WEBVTT - Meta to Become the Biggest Nuclear Buyer Among Hyperscalers 0:00:02.520 --> 0:00:07.040 Bloomberg Audio Studios, podcasts, radio news. 0:00:08.960 --> 0:00:10.639 This is Bloomberg Tech coming up. 0:00:10.760 --> 0:00:13.800 Meta agrees to a series of electricity deals to power 0:00:13.920 --> 0:00:17.240 data centers, making it the biggest fire of nuclear power 0:00:17.440 --> 0:00:21.560 among the hyperscalers plus Minimax, one of China's largest generative 0:00:21.560 --> 0:00:22.720 AI startups. 0:00:22.440 --> 0:00:23.599 Goes public in Hong Kong. 0:00:24.000 --> 0:00:27.760 For that means with China's AI ecosystem and Snowflake plans 0:00:27.800 --> 0:00:31.400 to buy Observe, an AI powered observability platform. We're going 0:00:31.440 --> 0:00:35.080 to discuss the move with Snowflake's c EO. Let's get 0:00:35.159 --> 0:00:37.640 right to it and what's happening in the markets. Later 0:00:37.640 --> 0:00:40.520 in the program, we're going to go big on Intel. Lipbutan, 0:00:40.600 --> 0:00:43.879 Intel CEO has spent the last twenty four hours in Washington, 0:00:43.960 --> 0:00:46.199 d C. In the White House with the President and 0:00:46.240 --> 0:00:49.839 with Howard Lutnik, the Commerce Secretary. Glowing remarks from President 0:00:49.880 --> 0:00:54.480 Trump about Litbutan's leadership of Intel, but also the progress 0:00:54.560 --> 0:00:56.600 that the US government's made in taking a steak in 0:00:56.640 --> 0:00:57.360 the chip maker. 0:00:57.680 --> 0:00:58.320 More to come. 0:00:58.600 --> 0:01:02.440 Then there's the big one investing in a multi gigawatt 0:01:02.480 --> 0:01:05.720 deal with some big names in nuclear energy. Opkloat up 0:01:05.720 --> 0:01:07.880 fourteen percent on that deal. Of course backed by some 0:01:08.000 --> 0:01:11.880 outman district corps. Up thirteen percent, Meta up eight ten 0:01:11.959 --> 0:01:14.600 to one percent. We said at the top of the program. 0:01:14.640 --> 0:01:18.360 This puts them among the biggest energy buyers, among the hyperscalers, 0:01:18.360 --> 0:01:22.399 and we remind ourselves Meta is not technically a hyperscaler. 0:01:22.520 --> 0:01:25.240 That's get the blue most Riley Griffin details here really 0:01:25.640 --> 0:01:30.199 important size of the deal, how many gigawatts, but also 0:01:30.240 --> 0:01:30.720 the structure. 0:01:30.720 --> 0:01:31.000 Please. 0:01:31.360 --> 0:01:34.800 Yeah, So three different agreements and they actually are different 0:01:34.800 --> 0:01:37.760 in form and functions. So this is about supporting up 0:01:37.840 --> 0:01:41.760 to six point six gigawatts in nuclear energy, but it 0:01:41.840 --> 0:01:46.280 is both about ensuring that existing nuclear power plants continue 0:01:46.319 --> 0:01:51.600 to thrive and investing in future nuclear power. Meta's head 0:01:51.600 --> 0:01:54.160 of global energy told me last night that they heard 0:01:54.200 --> 0:01:57.400 that there's real concern about the amount of energy that's 0:01:57.440 --> 0:02:00.480 out there. As you know ed, there's an insatiable demand 0:02:00.520 --> 0:02:04.280 for energy, and this cements their low carbon play. 0:02:04.960 --> 0:02:07.720 I'll find this fascinating because while Meta is not a 0:02:07.800 --> 0:02:10.560 hyperscale or a hyperscale of being a cloud computing company 0:02:10.600 --> 0:02:14.000 that basically leases capacity, Meta is doing this on its 0:02:14.040 --> 0:02:18.960 own behalf right. It has been aggressive in in securing 0:02:19.120 --> 0:02:22.399 the supply of energy it needs for its own data centers, 0:02:22.680 --> 0:02:25.799 which principally I use for training and inference of its 0:02:25.800 --> 0:02:28.680 own activity in AI. What else has it done and 0:02:28.680 --> 0:02:29.880 where does this bring them to date? 0:02:30.080 --> 0:02:32.320 Well, I want to note that you made a really 0:02:32.320 --> 0:02:36.919 important important point there. Prometheus and Hyperion. These are two 0:02:36.960 --> 0:02:41.000 of their biggest AI plays. Prometheus actually is in Ohio 0:02:41.120 --> 0:02:43.880 and these agreements are in Ohio and Pennsylvania. So some 0:02:43.919 --> 0:02:46.440 of this energy presumably is going to go to the 0:02:46.480 --> 0:02:49.320 Prometheus data center cluster, which is one of their biggest plays. 