WEBVTT - All Eyes on Nvidia Ahead of Earnings 0:00:02.560 --> 0:00:13.480 Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is alive 0:00:13.560 --> 0:00:17.360 from coast to coast with Caroline Hyde in New York 0:00:17.640 --> 0:00:19.640 and Va Loow in San Francisco. 0:00:22.880 --> 0:00:24.480 This is Bloomberg Tech coming up. 0:00:24.520 --> 0:00:28.040 All eyes on Nvidia earnings after the closing bell, Investors 0:00:28.240 --> 0:00:30.920 expecting to learn more about where those billions of dollars 0:00:30.920 --> 0:00:32.800 on AI spending are actually going. 0:00:32.880 --> 0:00:35.560 Plus in Video CEO Jensen Wang and Tesla CEO Elil 0:00:35.600 --> 0:00:38.480 Musk are speaking right now at the US Saudi Investment Forum. 0:00:38.760 --> 0:00:40.600 We'll bring you the latest, and. 0:00:40.680 --> 0:00:44.320 Brookfield Asset Management targets ten billion dollars of fund commitments 0:00:44.360 --> 0:00:47.640 for a global AI infrastructure program. 0:00:46.960 --> 0:00:50.360 In partnership with you guessed it in Video and. 0:00:50.360 --> 0:00:51.879 In Vidia dictates trade. 0:00:51.960 --> 0:00:54.440 Right now, we are seeing signs of stability in the 0:00:54.480 --> 0:00:57.480 Nasdaq one hundred. More broadly, I remind you that almost 0:00:57.480 --> 0:00:59.520 two trillion dollars have been wiped off of this benchmarks. 0:00:59.560 --> 0:01:01.840 It's the end of October in large part because in 0:01:01.920 --> 0:01:03.800 Video has been down, but all the key mag seven 0:01:03.880 --> 0:01:06.360 names have been under pressure. But today's some reprieve and 0:01:06.440 --> 0:01:09.000 video in the point's perspective helping within as that one 0:01:09.080 --> 0:01:11.399 hundred crypto though still in the EI of the storm 0:01:11.480 --> 0:01:13.679 or by two point eight percent, that anxiety driving it 0:01:13.720 --> 0:01:17.000 below ninety thousand dollars, so still some risk A version. 0:01:16.840 --> 0:01:19.720 Ed okay, in Vidia is up more than two and 0:01:19.720 --> 0:01:22.480 a half percent, but off its session high. It is 0:01:22.480 --> 0:01:25.160 a stock that's up almost forty percent year to date, 0:01:25.640 --> 0:01:28.320 outperforming double the performance of what we've seen of the 0:01:28.400 --> 0:01:31.720 NAZAQ one hundred in S and P five hundred. The 0:01:31.920 --> 0:01:36.600 expectation is revenue growth above fifty percent, net income growth 0:01:36.640 --> 0:01:39.520 above fifty percent. But all that matters is what CEO 0:01:39.640 --> 0:01:43.840 Jensmong tells us about the future. Let's get with Bloomberg'ssey 0:01:43.880 --> 0:01:46.479 and King, who leads our semiconductive coverage. I mean, that's 0:01:46.520 --> 0:01:50.160 what it comes down to. Either they will beat consensus 0:01:50.240 --> 0:01:53.880 or they won't. But expectations are really high for this quarter, 0:01:54.280 --> 0:01:57.480 as is the skepticism of what's happening bigger picture with 0:01:57.600 --> 0:01:59.800 data center instructure. Give us the things we need to 0:01:59.840 --> 0:02:02.040 look out for, and what's in your preview of this 0:02:02.480 --> 0:02:03.320 company's earnings? 0:02:03.440 --> 0:02:06.680 Yeah, I mean the numbers speak for themselves, right. We're 0:02:07.040 --> 0:02:11.000 looking for a prediction in the sixty billion range for revenue, 0:02:11.400 --> 0:02:14.200 and just to give that context, that's ten x where 0:02:14.200 --> 0:02:18.000 we were three years ago, ten x Okay on profit 0:02:18.040 --> 0:02:19.880 for this year, we're going to be looking at one 0:02:19.960 --> 0:02:23.840 hundred billion dollars of net income. That's more than Intel 0:02:23.960 --> 0:02:26.320 and AMD get in revenue combined. So the numbers have 0:02:26.360 --> 0:02:30.639 come off the charts. The key here is we all 0:02:30.639 --> 0:02:31.919 know the numbers are going to be good. We know 0:02:32.000 --> 0:02:34.120 the forecast is going to be But the key is, well, 0:02:34.160 --> 0:02:37.480 