01.920 Technology on Bloomberg TV Weekday at eleven am Wall Street time. 0:04:03.160 --> 0:04:05.400 I do want to ask about is there is the 0:04:05.440 --> 0:04:08.320 sigh of relief sort of coming too early? Did we 0:04:08.400 --> 0:04:11.160 get any answers about the impact of deep seek here. 0:04:11.280 --> 0:04:12.880 Well, I think there's flies in the ointment. 0:04:13.320 --> 0:04:14.720 I don't know how much you've talked about them, but I 0:04:14.720 --> 0:04:17.880 think the gross margin is clearly a sign that we're 0:04:17.920 --> 0:04:20.440 having perhaps a slight air pocket that many had talked 0:04:20.440 --> 0:04:23.680 about is where you flip from Hopper to Blackwell. And indeed, 0:04:23.720 --> 0:04:26.359 how much expense is spent on making sure that Blackwell 0:04:26.440 --> 0:04:27.960 is easy to use and get up and running as 0:04:28.000 --> 0:04:28.760 quickly as possible. 0:04:28.760 --> 0:04:30.919 That could be an issue. And you have also got. 0:04:30.720 --> 0:04:33.280 This potential guide of slightly weaker on the revenue for 0:04:33.320 --> 0:04:36.760 the fiscal first quarter plus or minus forty three billion. 0:04:37.160 --> 0:04:38.920 If they come into the low end, that does indeed 0:04:39.080 --> 0:04:42.160 just slightly lose out. But I think we are trying 0:04:42.200 --> 0:04:46.800 to hear ultimately whether that insane demand is now an 0:04:46.839 --> 0:04:49.800 amazing amount of demand and does that basically put off 0:04:49.800 --> 0:04:52.479 any concerns that we had about whether or not more 0:04:52.520 --> 0:04:54.760 can be done with less in the future. And I 0:04:54.760 --> 0:04:57.440 think we'll hear on the call, particularly from Jensen, as 0:04:57.480 --> 0:05:01.039 to ultimately whether inferenced demand is really there for his 0:05:01.200 --> 0:05:02.400 very expensive chick tyrol. 0:05:02.400 --> 0:05:04.159 It is like reading a statement from the Federal Reserve 0:05:04.279 --> 0:05:05.479 insane versus amazing. 0:05:06.520 --> 0:05:08.000 This is what I was going to say ask you, 0:05:08.160 --> 0:05:10.320 is that you know you follow this company so closely, 0:05:10.480 --> 0:05:12.800 and it's how do you read the tea leaves because 0:05:12.800 --> 0:05:16.320 he does tend to be very enthusiastic and maybe changes 0:05:16.360 --> 0:05:18.520 the word and maybe there's a nuance to that change, 0:05:18.520 --> 0:05:21.320 But how do you kind of read through what he says, 0:05:21.320 --> 0:05:22.760 because we know he's going to be a beat. 0:05:23.000 --> 0:05:24.960 I think he's going to be saying, look, trust us 0:05:25.000 --> 0:05:26.919 with this R and D, trust us on a slightly 0:05:27.080 --> 0:05:30.240 lower margin because we have got quote the fastest product 0:05:30.320 --> 0:05:31.279 ramp in our countries. 0:05:31.320 --> 0:05:32.480 His company's history. 0:05:32.600 --> 0:05:34.800 This is a company that suddenly become like Apple and 0:05:34.839 --> 0:05:36.440 giving us a new chip. 0:05:36.240 --> 0:05:38.599 Every year that ever used to happen before. 0:05:38.760 --> 0:05:42.760 There's an extraordinary pace which they're trying to innovate that 0:05:42.839 --> 0:05:45.599 costs money. And indeed, of course you have to be 0:05:45.680 --> 0:05:48.239 patient that sometimes there's going to be a slight friction 0:05:48.320 --> 0:05:51.200 when you come to actually putting these Blackwell supercomputers into 0:05:51.240 --> 0:05:51.520 the mix. 0:05:51.839 --> 0:05:54.640 She shares, we're as high as four percent higher in 0:05:54.680 --> 0:05:58.640 the after hours, now down about four tenths of one percent. Yeah, 0:05:58.720 --> 0:06:01.320 not so much, like though investors are really trying to 0:06:01.320 --> 0:06:04.479 make sense of these numbers. Yeah, it's not going and 0:06:04.520 --> 0:06:05.960 it's not moving in a clear direction. 