0:02:49.680 --> 0:02:53.960 But we're also seeing Meta turn with Hyperion to natural gas. 0:02:54.200 --> 0:02:57.440 At least three natural gas plants are being fired up 0:02:57.720 --> 0:03:00.560 just for that single facility, So it's not like nuclear 0:03:00.680 --> 0:03:03.160 is the only strategy that Meta is employing here. 0:03:03.440 --> 0:03:06.360 We're going to get to the sort of available energy 0:03:06.400 --> 0:03:09.120 sources around the world in a little bit. There is 0:03:09.160 --> 0:03:12.360 an interesting point you co reported this with Will Wade, 0:03:12.400 --> 0:03:15.239 who's just been really on top of the nuclear side 0:03:15.280 --> 0:03:17.880 of the story. There does seem to be some anxiety 0:03:18.240 --> 0:03:22.160 from the technology companies that the existing NUCLEME infrastructure that 0:03:22.240 --> 0:03:25.120 does exist in America, limited as it may be, is 0:03:25.160 --> 0:03:27.600 also at risk at being shut down. Is this kind 0:03:27.600 --> 0:03:28.840 of future proofing a little bit. 0:03:29.040 --> 0:03:31.120 Absolutely, that is what I'm hearing from metas head of 0:03:31.120 --> 0:03:35.400 Global Energy. They started back a December two years ago 0:03:36.400 --> 0:03:39.840 and they were looking into what nuclear needed to keep going, 0:03:40.280 --> 0:03:43.000 and they heard we need investment now into plants that 0:03:43.000 --> 0:03:45.520 could potentially shutter. So this is a future proofing, as 0:03:45.560 --> 0:03:45.839 you say. 0:03:46.040 --> 0:03:48.880 Bloomberg's Riley Griffin leading our coverage of Meta with a 0:03:48.960 --> 0:03:51.200 must read, Thank you very much. Access to power has 0:03:51.240 --> 0:03:53.640 been top of mind in AI. We spoke about that 0:03:53.720 --> 0:03:56.160 with Nvidia CEO Jensenmong earlier this week. 0:03:56.960 --> 0:03:59.760 In order for our new industry to emerge, you need energy, 0:04:00.080 --> 0:04:02.360 and so I think it's safe to say that we 0:04:02.400 --> 0:04:04.920 wish we had more energy in the United States. Your 0:04:05.040 --> 0:04:07.520 wis should have more energy. I think the world all 0:04:07.560 --> 0:04:09.720 wish we have more energy, and so we have to 0:04:09.720 --> 0:04:13.280 invest in all sorts of different forms of energy. 0:04:13.800 --> 0:04:15.320 Let's bring in Paul Meeks for more. 0:04:15.360 --> 0:04:19.000 He's managing director and head of Technology research of Freedom 0:04:19.040 --> 0:04:22.039 Capital Markets. And when I was reading your research last night, Paul, 0:04:22.640 --> 0:04:24.600 A little bit relieved coming up this morning when I 0:04:24.640 --> 0:04:28.440 saw the Meta headline, because you're very focused on the 0:04:28.480 --> 0:04:31.960 real terms footprint of data centers. That's what we're talking 0:04:31.960 --> 0:04:35.480 about here. You heard Riley's reporting on the specifics of 0:04:35.480 --> 0:04:40.120 what Meta has done to secure future capacity. How much 0:04:40.160 --> 0:04:42.320 of a bottleneck is that to your mind right now? 0:04:42.440 --> 0:04:43.200 For this sector? 