do we really believe the basis for those numbers? And 0:02:37.520 --> 0:02:39.800 that's going to be the key question he's going to face. 0:02:40.560 --> 0:02:43.840 So in what can his response be that's more than 0:02:43.840 --> 0:02:46.359 what are you already signals in GtC that he has 0:02:46.800 --> 0:02:49.959 line of sight on half a trillion dollars worth of 0:02:50.080 --> 0:02:53.200 orders not you black Bell, but Roobin into twenty twenty six? 0:02:53.440 --> 0:02:56.400 How much more can he signal that they will remain 0:02:56.480 --> 0:02:58.440 integral to inference as well as training. 0:03:00.080 --> 0:03:03.920 Absolutely right, he's essentially played all his cards in that respect. 0:03:03.960 --> 0:03:07.520 What's going to happen will be that the investment community 0:03:07.560 --> 0:03:11.840 are getting their first chance to sort of pull that apart, 0:03:11.919 --> 0:03:14.360 to ask him about the details, to ask him about 0:03:14.560 --> 0:03:17.119 the new products, to ask about the margins, to ask 0:03:17.160 --> 0:03:20.359 about when exactly these sales will kick in and we haven't. 0:03:20.560 --> 0:03:22.639 They haven't really had that chance. So that's what they'll 0:03:22.680 --> 0:03:25.639 dig into today. And his response is how precise he 0:03:25.880 --> 0:03:28.160 is will condition how they feel. 0:03:28.520 --> 0:03:30.839 There are some real concerns and there are some real 0:03:30.960 --> 0:03:32.960 questions from videos. Some of those will get to pose 0:03:33.000 --> 0:03:35.960 this evening to gentlemen self. There is the idea of 0:03:36.040 --> 0:03:40.920 depreciation on older chips and circular financing that's just not going. 0:03:40.720 --> 0:03:43.840 Away, absolutely not. I mean, we saw a deal announced 0:03:43.880 --> 0:03:48.600 with Anthropic yesterday, big commitment to use a lot more 0:03:48.720 --> 0:03:51.480 of in Video's chips. But guess what Invidea is putting 0:03:51.520 --> 0:03:54.440 ten billion dollars to work in that company over time. 0:03:54.520 --> 0:03:58.040 So yes, that concern is absolutely not going away. And 0:03:58.080 --> 0:04:00.760 if anything is going to accelerate, who we get a 0:04:00.840 --> 0:04:02.080 clear outcome. 0:04:02.160 --> 0:04:05.560 Bloomg's inking, who will be across all those earnings after 0:04:05.600 --> 0:04:08.240 the bell alongside ed We so appreciate it. Meanwhile, look, 0:04:08.320 --> 0:04:11.160 we know Wall Street is eagerly awaiting in Vidia's results 0:04:11.160 --> 0:04:14.040 for a clearer picture on AI spending because it impacts 0:04:14.280 --> 0:04:17.400 the whole rest of the market. Bloomberg's US Equities report 0:04:17.400 --> 0:04:19.839 a common rhyining keys here to just bring us the context. 0:04:20.080 --> 0:04:22.080 We know the question is gonna be asked of Nvidia, 0:04:22.120 --> 0:04:24.359 But what does it signal about the commitment of the 0:04:24.400 --> 0:04:27.359 Magnificent seven of the key hyperscalers and more broadly the 0:04:27.360 --> 0:04:28.400 rest of the AI trade. 0:04:29.360 --> 0:04:31.719 Yeah, so this is a huge moment for the AI trade. 0:04:31.760 --> 0:04:33.440 I think a lot of people are really looking to 0:04:33.560 --> 0:04:37.320 Invidia and the stocks reaction to sort of decide the 0:04:37.360 --> 0:04:41.560 next direction of where the market, the entire market is 0:04:41.560 --> 0:04:43.919 going to go. You know, in Nvidia is the largest 0:04:43.920 --> 0:04:46.720 waiting in the S and P five hundred and the biggest. 0:04:46.400 --> 0:04:47.039 Name in AI. 0:04:47.120 --> 0:04:49.960 And what we also know about in Vidia is that 0:04:50.160 --> 0:04:53.520 its biggest four clients are some of the other MAGS 0:04:53.560 --> 0:04:56.359 seven members. So we're going to see where they're spending 0:04:56.440 --> 0:04:58.160 is flowing if it's still flowing to. 0:04:58.200 --> 0:04:59.160 Nvidia how much. 0:05:00.680 --> 0:05:03.120 When I was reading your your piece which you co 0:05:03.200 --> 0:05:06.359 wrote with Rhyan Vla Seleca, the data is really interesting. 