0:06:06.440 --> 0:06:09.120 I think the clear direction is still always growth, isn't it? 0:06:09.200 --> 0:06:12.400 And the idea that sure and one hundred percent increase 0:06:12.440 --> 0:06:14.640 in revenue that got for their fiscal twenty twenty five, 0:06:14.680 --> 0:06:17.240 which is the one that's just gone, and the one 0:06:17.600 --> 0:06:20.400 prior to that, for two straight years we've had a 0:06:20.440 --> 0:06:21.560 doubling of revenue. 0:06:21.680 --> 0:06:23.080 We're going to have to start to get used. 0:06:23.000 --> 0:06:25.200 To a fifty percent increase in revenue and then a 0:06:25.240 --> 0:06:28.440 twenty three percent growth in revenue. The key question I 0:06:28.480 --> 0:06:31.000 think there is still going to be what about the 0:06:31.080 --> 0:06:34.840 cheap alternatives? What about Trainium too that Amazon has that 0:06:34.960 --> 0:06:37.240 is much cheaper, What about some of these startups that. 0:06:37.240 --> 0:06:40.920 You're seeing, does Tridium too, does that compare with Hopper? 0:06:40.960 --> 0:06:41.960 Does it compare with black? 0:06:42.000 --> 0:06:44.599 Well? About inference much more as well? 0:06:44.680 --> 0:06:46.600 And so this is when they're trying to say, hey, 0:06:46.960 --> 0:06:49.599 you've got your beautiful large language model, come and now 0:06:49.839 --> 0:06:51.800 use it with us, and we will save you so 0:06:51.960 --> 0:06:53.440 much more money about thirty percent. 0:06:53.640 --> 0:06:56.600 So after you've used Nvidia to train it, yes, we 0:06:56.640 --> 0:06:59.120 will process those requests for you exactly. 0:06:59.200 --> 0:07:00.600 And I think that's where you're to start to get 0:07:00.600 --> 0:07:00.880 the edge. 0:07:00.880 --> 0:07:02.839 And in fact, there was an interesting interview done with 0:07:02.880 --> 0:07:06.080 a rival with our CEO of course knows data centers 0:07:06.080 --> 0:07:09.559 and mobiles deeply, and his own design work of course 0:07:09.640 --> 0:07:11.920 helping a lot of those out there, including in Video, 0:07:11.960 --> 0:07:14.280 but he himself saying, look, this is where the opening 0:07:14.520 --> 0:07:15.640 is for competitors. 0:07:15.680 --> 0:07:18.400 It's the inference stage. No one can really compete with the. 0:07:18.360 --> 0:07:20.400 GPU power that in Video has when it comes to 0:07:20.400 --> 0:07:22.760 training these models. But the next layer, when we actually 0:07:22.800 --> 0:07:25.480 get it in and use these incredibly powerful things, that's 0:07:25.520 --> 0:07:26.400 where the edge can come. 0:07:27.040 --> 0:07:30.240 I mean, I don't know. In terms of market expectations, 0:07:30.720 --> 0:07:31.520 not bad. 0:07:31.520 --> 0:07:34.440 Yeah, signaling life kind of everyone was on the money, 0:07:34.920 --> 0:07:37.560 alas did their job well. Jensen's done his job well 0:07:37.600 --> 0:07:39.760 in terms of guiding. But had we got used to 0:07:40.240 --> 0:07:42.360 beaten raises. But look look back to when we were 0:07:42.360 --> 0:07:44.440 sit here together with Meta and Meta's numbers were a 0:07:44.560 --> 0:07:46.960 really late be in miss the forecast, and then Zuck 0:07:47.000 --> 0:07:49.440 gone on and just talked really optimistically and we all 0:07:49.440 --> 0:07:49.800 bought it. 0:07:50.000 --> 0:07:52.400 Yeah that's a good point. Yeah, wait to see what 0:07:52.400 --> 0:07:54.680 happens in the cash session. Wait to hear from Jensen 0:07:54.720 --> 0:07:57.040 a little later. We can only take the moves as 0:07:57.040 --> 0:07:58.480 soon as the numbers come out with a grain of 0:07:58.480 --> 0:07:59.200 solid red brown. 