0:04:45.120 --> 0:04:49.320 It is absolutely critical of all the bottomnecks. It is 0:04:49.600 --> 0:04:53.880 the most important. I like what Meta is doing here, 0:04:53.920 --> 0:04:59.480 but folks need to realize that if you ramp up 0:04:59.640 --> 0:05:03.040 nuclear capacity, we're not going to really see power generation 0:05:03.320 --> 0:05:07.920 until twenty thirty earliest, maybe even not until twenty thirty two. 0:05:08.279 --> 0:05:10.479 So what do we do for the next four to 0:05:10.520 --> 0:05:16.960 six years? So the immediate draw is probably that gas. 0:05:17.160 --> 0:05:19.360 But when I look at all of these companies and 0:05:19.400 --> 0:05:23.200 we cover the AI hyperscalers and the neo clouds and 0:05:23.240 --> 0:05:28.560 everybody adjacent to them, we are looking for their availability 0:05:29.360 --> 0:05:34.080 for power, their capacity to find power, because it's not 0:05:34.279 --> 0:05:39.040 a demand issue, it is a supply issue focused on 0:05:39.120 --> 0:05:41.159 this particular metric. 0:05:41.520 --> 0:05:42.320 You know, pool. 0:05:42.360 --> 0:05:45.480 I played you just just some of the interview that 0:05:46.000 --> 0:05:48.280 we conducted with Jensen Wong earlier this week, and when 0:05:48.279 --> 0:05:51.400 I was sitting in front of him, you don't really 0:05:51.400 --> 0:05:54.919 get any sense of anxiety from him, right Think about 0:05:55.200 --> 0:05:58.159 the mechanics of how this transaction works and video brings 0:05:58.320 --> 0:06:01.760 GPUs actually that's probably not fair. Increasingly they bring more 0:06:01.839 --> 0:06:05.279 of the content of the server, but there is an 0:06:05.320 --> 0:06:09.719 acknowledgment that something needs to change, principally in America. He 0:06:09.760 --> 0:06:12.360 did go on to talk about the differences between Europe, 0:06:12.440 --> 0:06:15.520 Asia and the United States and where we're sourcing energy. 0:06:16.080 --> 0:06:18.159 Do you get a sense from your research that the 0:06:18.200 --> 0:06:23.280 companies you cover acknowledge the severity of that deficit right now? 0:06:24.960 --> 0:06:28.599 So when they are public facing, particularly somebody like Jensen Wong, 0:06:28.640 --> 0:06:32.599 who is essentially the spokesperson for the entire industry, they 0:06:32.640 --> 0:06:38.200 have to exude optimism. However, behind the scenes, particularly when 0:06:38.200 --> 0:06:42.400 I'm talking to companies about my financial models, which of 0:06:42.440 --> 0:06:45.920 course the revenues and then the follow on cash flow 0:06:45.960 --> 0:06:47.920 and earnings come from. 0:06:48.400 --> 0:06:49.760 Do you have the power? 0:06:50.200 --> 0:06:54.440 Yes, they are anxious, and I think they're actually very anxious. 0:06:56.320 --> 0:06:58.720 There will be people watching Bloomberg Tech that work at 0:06:58.760 --> 0:07:01.240 Meta or in the AI industry. There will be Meta 0:07:01.360 --> 0:07:05.200 investors watching this program, and some of these people may 0:07:05.240 --> 0:07:06.520 never have heard of an outclow. 0:07:07.160 --> 0:07:08.040 Have you had to. 0:07:08.120 --> 0:07:12.120 Change the remit and breadth of your coverage if you 0:07:12.160 --> 0:07:16.920 say I cover AI data center to include future sources 0:07:16.920 --> 0:07:19.480 of electricity supply, Yeah. 0:07:19.320 --> 0:07:23.480 It's really interesting, ed that years ago, you know, nobody 0:07:23.520 --> 0:07:26.040 in my line of work covering tech sector like I 0:07:26.120 --> 0:07:29.840 have for decades, would be interested in any utility company. 0:07:29.920 --> 0:07:33.679 It's the antithesis of tech. They are boring, regulated, slow 0:07:33.760 --> 0:07:37.400 growth businesses. However, now they've come to the four and 0:07:37.480 --> 0:07:40.160 of course these are not just your regular utilities, you know, 0:07:40.240 --> 0:07:43.840 they're your new era utilities. But yes, if you are 0:07:43.880 --> 0:07:47.440 a good analyst, you have to cover the entire supply chain, 0:07:47.920 --> 0:07:52.520 the entire ecosystem. And because demand is so robust, we're 0:07:52.560 --> 0:07:58.600 focusing on supply absolutely critical. And it's not just the power. 