0:05:06.400 --> 0:05:07.840 It's gonna be a big part of the Bluma tech 0:05:07.880 --> 0:05:09.680 audience that don't know some of that. So you just 0:05:09.720 --> 0:05:12.719 talked about waiting right in video is about eight percent 0:05:12.720 --> 0:05:14.960 of the S and P five hundred. That's a factor 0:05:15.279 --> 0:05:18.159 you also looked at in Nvidia's multiples relative to the 0:05:18.160 --> 0:05:20.640 index level the Nasdaq one hundred, for example. What are 0:05:20.680 --> 0:05:22.760 the other key data points that have us on such 0:05:23.240 --> 0:05:25.760 edge ahead of the earnings report later this evening. 0:05:26.440 --> 0:05:28.280 Well, you know, as Ian said, I think a lot 0:05:28.279 --> 0:05:31.080 of people are really looking at the forward guidance. You know, 0:05:31.200 --> 0:05:34.039 in video is expected to continue to grow revenue, even 0:05:34.080 --> 0:05:37.120 though that growth is expected to slow in the coming years. 0:05:37.360 --> 0:05:39.159 I think the other thing that people are really looking 0:05:39.200 --> 0:05:42.400 for is what Jensen's going to say about what they're 0:05:42.480 --> 0:05:45.159 seeing in the future. Right Also, is Ian pointed out 0:05:45.279 --> 0:05:48.000 the guidance is very important here, and that's I mean, 0:05:48.040 --> 0:05:50.839 probably even more important than the numbers that in Vidia 0:05:51.160 --> 0:05:54.159 actually reports. We want to see the line of sight 0:05:54.360 --> 0:05:57.320 into revenue growth going forward. There are also still some 0:05:57.480 --> 0:06:00.320 questions about China, how much revenue can be expected there 0:06:00.440 --> 0:06:02.760 or not, And so I think the sentiment here, what 0:06:02.920 --> 0:06:06.960 investors take away about their confidence from the report, is 0:06:06.960 --> 0:06:10.080 going to be maybe even more important than the actual numbers. 0:06:10.600 --> 0:06:13.280 Pretty most common Rhino Keey, it's a great story. Thank 0:06:13.320 --> 0:06:15.560 you very much indeed for bringing us the data as well. 0:06:15.600 --> 0:06:18.040 And now we bring you the investment perspective. Martin Norton's 0:06:18.080 --> 0:06:21.280 here with us chief investment strategistic and Power, which administers 0:06:21.320 --> 0:06:24.240 more than one point six trillion dollars in assets, and 0:06:24.240 --> 0:06:25.720 that's about the same amount that's been wiped off on 0:06:25.760 --> 0:06:28.120 nude like one hundred since the end of October. 0:06:28.320 --> 0:06:30.160 Yes, is there room to the. 0:06:30.120 --> 0:06:32.560 Downside here or are you thinking there's some sort of 0:06:32.600 --> 0:06:35.599 relief rally from any data we get from Jensen later? 0:06:35.920 --> 0:06:37.880 Well, I don't think you can ever count out a 0:06:37.960 --> 0:06:40.359 relief rally. But what I come back to when we 0:06:40.440 --> 0:06:43.080 look at the price action that we've seen over the 0:06:43.120 --> 0:06:46.040 past few weeks is that we have taken some froth 0:06:46.120 --> 0:06:49.440 off the top, but we're not looking at really attractive 0:06:49.520 --> 0:06:53.200 valuations at this point. We're still at elevated valuations for 0:06:53.240 --> 0:06:56.240 a lot of these names. And what's interesting, and it 0:06:56.360 --> 0:06:59.520 alluded to this, you know, the questions that people are 0:06:59.600 --> 0:07:02.240 raising the current environment, I don't think they're necessarily going 0:07:02.279 --> 0:07:07.720 to be conclusively answered within Video's report, right, the depreciation question, 0:07:07.880 --> 0:07:10.680 the demand question, you know, extending to the rest of 0:07:10.720 --> 0:07:13.640 the economy. I think those doubts are are with us, 0:07:13.800 --> 0:07:16.480 which would mean potentially more volatility. 