0:07:59.400 --> 0:08:02.440 I put deep seek in the search and nothing, nothing's 0:08:02.440 --> 0:08:02.800 coming up. 0:08:03.320 --> 0:08:05.240 Let's just say it'll likely be asked about by at 0:08:05.280 --> 0:08:06.800 least one analyst tonight. 0:08:07.160 --> 0:08:07.560 Red round. 0:08:07.560 --> 0:08:08.680 I want to bring you in here on sort of 0:08:08.720 --> 0:08:12.240 the broader market reaction to this. There was this idea 0:08:12.360 --> 0:08:14.680 that given the segiment we've seen in the last five 0:08:14.800 --> 0:08:17.600 six days, uh in the broader market, the cell off 0:08:17.600 --> 0:08:20.360 that we've seen in the megacap tech companies this year 0:08:20.400 --> 0:08:22.360 and in the last few days, this would sort of 0:08:22.440 --> 0:08:23.840 change the tone, does it. 0:08:24.200 --> 0:08:25.840 I mean, it doesn't really seem like the tone has 0:08:25.880 --> 0:08:28.160 changed at all. We're looking at the numbers is flat right, 0:08:28.800 --> 0:08:31.200 kind of across the board, with everybody not really reacting. 0:08:32.320 --> 0:08:34.600 You know, I think a lot of the narrative kind 0:08:34.640 --> 0:08:36.720 of has been said. You know, the spending is still there. 0:08:36.800 --> 0:08:38.600 I think if we read all the analysts reports heading 0:08:38.640 --> 0:08:41.200 into this, that was the one data point that everyone 0:08:41.240 --> 0:08:43.040 was hanging on to that there hasn't been anyone coming 0:08:43.040 --> 0:08:44.720 out and saying, you know, we're going to be adjusting 0:08:44.720 --> 0:08:47.680 our AI spending. They've stuck to you know, those massive 0:08:47.679 --> 0:08:51.000 capex numbers that we saw throughout the quarter, so I think, yeah, 0:08:51.040 --> 0:08:53.160 maybe a little bit of a wait and see still unfortunately, 0:08:53.920 --> 0:08:55.160 you know, for the time being. 0:08:55.280 --> 0:08:57.199 Well, and they did note I'm looking at our live 0:08:57.240 --> 0:08:59.760 blog noted margins are being squeezed by the cost of 0:08:59.760 --> 0:09:02.480 bringing new, more complicated products to market, and it sounds 0:09:02.520 --> 0:09:02.880 like what. 0:09:03.120 --> 0:09:06.160 Well, Carolyn explained that means because you know Kunjohn also 0:09:06.200 --> 0:09:08.160 mentioned this idea of an air pocket or the transition 0:09:08.320 --> 0:09:12.040 from Harper to Blackwell. What exactly happens behind the scenes 0:09:12.040 --> 0:09:13.240 here and why is it so expensive? 0:09:13.400 --> 0:09:15.400 Well, I think you've just ultimately got to make sure 0:09:15.440 --> 0:09:18.600 that it works hardware seamlessly with the software that you've 0:09:18.640 --> 0:09:22.400 got going, and ultimately that becomes more a powerful piece 0:09:22.400 --> 0:09:25.480 of equipment that is able to use your data and 0:09:25.520 --> 0:09:27.160 do what you want with it. I think when it 0:09:27.160 --> 0:09:30.400 came to Hopper, there was that was working beautifully for many, 0:09:30.400 --> 0:09:32.360 but everyone got deeply excited about the fact that there 0:09:32.400 --> 0:09:35.120 was the latest greatest innovation suddenly coming out and sort 0:09:35.120 --> 0:09:38.400 of decided to go long on Blackwell instead. We actually 0:09:38.440 --> 0:09:41.720 saw a repositioning of that by many customers when they realized, oh, 0:09:41.720 --> 0:09:44.440 actually this is going to take work for everyone to 0:09:44.559 --> 0:09:46.160 ensure to pull this in, to get this into my 0:09:46.240 --> 0:09:47.960 data centers, and to ultimately make sure that. 0:09:48.000 --> 0:09:50.280 It's doing what it wants to do with the most efficiency. 