0:07:59.240 --> 0:08:03.920 There are physical constraints here, like simply pouring the cement 0:08:04.600 --> 0:08:08.320 to build the next data center, which can't be a bottleneck, 0:08:08.360 --> 0:08:11.240 even though it sounds, you know, so old school in 0:08:11.320 --> 0:08:14.920 a new school business. But yes, these are some factors 0:08:14.920 --> 0:08:17.800 that are very, very interesting that people didn't even conceive 0:08:17.880 --> 0:08:19.280 of a couple of years ago. 0:08:20.400 --> 0:08:27.560 Pol is there an adequate federal level framework regulation to 0:08:27.680 --> 0:08:31.520 support the energy requirements of the AI industry in this country? 0:08:32.520 --> 0:08:35.240 You know, there is not. And here is my problem. 0:08:35.559 --> 0:08:39.479 We are still suffering from a lack of solid regulation 0:08:40.160 --> 0:08:42.800 for the Internet because even today in the US the 0:08:42.880 --> 0:08:47.040 Internet is operating under nineteen ninety six legislation when Bill 0:08:47.080 --> 0:08:51.760 Clinton was president, and so I'm hoping that we're not 0:08:51.880 --> 0:08:52.760 too restrictive. 0:08:52.880 --> 0:08:53.040 Right. 0:08:53.080 --> 0:08:55.960 We need to embrace our entrepreneurial companies and give them 0:08:56.480 --> 0:08:59.959 room to roll, but we need to have some guards 0:09:00.120 --> 0:09:04.880 else on everything AI because what happened last time is 0:09:04.920 --> 0:09:09.640 we end up abdicating the legislation to the EU because 0:09:09.679 --> 0:09:12.800 we didn't have the will to do it ourselves. So 0:09:12.840 --> 0:09:15.839 we missed that opportunity in the Internet. Here is the 0:09:15.880 --> 0:09:19.520 next wave. We need to embrace it and at least 0:09:19.520 --> 0:09:21.600 come up with some guardrails that are going to guide 0:09:21.640 --> 0:09:24.080 all this because there are some important decisions on the 0:09:24.120 --> 0:09:26.040 federal level to be made. 0:09:27.240 --> 0:09:29.320 In your research pro it's very interesting to see you 0:09:29.400 --> 0:09:32.320 approach data center from the perspective of yes, you know, 0:09:32.840 --> 0:09:38.040 energy supply, but also real estate. The reats right, and 0:09:38.360 --> 0:09:41.760 what you want to ask you is away from the GPUs, 0:09:42.480 --> 0:09:45.679 where are the frailties in that industry? Is it literally 0:09:45.720 --> 0:09:48.880 the folks building the tin can around the outside of 0:09:48.920 --> 0:09:51.559 the data center, the labor, the concrete. 0:09:52.880 --> 0:09:56.680 Yeah, we've seen some hit in short term revenues because 0:09:56.720 --> 0:09:59.280 a couple of the number or a few of the 0:09:59.600 --> 0:10:03.440 players have subcontractors. Now this is not them, this is 0:10:03.480 --> 0:10:05.679 their subcontractors. 0:10:04.960 --> 0:10:06.280 That are delayed. 0:10:06.320 --> 0:10:10.640 What we call building a powered shell. So it's more 0:10:10.679 --> 0:10:13.080 than a quandset hut that you might find the army in. 0:10:13.520 --> 0:10:16.800 These powered shells are a little bit more sophisticated. But 0:10:16.920 --> 0:10:19.920 one company I cover had a one hundred and fifty 0:10:20.000 --> 0:10:23.960 million dollars revenue deferral from twenty five to twenty six 0:10:24.480 --> 0:10:28.200 because at one of their forty one data centers someone 0:10:28.200 --> 0:10:31.960 couldn't pour concrete because it was raining. And so yes, 0:10:32.600 --> 0:10:35.440 all these things come into play, and it's interesting that 0:10:35.520 --> 0:10:40.480 they may not necessarily be technical bottlenecks, but old school 0:10:40.679 --> 0:10:44.600 physical including construction bottlenecks. In addition to the power. 0:10:45.200 --> 0:10:46.599 I just go back from Vegas. 0:10:46.840 --> 0:10:50.560 With all the attention on artificial intelligence in the digital world, 0:10:50.880 --> 0:10:53.880 we have an industry that's getting hit by a rainstorm. 