0:07:17.040 --> 0:07:19.760 Well, if Jensen can't tell us how long his GPUs 0:07:19.800 --> 0:07:21.640 are going to last and what the depreciation of them are. 0:07:21.800 --> 0:07:23.920 I'm not sure anyone have, but you're so right that 0:07:23.920 --> 0:07:26.720 that has been again an argument about why perhaps. 0:07:26.480 --> 0:07:27.680 Evaluations are flush. 0:07:27.840 --> 0:07:30.200 If you look at a video trenting about thirty times 0:07:30.240 --> 0:07:34.120 future earnings, that's not that elevated. So is it the 0:07:34.120 --> 0:07:37.560 rest of the trade, the palenteers, or perhaps some of 0:07:37.600 --> 0:07:40.640 the energy stocks that have risen to extraordinary degrees. 0:07:40.720 --> 0:07:42.600 Well, I think you raise a good point. It's not 0:07:42.720 --> 0:07:46.520 that elevated, especially if you're focused on that right side 0:07:46.560 --> 0:07:49.920 of the probability distribution, if you're looking at a full 0:07:50.040 --> 0:07:53.360 probability distribution, and you come to the conclusion that I 0:07:53.400 --> 0:07:55.760 don't know if I share, but some folks are raising, hey, 0:07:55.800 --> 0:07:59.360 this isn't as transformative as people have suggested it is. 0:07:59.400 --> 0:08:02.720 We're not going to see every application take off the 0:08:02.760 --> 0:08:05.560 way people suggest it would. Then I think there's room 0:08:05.600 --> 0:08:07.480 for some of these stacks to come down. Now, I'm 0:08:07.480 --> 0:08:09.440 not sure I share that view. I tend to believe 0:08:09.440 --> 0:08:12.000 that the AI supercycle is real, that it is going 0:08:12.080 --> 0:08:14.400 to have a pretty profound impact on the economy, but 0:08:14.480 --> 0:08:16.160 we have to make room for the full range of 0:08:16.160 --> 0:08:17.679 outcomes when we price these things. 0:08:18.960 --> 0:08:21.120 Marta, it's great to have you back on Bloomberg Tech. 0:08:21.480 --> 0:08:24.320 As is quoted in that well read story on the 0:08:24.360 --> 0:08:27.960 Bloomberg term on Bloomberg dot com from one investor, this 0:08:28.040 --> 0:08:31.400 is a quote, so goes Nvidia, so goes the market 0:08:31.600 --> 0:08:35.120 kind of report from your desk, and your perspective is 0:08:35.440 --> 0:08:37.520 that the situation here. 0:08:37.720 --> 0:08:39.640 Well, I think there's no question that this is a 0:08:39.679 --> 0:08:42.959 Capstone report. It's a macro indicator, and I think has 0:08:43.000 --> 0:08:45.760 the potential and you see that in pricing around it, 0:08:45.840 --> 0:08:47.520 right this idea that you could have swings up to 0:08:47.559 --> 0:08:50.120 seven percent either way. I think this is the kind 0:08:50.160 --> 0:08:52.200 of thing that people are really going to key off of. 0:08:52.400 --> 0:08:56.960 My question is whether this is going to answer all 0:08:57.040 --> 0:08:59.880 the AI doubts that are out there or whether it's 0:09:00.120 --> 0:09:02.840 just going to arrest them for now. But we're still 0:09:02.880 --> 0:09:05.000 going to be wrestling with some of these questions going forward. 0:09:05.920 --> 0:09:09.000 Depreciation is the most common concern or question that I've 0:09:09.040 --> 0:09:11.959 received for Gensen so far. You know, get out to 0:09:11.960 --> 0:09:14.720 people and said what would you ask? There are people 0:09:14.760 --> 0:09:17.200 that look at the depreciation issue and say, how can 0:09:17.240 --> 0:09:19.760 I model for that impact on the balance sheet of 0:09:19.800 --> 0:09:23.920 those key customers of Nvidia, But others on social media 0:09:23.960 --> 0:09:26.560 are talking about a different data set which is utilization. 