0:09:50.480 --> 0:09:53.600 Heating problems for example, just some of the technicalities that 0:09:53.640 --> 0:09:55.880 they hadn't been able to stress test at this rapid 0:09:55.920 --> 0:09:56.600 rate of innovation. 0:09:57.160 --> 0:10:00.000 All right, just to rehash, Nvidier giving a bullish revenue 0:10:00.000 --> 0:10:02.960 forecast for the current quarter, reassuring investors that spending on 0:10:03.000 --> 0:10:06.240 AI computing remains strung. Just reading some of the reporting 0:10:06.280 --> 0:10:08.840 out by our Oni and King, who follows this sector, 0:10:08.920 --> 0:10:11.200 sales will be about forty three billion in the fiscal 0:10:11.240 --> 0:10:14.320 first quarter, which runs through April. Analystaid An estimated about 0:10:14.360 --> 0:10:17.040 forty two point three billion on average, with some projections 0:10:17.440 --> 0:10:19.680 ranging as high as forty eight billion. So there were 0:10:19.760 --> 0:10:24.360 some outliers and some very high expectations. Carolyn, how are 0:10:24.400 --> 0:10:26.600 you thinking about this story? Obviously we want to see 0:10:26.920 --> 0:10:29.400 what more commentary we get on the call the follow through. 0:10:29.440 --> 0:10:31.880 In terms of the trade, like I'm looking at NASDAQ 0:10:31.880 --> 0:10:36.480 one hundred mini futures, they're relatively little change, maybe up 0:10:36.480 --> 0:10:40.240 about one tenth of one percent. S and P five 0:10:40.280 --> 0:10:43.240 hundred mini futures, they are just a hair lower, pretty flat, 0:10:43.280 --> 0:10:46.719 So it's really very calm. It feels like you're in 0:10:46.760 --> 0:10:47.520 the aftermarket. 0:10:47.640 --> 0:10:49.480 And maybe we just sort of braced ourselves because the 0:10:49.559 --> 0:10:52.839 volatility they experienced yesterday, and then say we got used 0:10:52.880 --> 0:10:55.560 to the fact that the market is trying to find 0:10:55.559 --> 0:10:58.400 its direction when it comes to what is the question 0:10:58.480 --> 0:11:00.280 of capacity, then we've got to go back to what 0:11:00.320 --> 0:11:01.679 happened in the beginning of the week, and we're all 0:11:01.720 --> 0:11:05.119 asking questions about TD Cowan's note over the weekend about Microsoft. 0:11:05.640 --> 0:11:08.280 I'm sure that it will be questions put to Jensen 0:11:08.360 --> 0:11:12.360 about ultimately is there a capacity over stretch or underreach 0:11:12.400 --> 0:11:14.360 at the moment, and what about these so called scaling 0:11:14.440 --> 0:11:16.720 rules that he's trying to combat in every feat, whether 0:11:16.800 --> 0:11:18.760 or not it's about the scaling rules of how much 0:11:18.760 --> 0:11:21.199 better llms can become with the amount of data on 0:11:21.280 --> 0:11:23.920 compute you throw at it, but ultimately also the scaling 0:11:24.000 --> 0:11:26.600 rules of China or whether that becomes something you can 0:11:26.679 --> 0:11:28.720 still scale into in any way with an H twenty. 0:11:28.880 --> 0:11:30.320 I'm glad he brought that up because one of the 0:11:30.360 --> 0:11:33.320 questions that kun John mentioned was getting more detail on 0:11:33.400 --> 0:11:36.679 views of more US restrictions, specifically with regard to China. 