0:10:54.000 --> 0:10:57.440 Pool meeks, buckets, unbelievable. It's great to have you on 0:10:57.440 --> 0:11:00.000 Bloomberg Tech. A lot more to come and coming out 0:11:00.200 --> 0:11:03.240 an exclusive interview with the co founder and COO of 0:11:03.280 --> 0:11:06.679 Minimax after the company just went public in Hong Kong, 0:11:07.040 --> 0:11:08.400 and that's the debut trade. 0:11:08.600 --> 0:11:10.120 And this is Bloomberg Tech. 0:11:23.679 --> 0:11:27.560 Minimax, one of China's largest generative AI startups, backed by 0:11:27.559 --> 0:11:30.760 Ali Barber and Abadavi Sovereign Well Fund, is now public. 0:11:31.160 --> 0:11:33.960 Shares surge in Hong Kong after an IPO that raised 0:11:34.000 --> 0:11:37.559 six hundred and nineteen million dollars, closing up one hundred 0:11:37.600 --> 0:11:41.600 and nine percent in its Friday debut trading. Bloomberg Stephen 0:11:41.640 --> 0:11:46.640 Engel spoke exclusively with co founder and COO Yee Yun 0:11:46.880 --> 0:11:50.760 about growth and expansion plans as Minimax faces hot AI 0:11:50.800 --> 0:11:53.320 competition both locally and globally. 0:11:54.240 --> 0:11:58.040 We really focus out that capital efficiency caused the efficiency, 0:11:58.320 --> 0:12:01.520 so we had always spent a round five hundred media 0:12:01.640 --> 0:12:04.959 USC in the in total, probably only one or two 0:12:05.040 --> 0:12:08.640 percent of the biggen suspending, So we did all the organization, 0:12:08.840 --> 0:12:12.000 mad creativity innovations on. 0:12:11.160 --> 0:12:13.800 The h How do you compete though, if you're going 0:12:13.840 --> 0:12:16.120 to go global, how do you compete with an open 0:12:16.160 --> 0:12:20.000 AI or others that are throwing lots of capital at 0:12:20.000 --> 0:12:21.640 their business? 0:12:22.000 --> 0:12:23.360 Is it viable and what is the. 0:12:23.320 --> 0:12:26.160 Technology gap looking like in the next few years. 0:12:26.640 --> 0:12:30.400 Actually, I don't think it's a competition. It's more about 0:12:30.800 --> 0:12:34.240 all the top talents across all the world. They bring 0:12:34.400 --> 0:12:38.000 the technology break through to the society, to the end users. 0:12:38.200 --> 0:12:40.839 So you will see each models have pros and accounts, 0:12:41.000 --> 0:12:44.199 so we focus more resources on building the models, building 0:12:44.200 --> 0:12:48.480 the best product experience to a more and more uses 0:12:48.480 --> 0:12:49.439 from all the world. 0:12:49.320 --> 0:12:49.880 To use it. 0:12:50.240 --> 0:12:53.199 We're hearing as well, the Chinese government might approve in 0:12:53.360 --> 0:12:57.760 Vidia's H two hundred for China in this quarter. Perhaps 0:12:58.080 --> 0:13:02.040 is that something you need to look at importing more 0:13:02.040 --> 0:13:04.240 advanced accelerators for your training. 0:13:03.960 --> 0:13:04.320 And the like. 0:13:04.679 --> 0:13:08.000 I think in every industry with a really high growth 0:13:08.200 --> 0:13:11.120 when the new technology comes here, so you will see 0:13:11.320 --> 0:13:15.440 every startups, all the companies. This please resources comestring probably 0:13:15.480 --> 0:13:19.000 chief constring computing constring is only one of the challenges. 0:13:19.400 --> 0:13:23.040 So in the past you will see our model performers 0:13:23.080 --> 0:13:26.560 and our product performers, so our resources is stable. We 0:13:26.559 --> 0:13:30.080 can get access to all of them, so we focus 0:13:30.160 --> 0:13:32.720 more on our R and D and foundation model. So 0:13:32.760 --> 0:13:36.000 we just use whatever chiefs who make sense for us 0:13:36.000 --> 0:13:37.720 at that specific stage. 