0:09:27.080 --> 0:09:30.120 If those older generation GPUs are running at one hundred percent, 0:09:30.160 --> 0:09:33.400 that's a really good problem to have, right. This is 0:09:33.480 --> 0:09:36.880 kind of very specific, but are those soft data sets 0:09:36.880 --> 0:09:38.760 that your team are looking at to work out what 0:09:38.960 --> 0:09:39.600 neof's going on? 0:09:40.160 --> 0:09:42.560 Well, you know, I think the thing that we're focused 0:09:42.600 --> 0:09:45.640 most on as we turn our gaze to twenty twenty 0:09:45.679 --> 0:09:49.800 six is this question in particular about the capacity bill 0:09:49.880 --> 0:09:52.080 that we expect in twenty twenty six and frankly in 0:09:52.160 --> 0:09:55.079 twenty twenty seven. So our view is that, you know, 0:09:55.160 --> 0:09:58.440 we're going to have these questions to your point about utilization, 0:09:58.679 --> 0:10:02.000 about appreciation, but our view is that those questions are 0:10:02.000 --> 0:10:05.040 going to linger as we build out the capacity for 0:10:05.120 --> 0:10:08.080 AI in twenty twenty six, in twenty twenty seven, because 0:10:08.080 --> 0:10:09.960 there is so much that you have to put in 0:10:10.000 --> 0:10:13.080 place to be able to see the demand come through. 0:10:13.400 --> 0:10:15.840 And so for us, as we go through twenty twenty six, 0:10:15.840 --> 0:10:17.760 we're going to be watching, of course, like everybody else. 0:10:17.960 --> 0:10:20.880 Are we seeing that capex is the conviction still there 0:10:20.920 --> 0:10:24.400 to build out the massive infrastructure that you need for AI, 0:10:24.440 --> 0:10:26.920 And I think that's the key question that we're wrestling with. 0:10:27.400 --> 0:10:30.960 It was interesting that you said you don't align yourself 0:10:31.000 --> 0:10:34.320 with the negativity and actually applications of AI, right, you 0:10:34.360 --> 0:10:36.080 do think the supercycle's real? 0:10:36.600 --> 0:10:39.520 What data set are you looking for that? Because we do. 0:10:39.440 --> 0:10:43.080 Have the MIT pilots aren't working, there's always a counter 0:10:43.160 --> 0:10:46.040 example for every time there's a negative, but Intelligent just 0:10:46.040 --> 0:10:49.040 have some great analysis out showing that actually only ten 0:10:49.080 --> 0:10:51.920 percent of companies at the moment or people surveyed are 0:10:51.920 --> 0:10:54.440 saying that they're using generator of AI for revenue or 0:10:54.440 --> 0:10:55.080 for new product. 0:10:55.200 --> 0:10:58.200 Right, I think there's still an I mean, first of all, 0:10:58.240 --> 0:10:59.959 if you take a look at kind of adoption rates 0:11:00.160 --> 0:11:02.959 for the AI cycle relative to the Internet and things 0:11:03.000 --> 0:11:04.800 like that, people are pointing out the Federal Reserve and 0:11:04.840 --> 0:11:08.880 others that you've seen much faster acceleration of adoption. But 0:11:09.000 --> 0:11:12.280 to your point, we're still very much early days, and 0:11:12.360 --> 0:11:14.440 so I think one of the things that we look 0:11:14.440 --> 0:11:18.040 at is just at the earnings season themselves and looking 0:11:18.080 --> 0:11:20.240 at what companies are saying. And a lot of the 0:11:20.440 --> 0:11:23.760 commentary around AI at this point is still very generic. 0:11:23.840 --> 0:11:25.079 It's not very. 0:11:24.840 --> 0:11:28.520 Specific in terms of how AI is actually transforming those businesses. 0:11:28.520 --> 0:11:31.400 And I think also watching to see how earnings in 0:11:31.440 --> 0:11:33.720 the broader economy are responding, Are you starting to see 0:11:33.720 --> 0:11:36.079 those cost savings come through, which I think is of 0:11:36.120 --> 0:11:38.319 course the first leg, and then ultimately you'd also want 0:11:38.320 --> 0:11:40.880 to see the revenue which we're seeing from the hyperscalers. 0:11:41.040 --> 0:11:44.120 But I think those, you know, that real world application 0:11:44.240 --> 0:11:45.800 is really what we're going to be keying off of it, 0:11:45.800 --> 0:11:48.000 and I think it's going to take time. I think 0:11:48.000 --> 0:11:50.640 you still need to build that capacity to see the 0:11:51.240 --> 0:11:52.040 earnings impact. 0:11:52.559 --> 0:11:55.120 Broadly speaking, Macha, so it's great to have you in 0:11:55.280 --> 0:11:57.920 good thank you for coming to this year. Mantulton close 0:11:57.920 --> 0:12:00.560 side of empower. Meanwhile, in Vidia CEO and Hung it's 0:12:00.600 --> 0:12:02.640 already on stage ahead of his earnings publication. 