0:11:36.760 --> 0:11:40.240 Read I know you've been doing some searching within the 0:11:40.360 --> 0:11:43.280 press release. Are you seeing anything in there about China sales, 0:11:43.320 --> 0:11:45.000 about a commentary about China. 0:11:44.800 --> 0:11:46.920 No mention about China. So then Kunjohn is correct that 0:11:46.920 --> 0:11:48.560 we that's definitely going to be a top of mind 0:11:49.200 --> 0:11:53.040 topic for analysts on the call. You know right now 0:11:53.080 --> 0:11:54.960 there's a lot of question marks looming around what they 0:11:55.000 --> 0:11:56.720 can do there. It gets they get around twelve percent 0:11:56.760 --> 0:11:58.920 of their sales over the past few years, most of 0:11:58.920 --> 0:12:01.439 that coming across from game being automotive end data centers. 0:12:01.480 --> 0:12:03.600 So I think there will be questions about, you know, 0:12:03.720 --> 0:12:06.760 what sort of contingencies maybe they have in place if 0:12:07.080 --> 0:12:09.800 any sort of rediction restrictions do come down. 0:12:10.000 --> 0:12:12.240 Well, and Caroline, I mean jensenmong has made his way 0:12:12.280 --> 0:12:15.120 to the way that's read. He said, conversations with President 0:12:15.160 --> 0:12:17.439 Trump about these issues and his concerns. 0:12:17.679 --> 0:12:21.199 Yeah, certainly, and that had been pre premeditated. He'd been 0:12:21.200 --> 0:12:23.800 going to go and see President Trump. We know that 0:12:23.840 --> 0:12:25.960 he'd wanted to. When our own ed Ludlow sat down 0:12:25.960 --> 0:12:28.400 with him previously, he said, I'm ready willing and able 0:12:28.440 --> 0:12:30.400 to go and meet President Trump, and certainly they wanted 0:12:30.440 --> 0:12:32.280 me to go. She because remembers the end of the administration, 0:12:32.360 --> 0:12:35.560 the Biden administration that's suddenly added all these curtailments to 0:12:35.559 --> 0:12:39.400 how much they can place chips in other geographies, not 0:12:39.440 --> 0:12:41.360 just China, because there's this worry that you sell into 0:12:41.360 --> 0:12:44.080 a different nation to Singapore, and actually that's suddenly how 0:12:44.080 --> 0:12:47.199 it's getting dispersed into China as well. So he's got 0:12:47.240 --> 0:12:50.319 a lot to be debating. I do note in I mean, 0:12:50.320 --> 0:12:52.079 I hate to be a cynic when you're looking at 0:12:52.120 --> 0:12:56.000 these press releases, but he really only mentions US companies 0:12:56.040 --> 0:12:57.920 at the beginning of how much that they're working with 0:12:58.000 --> 0:13:01.920 the HWS is the Cisco's the verizons. I mean, it 0:13:02.000 --> 0:13:04.600 really is a focus in on us. 0:13:04.760 --> 0:13:06.920 It's just this week things I think now increasing a 0:13:07.000 --> 0:13:08.439 partnership right with Cisco. 0:13:08.280 --> 0:13:10.679 Exactly to get the networking better again, just get these 0:13:10.720 --> 0:13:12.240 things into people's hands more efficiently. 0:13:12.400 --> 0:13:15.480 We've talked about hop Hopper, We've talked a lot about Blackwell. 0:13:15.760 --> 0:13:17.360 Is it too early for us to be talking about 0:13:17.360 --> 0:13:19.400 Reuben I don't. 0:13:19.440 --> 0:13:23.640 I mean, what in the next innovation, This is the 0:13:23.720 --> 0:13:26.240 next year, this is the next I don't think Jadson 0:13:26.320 --> 0:13:27.960 is ever too early to talk about what he's got 0:13:28.040 --> 0:13:28.720 up next to sleep. 