0:13:37.920 --> 0:13:41.480 Is your goal to stay indigenously with your chips locally 0:13:41.559 --> 0:13:45.720 made chips or would you buy nvidio chips as well? 0:13:46.160 --> 0:13:48.840 It doesn't matter what kinds of chiefs it is. It's 0:13:48.920 --> 0:13:52.560 more about which chiefs can give us best ROI, which 0:13:52.600 --> 0:13:55.520 can help us to achieve our mission to make the 0:13:55.559 --> 0:13:59.640 best technology too accessible to all of the users across 0:13:59.760 --> 0:14:00.000 the world. 0:14:00.280 --> 0:14:04.840 So when does the emphasis turn from efficiency to profitability 0:14:05.240 --> 0:14:08.800 and also getting that steady revenue growth that will then 0:14:09.080 --> 0:14:11.880 justify a higher stock price valuation. 0:14:12.480 --> 0:14:14.920 So we put loss of R and D resources and 0:14:15.040 --> 0:14:19.360 innovations on the efficiency part. You will see our API 0:14:19.960 --> 0:14:21.960 we have the enterprise business and you will see the 0:14:22.000 --> 0:14:24.920 gross profit margin of our API is more than sixty 0:14:24.920 --> 0:14:28.920 five percent, which is probably already one of the highest globally. 0:14:29.280 --> 0:14:32.840 So we focus more on make the best technology and 0:14:32.880 --> 0:14:35.320 the best the user experience that people would love to 0:14:35.400 --> 0:14:39.520 pay for the best performers models for their use cases. 0:14:39.800 --> 0:14:43.840 You've also come out of a price war essentially a 0:14:43.960 --> 0:14:47.120 race to the bottom some have called, because of you're 0:14:47.120 --> 0:14:49.400 trying to keep costs down and you want to survive 0:14:49.480 --> 0:14:52.240 as well. There was a battle of survival and you 0:14:52.360 --> 0:14:55.680 have been one of the survivors. But at the same time, 0:14:56.720 --> 0:14:59.200 the startups are not the only ones. The big players 0:14:59.240 --> 0:15:02.480 Ali Baba, by Dads, tens and others. How do you 0:15:02.520 --> 0:15:05.440 compete with them at a low cost basis? 0:15:05.920 --> 0:15:09.880 Actually it's not. I don't think that's a press word. 0:15:10.160 --> 0:15:14.160 It's more about the performers the competition, So you need 0:15:14.200 --> 0:15:17.000 to make your model best. And the people I mean 0:15:17.040 --> 0:15:19.920 and the user are interpress customers. They don't choose your 0:15:20.000 --> 0:15:23.280 model because of its cheap or expensive. It's more about 0:15:23.400 --> 0:15:26.720 its performers. So that's the key. As you said, I 0:15:26.720 --> 0:15:29.880 think it's not about the competition between like all these 0:15:29.920 --> 0:15:32.800 big chenz, but it's more about some of them. They 0:15:32.800 --> 0:15:35.880 are like our cloud providers and some of their products 0:15:36.240 --> 0:15:39.520 you see our models, So it's more about the collaborations. 0:15:39.720 --> 0:15:43.040 All this company brings the foundation models to all the world. 0:15:44.720 --> 0:15:48.200 Many Max co founder and COO Yaun speaking exclusively with 0:15:48.240 --> 0:15:51.120 Bloomberg Steven Engel. Public listings in Hong Kong are hot 0:15:51.160 --> 0:15:54.840 right now. Chinese industrial robot maker in Events is now 0:15:54.880 --> 0:15:58.960 also considering a second listing in Hong Kong, according to sources. 0:15:59.080 --> 0:16:02.960 Let's get out to Bloomberg Executive Tech editor Peter Elstrom. 0:16:02.240 --> 0:16:02.880 To discuss this. 0:16:02.960 --> 0:16:06.040 But also like this growing trend of more of China's 0:16:06.120 --> 0:16:09.760 sort of post chat gbtai firms going public. The team 0:16:09.800 --> 0:16:13.680 that you lead out in Asia busy right now. In 0:16:13.960 --> 0:16:16.560 Events's case, what are the details in our reporting that 0:16:16.560 --> 0:16:17.400 we need to know about. 