0:12:02.679 --> 0:12:03.760 Of course, I mean while test to. 0:12:05.280 --> 0:12:07.800 Alongside in DC as part of the US Saudi Investment 0:12:07.840 --> 0:12:08.880 Forum in Washington, d C. 0:12:09.120 --> 0:12:10.000 Take a listen to his. 0:12:10.000 --> 0:12:14.320 Royal Highness announced the AI strategic Framework and partnership. Today 0:12:14.600 --> 0:12:17.840 we're going big with Elon and Jensen. 0:12:17.559 --> 0:12:19.359 So thank you for those opportunities. 0:12:23.880 --> 0:12:27.520 Now they told me I have time for two last questions. 0:12:27.559 --> 0:12:31.120 So last night at the dinner, I got a number 0:12:31.160 --> 0:12:34.880 of questions because it seems that the schedule leaked and 0:12:35.200 --> 0:12:37.839 everybody was giving me hints about the last two questions 0:12:37.920 --> 0:12:40.160 I'm going to do. So the first one was for 0:12:40.240 --> 0:12:44.760 you Elon, and there's a big one for you Jensen, 0:12:44.840 --> 0:12:45.960 So prepare for that one. 0:12:47.240 --> 0:12:49.880 AI in space Is that possible? 0:12:51.800 --> 0:12:56.559 Yes, if civilization continues, which it probably will, then AI 0:12:56.559 --> 0:13:03.840 in space is inevitable. You know, I always have to 0:13:03.880 --> 0:13:08.240 like preface that, you know, we shouldn't take civilization for granted. 0:13:08.360 --> 0:13:11.640 We need to make sure to take care to ensure 0:13:11.640 --> 0:13:15.240 that civilization hasn't an upward arc. I mean, any student 0:13:15.240 --> 0:13:18.200 of history knows that civilization does not always have an 0:13:18.240 --> 0:13:21.520 upward arc, and in fact, civilizations have life life cycles. 0:13:22.000 --> 0:13:24.560 So hopefully we are in a strong upward arc. I 0:13:24.559 --> 0:13:26.719 think we are for now, but we don't want to 0:13:26.760 --> 0:13:31.320 take that for granted or becomplacent. But in order to 0:13:31.840 --> 0:13:34.439 the way to think of AI in space is that 0:13:34.520 --> 0:13:37.520 in order to achieve any meaningful percentage of a Kotdashev 0:13:37.559 --> 0:13:41.319 two scale civilization where you're using even a millionth of 0:13:41.480 --> 0:13:46.120 a millionth of the Sun's energy, you must have solar 0:13:46.200 --> 0:13:52.720 powered AI satellites in deep space, so that once you realize, 0:13:52.720 --> 0:13:55.000 like once you think in terms of a Cottashev two 0:13:55.040 --> 0:13:59.080 scale civilization, which is what percentage of the Sun's energy 0:13:59.360 --> 0:14:03.200 are you turning into useful work? Then you then it 0:14:03.240 --> 0:14:08.080 becomes obvious that space is overwhelmingly what matters. 0:14:08.640 --> 0:14:09.280 Overwhelmingly. 0:14:09.559 --> 0:14:12.240 So the Sun only receives one roughly one two billionth 0:14:12.760 --> 0:14:17.199 of the Earth only receives roughly one two billionth of 0:14:17.240 --> 0:14:21.120 the Sun's energy. So if you want to have something 0:14:21.120 --> 0:14:24.760 that is, say a million times more energy than Earth 0:14:24.800 --> 0:14:27.720 could possibly produce, you must. 0:14:27.520 --> 0:14:31.040 Go into space. It's and so. 0:14:33.720 --> 0:14:35.200 This is where it's kind of handy to have a 0:14:35.240 --> 0:14:36.800 space company, I guess. 0:14:38.560 --> 0:14:42.240 Sell the books cold chips in space too, Yes, easier 0:14:42.240 --> 0:14:43.320 to cool chips in space. 0:14:43.720 --> 0:14:45.640 Yes, there's definitely no water in space, so you're gonna 0:14:45.680 --> 0:14:48.080 have to do something that doesn't involve water. 0:14:49.320 --> 0:14:51.360 Well, it's just got to radiate, that's right. 0:14:52.760 --> 0:14:58.119 So my estimate is that actually that that that the 0:14:58.200 --> 0:15:04.600 cost of electricity, like the cost effectiveness of AI in 0:15:04.720 --> 0:15:08.720 space will be overwhelmingly better than AI on the ground, 0:15:08.760 --> 0:15:13.640 so far, long before you exhaust potential energy sources on Earth, 0:15:14.240 --> 0:15:15.359 long before. 