0:13:28.880 --> 0:13:31.200 So that's that's the question. I mean, our investors looking 0:13:31.320 --> 0:13:33.240 for you know, this is this is the This is 0:13:33.280 --> 0:13:36.040 what happens with compute, right, You build something and then 0:13:36.040 --> 0:13:37.920 you build something better, and you build something better. That 0:13:38.040 --> 0:13:40.280 just happens over and over again, and it costs money. 0:13:40.400 --> 0:13:43.480 It costs money. So how important is you know the 0:13:43.559 --> 0:13:47.240 next generation what comes after blackwell apparently code named Rubin. 0:13:47.440 --> 0:13:50.000 How important is that to where we are in the 0:13:50.040 --> 0:13:51.000 timeline of Nvidia. 0:13:52.280 --> 0:13:56.040 I think it's important to show that they are continuing 0:13:56.240 --> 0:13:59.719 to just innovate ahead of any of the competition. That 0:14:00.120 --> 0:14:02.400 the more it's more Lisa Sue who pays attention to 0:14:02.480 --> 0:14:06.120 Ruben and things up like, yeah, how am I managing 0:14:06.200 --> 0:14:07.959 to get any sort of in here in terms of 0:14:08.080 --> 0:14:10.559 owning the data center space when it comes to just 0:14:10.679 --> 0:14:14.000 the ability that Jensen has had to see around corners, 0:14:14.000 --> 0:14:15.679 when it comes to the use and application of his 0:14:15.800 --> 0:14:17.360 GPUs for gaming. 0:14:17.080 --> 0:14:19.200 For crypto and then for generative AI. 0:14:19.760 --> 0:14:22.600 But maybe Lisa Su starts to play the tack of okay, 0:14:22.680 --> 0:14:25.080 but it's me with cheaper price points, and I wonder 0:14:25.080 --> 0:14:27.400 whether the margin starts to creep into that narrative as 0:14:27.440 --> 0:14:30.480 well as maybe yes, it's a deeply expensive piece of 0:14:30.560 --> 0:14:32.760 kit with everyone offering cheaper alternatives. 0:14:32.840 --> 0:14:33.200 What do you do? 0:14:33.280 --> 0:14:33.960 How do you respond? 0:14:34.040 --> 0:14:36.000 All right, folks, we're talking with Caroline Hyde, co host 0:14:36.000 --> 0:14:39.200 of Bloomberg Technology on Bloomberg TV at eleven am Wall 0:14:39.280 --> 0:14:42.720 Street Time. Let's bring into the conversation Mandy Seeing walking 0:14:42.800 --> 0:14:46.200 into our studio Bloomberg Intelligence, Senior Tech Industry Analysis. He 0:14:46.240 --> 0:14:48.000 walked in because we asked him to and we were 0:14:48.760 --> 0:14:51.400 hoping you would. Hey, and video shares have been bouncing around. 0:14:51.920 --> 0:14:54.960 Should any of us who are thinking about the AI industry? 0:14:55.080 --> 0:14:56.400 Is this a dose of optimism? 0:14:57.440 --> 0:15:00.400 It is? But when it comes to Nvidia specific I 0:15:00.480 --> 0:15:04.160 would say, you know, the fact that networking revenue declines 0:15:04.240 --> 0:15:07.720 sequentially for the second quarter in a row again in 0:15:07.840 --> 0:15:11.240 Vidio is very high bar. And you know, given their 0:15:11.280 --> 0:15:14.800 bed and raises, everyone expects them to do, you know, 0:15:14.960 --> 0:15:18.000 well on every line. So networking going down, to me, 0:15:18.120 --> 0:15:21.680 that's a sticking point, along with gross margins going down 0:15:22.200 --> 0:15:25.240 and just the sequential increases that we have seen in 0:15:25.320 --> 0:15:29.280 the past quarters. It looks like it's tapering again. It's 0:15:29.480 --> 0:15:32.560 double digit. But when it comes to the overall growth rate, 0:15:32.680 --> 0:15:35.440 it feels like it's decelerating. And the last data point 0:15:35.480 --> 0:15:40.520 I would offer is the cloud exposure, the hyperscale cloud exposure. 