0:15:15.200 --> 0:15:16.880 Meaning like I think even PEFs in. 0:15:16.880 --> 0:15:21.920 The four or five year timeframe, the lowest cost way 0:15:22.040 --> 0:15:26.800 to do AI compute will be with solar powered AI satellites. 0:15:28.320 --> 0:15:31.720 So I'd say not more than five years from now. 0:15:32.080 --> 0:15:36.280 Wow, And just look at the supercomputers we're building together. 0:15:36.600 --> 0:15:38.600 Let's say each one of the racks is two tons. 0:15:39.000 --> 0:15:41.520 Out of that two tons, one point nine to five 0:15:41.520 --> 0:15:42.880 of it is probably for cooling. 0:15:43.200 --> 0:15:43.440 Right. 0:15:44.800 --> 0:15:47.720 Just imagine how tiny that little supercomputer is, right, each 0:15:47.760 --> 0:15:49.440 one of these GB three hundred racks and. 0:15:49.480 --> 0:15:50.680 Mull just be a little tiny thing. 0:15:50.920 --> 0:15:55.440 And just electricity generation is already becoming a challenge. So 0:15:55.680 --> 0:15:59.160 if you start doing any kind of scaling for both 0:15:59.200 --> 0:16:04.640 electricity generation and cooling, you realize, okay, space is incredibly compelling. 0:16:06.120 --> 0:16:10.640 So like, let's say you wanted to do I don't know, 0:16:10.680 --> 0:16:16.720 two or three hundred gigawats per year of AI computed, 0:16:17.600 --> 0:16:20.800 It's very difficult to do that on Earth. So the 0:16:21.560 --> 0:16:24.840 US average electricity usage last time I checked it was 0:16:24.840 --> 0:16:28.680 around four hundred and sixty gigawats per year average usage. 0:16:29.800 --> 0:16:34.800 So so something like say you know three hundred, If 0:16:34.800 --> 0:16:35.280 you're doing three. 0:16:35.200 --> 0:16:37.240 Hundred giga what's a year, that would be like two 0:16:37.280 --> 0:16:41.080 thirds of US electricity production per year. There's no way 0:16:41.120 --> 0:16:44.520 you're building power plants at that level. And then if 0:16:44.520 --> 0:16:46.080 you take it up to say a taro wat per 0:16:46.160 --> 0:16:49.280 year impossible, Like you have to do that in space 0:16:50.120 --> 0:16:52.080 there they're just is there just is. 0:16:52.040 --> 0:16:55.280 No way to do a tarow per year on Earth. 0:16:57.280 --> 0:17:03.760 And in space you've got tenuous solar you've got you 0:17:03.800 --> 0:17:06.240 don't You actually don't need batteries because it's always sunny 0:17:06.280 --> 0:17:10.920 in space, right exactly, And and the solar panels actually 0:17:10.960 --> 0:17:13.399 become cheaper because you don't need glass or framing. 0:17:14.640 --> 0:17:18.440 And the cooling is just radiative. So that's that's why 0:17:18.480 --> 0:17:18.919 I think. 0:17:18.760 --> 0:17:21.399 That's the dream. Yes, that's the dream. 0:17:21.800 --> 0:17:24.640 So Jensen, everybody last night was asking me, and I'm 0:17:24.680 --> 0:17:28.960 mindful it's Earning's call for you today. So I'm gonna 0:17:28.960 --> 0:17:32.320 say this delicately. Everybody has been asking me to ask you. 0:17:32.720 --> 0:17:34.080 Are we going to have an AI bubble? 0:17:35.640 --> 0:17:36.560 That's the last question? 0:17:37.240 --> 0:17:38.000 All right, let's not. 0:17:39.960 --> 0:17:41.359 All right, let me see well, let me just tell 0:17:41.400 --> 0:17:43.960 you what we see, okay. So, so I think it's 0:17:44.000 --> 0:17:46.520 really important when you look at what's happening around the 0:17:46.560 --> 0:17:48.520 world and go back to the first principles of what's 0:17:48.560 --> 0:17:51.760 happening in computer science and computing. There are three things 0:17:51.760 --> 0:17:55.359 that's happening. The first thing is that we all know 0:17:55.440 --> 0:17:58.080 that Moore's laws run its course, and the ability the 0:17:58.080 --> 0:18:01.040 amount of demand for computing, since the amount of computation 0:18:01.160 --> 