0:15:40.800 --> 0:15:44.600 It was forty five percent before now it's fifty percent Now. 0:15:44.760 --> 0:15:47.880 What it tells you is they kind of we are exposed. 0:15:48.480 --> 0:15:51.520 And so the fact that you know networking is going down, 0:15:51.640 --> 0:15:56.960 it's a sign that you know, training versus inferencing component 0:15:57.560 --> 0:16:01.200 that seems to be really in question here because networking 0:16:01.280 --> 0:16:06.040 does well. If training is growing, if networking is going down, 0:16:06.240 --> 0:16:09.560 that's a sign that they may be facing some sort 0:16:09.640 --> 0:16:13.960 of challenges or starting to get more questions around the 0:16:14.000 --> 0:16:16.200 scale of training. And I think that's a big question, 0:16:16.320 --> 0:16:16.840 So go. 0:16:16.920 --> 0:16:19.240 Deeper on that with training versus inference. We're talking a 0:16:19.240 --> 0:16:22.920 little bit about it with Caroline context Amazon and what 0:16:23.000 --> 0:16:26.880 Amazon has when it comes to in France. But what 0:16:27.000 --> 0:16:28.960 do we learn in video within Vidia today. 0:16:29.440 --> 0:16:33.680 Look, if you're setting up a giant cluster like Xai 0:16:33.880 --> 0:16:36.400 has with you know, one hundred and two hundred thousand GPUs, 0:16:36.680 --> 0:16:38.760 you need a lot of networking there. But if you 0:16:38.800 --> 0:16:41.240 are setting up a data center with let's say a thousand, 0:16:41.320 --> 0:16:45.120 two thousand GPUs, you don't need as much of Nvidia 0:16:45.160 --> 0:16:49.040 as networking, and that's where you know, training versus inferencing 0:16:49.120 --> 0:16:53.600 really matters, because if you know, companies are leaning towards 0:16:53.640 --> 0:16:58.840 setting up more data centers with fewer chips. I mean 0:16:58.880 --> 0:17:00.800 when I say, if you are still thousands of chips 0:17:00.840 --> 0:17:03.120 but not to the scale of one hundred thousand, then 0:17:03.280 --> 0:17:05.720 networking won't do as well. And that's the whole debate 0:17:05.760 --> 0:17:08.600 about training versus inferencing. So to my mind, you know, 0:17:08.800 --> 0:17:13.000 deep seek and all the recent developments from hyperscalers suggests 0:17:13.040 --> 0:17:17.040 that you may not need giant clusters and that is 0:17:17.520 --> 0:17:21.240 I think going to affect Nvidia going forward. So when 0:17:21.280 --> 0:17:22.920 I look at their guide, when I look at the 0:17:23.040 --> 0:17:26.760 fact that you know, networking declined sequentially, again, it's a 0:17:26.880 --> 0:17:30.760 great print, but to me, the path seems to be 0:17:31.000 --> 0:17:33.560 that it will slow down on the networking side, and 0:17:33.600 --> 0:17:36.080 then you have to ask yourself where will the upside 0:17:36.160 --> 0:17:38.919 come from. Maybe on the call they expand on starlink 0:17:39.119 --> 0:17:42.359 and you know, the sovereigns that is a big pocket. 0:17:42.440 --> 0:17:45.480 If that pans out in terms of training, like setting 0:17:45.600 --> 0:17:49.280 up big clusters that Sam Altman has talked about, you know, 0:17:49.400 --> 0:17:52.160 with projects started, then you will start to see networking 0:17:52.200 --> 0:17:55.000 bounce but for me, there is a clear connection between 0:17:55.040 --> 0:17:58.000 networking and size of cluster and also software. There was 0:17:58.119 --> 0:18:01.200 nothing in the print around software, even though it's you know, 0:18:01.320 --> 0:18:04.080