0:18:03.960 we can get out of general purpose computing is really challenging, 0:18:04.320 --> 0:18:06.919 and so the world's been moving to accelerated computing for 0:18:06.960 --> 0:18:08.840 some time. We've been pushing this now for some over 0:18:08.880 --> 0:18:11.879 twenty years. Let me give you one statistic. I was 0:18:11.960 --> 0:18:19.480 just at supercomputing six years ago. CPUs were ninety percent 0:18:20.119 --> 0:18:23.480 of the world supercomputers top five hundred supercomputers. Six years ago. 0:18:23.960 --> 0:18:28.439 This year less than fifteen percent. Went from ninety percent 0:18:28.520 --> 0:18:31.760 to ten percent. And meanwhile accelerated computing went from the 0:18:31.800 --> 0:18:35.239 other way ten percent to now ninety percent. Okay, so 0:18:35.320 --> 0:18:39.000 you're seeing that inflection point, the transition in high performance 0:18:39.040 --> 0:18:42.240 computing from general purpose of computing to accelerated computing. Well 0:18:42.440 --> 0:18:45.080 of the one of the most data intensive, one of 0:18:45.119 --> 0:18:47.359 the most intensive computation things that the world does in 0:18:47.440 --> 0:18:52.119 cloud is data processing. Several hundred billion dollars of computation 0:18:52.280 --> 0:18:54.560 is done on just raw data. Process has nothing to 0:18:54.560 --> 0:18:58.520 do with AI, just CFL processing data frames. You know 0:18:58.560 --> 0:19:02.720 everybody's names, address, their sex, their age, where do they live, 0:19:02.840 --> 0:19:04.760 you know how much money they make. All of that 0:19:04.880 --> 0:19:07.320 sits into a data frame, and that data frame drives 0:19:07.359 --> 0:19:09.919 the world today, whether it's in banking or you know, 0:19:09.960 --> 0:19:12.840 whether it's in credit cards or of course e commerce 0:19:12.920 --> 0:19:17.080 and everything from ad recommendation and everything is driven off 0:19:17.119 --> 0:19:19.320 of that data frame. That data frame costs hundreds of 0:19:19.320 --> 0:19:21.520 billions are always to go compute. And so that's the 0:19:21.600 --> 0:19:23.760 number one thing end of More's lows. The second thing 0:19:24.200 --> 0:19:31.000 is generative AI. The most important application of the last 0:19:31.119 --> 0:19:34.879 fifteen years is called rexis recommended systems. How do we 0:19:35.000 --> 0:19:38.720 know what information to recommend to us in a social feed? 0:19:38.840 --> 0:19:41.080 How do you know what ad to recommend to somebody, 0:19:41.480 --> 0:19:44.320 what book to recommend, what movie to recommend? The world 0:19:44.400 --> 0:19:47.520 is the Internet is so gigantic without a recommended system 0:19:47.720 --> 0:19:50.359 that the little tiny phone of us will have no chance. 0:19:50.400 --> 0:19:54.040 I've ever seen the right information that REXUS is the 0:19:54.080 --> 0:19:57.639 engine of the Internet today. That's going generative AI. It 0:19:57.760 --> 0:20:00.000 used to be running on CPUs, now it runs on GPS, 0:20:00.160 --> 0:20:03.320 which then says the third thing. When if you just 0:20:03.359 --> 0:20:07.800 look at those two applications, many of the Internet companies 0:20:08.160 --> 0:20:12.560 can build an enormous number of GPUs supercomputers. Just doing that, 0:20:13.040 --> 0:20:16.320 of course, then it creates this the third opportunity on. 0:20:16.200 --> 0:20:17.800 Top of it, which is agentic AI. 0:20:17.880 --> 0:20:19.840 This is Grock and this is open AI, this is 0:20:19.880 --> 0:20:21.840 anthropic you know, this is Gemini. 0:20:22.080 --> 0:20:23.800 Agentic AI sits on top of that. 0:20:24.040 --> 0:20:26.879 But don't you know, don't forget to think about what 0:20:27.040 --> 0:20:32.720 is happening above, underneath what everybody sees as AI today, 0:20:33.000 --> 0:20:36.399 there's a whole movement of computing from jenniferpose computing to 0:20:36.440 --> 0:20:39.400 accelerated computing, and that if you just if you take 0:20:39.440 --> 0:20:43.399