0:16:18.800 --> 0:16:21.240 Yeah, there are a lot of these IPOs that we're seeing, 0:16:21.360 --> 0:16:24.840 and we're seeing a much more forward move into the 0:16:24.880 --> 0:16:27.560 public markets for a bunch of these companies and events 0:16:27.560 --> 0:16:30.200 has already gone public in the mainland and Shenjen now 0:16:30.240 --> 0:16:32.120 it's looking at an IPO in Hong Kong. 0:16:32.160 --> 0:16:32.320 Two. 0:16:32.560 --> 0:16:36.080 That's partly a reflection of how hot demand is from 0:16:36.240 --> 0:16:39.080 investors for these kinds of companies. When you're talking about 0:16:39.080 --> 0:16:41.840 Minimax in particular, they went public, as you mentioned, their 0:16:41.840 --> 0:16:44.360 shares more than doubled in the debut. The founder of 0:16:44.400 --> 0:16:47.280 the company is very fascinating. These are some people called 0:16:47.280 --> 0:16:50.160 them the AI tigers, some people the AI dragons. But 0:16:50.200 --> 0:16:53.359 these are startups that really have very very different strategies 0:16:53.400 --> 0:16:55.520 for how they're approaching the market. Compared with some of 0:16:55.520 --> 0:16:58.360 the US competitors, Like in open Ai, they tend to 0:16:58.360 --> 0:17:00.720 be much lower cost models. They tend to be much 0:17:00.760 --> 0:17:04.520 more aggressive about implementing the technology with their customers, with 0:17:04.560 --> 0:17:07.680 corporate customers, and giving them very low cost, very low, 0:17:07.800 --> 0:17:10.199 very efficient models that they can use and they can 0:17:10.240 --> 0:17:12.960 actually deploy more quickly than some of the US counterparts. 0:17:13.480 --> 0:17:16.120 PETE in the world of technology or in financial market's 0:17:16.119 --> 0:17:18.600 like a term like an AI tiger or AI dragon. 0:17:18.960 --> 0:17:21.080 You know, it's kind of commonplace. We bandy it around. 0:17:21.080 --> 0:17:23.399 But for those watching Bloomberg Tech that are saying, what 0:17:23.480 --> 0:17:26.560 are you talking about? How do we define the AI tiger? 0:17:26.600 --> 0:17:27.560 What are we talking about here? 0:17:28.760 --> 0:17:31.240 Well, we're talking about a whole series of companies coming 0:17:31.240 --> 0:17:35.959 out of China. In Stevens interview with the Minimax COO, 0:17:36.119 --> 0:17:38.680 they talked about the competition there. We've of course heard 0:17:38.720 --> 0:17:41.320 a lot about Deep Seek and how good their model is, 0:17:41.400 --> 0:17:43.560 how technically good it is, but also how low cost 0:17:43.600 --> 0:17:45.720 it is. Behind Deep Seek, you have a whole bunch 0:17:45.760 --> 0:17:48.800 of other companies like Minimax, like Jipou, which just went 0:17:48.880 --> 0:17:51.600 public on Thursday. So yeah, the IPO of Geepoo on Thursday, 0:17:51.920 --> 0:17:54.879 mini Max on Friday investors showed a lot of demand 0:17:54.960 --> 0:17:57.000 for those shares. And what these companies have been able 0:17:57.040 --> 0:17:59.080 to show is that with a very low cost model, 0:17:59.280 --> 0:18:01.720 they're out there in the market competing. It's not clear 0:18:01.800 --> 0:18:03.879 that they're going to be able to compete directly with 0:18:04.080 --> 0:18:07.520 Open AI or Anthropic, but they have very innovative models. 0:18:07.520 --> 0:18:09.000 I mean, first of all, the US models are not 0:18:09.040 --> 0:18:11.080 going to be let into the China market. But these 0:18:11.119 --> 0:18:14.119 companies are also going out very aggressively and competing in 0:18:14.240 --> 0:18:16.560 some other parts of the world. We've written about how 0:18:16.640 --> 0:18:19.480 much traction Deep Seek is getting in Africa in particular, 0:18:19.720 --> 0:18:22.560 because their companies are sensitive to costs. They're sensitive to 0:18:22.600 --> 0:18:26.200 how much power, how much computer resources they need to 0:18:26.240 --> 0:18:28.400 be able to deploy these models. So they are having 0:18:28.480 --> 0:18:31.000 some successes. We're going to see a lot of competitive 0:18:31.040 --> 0:18:32.200 clashes in the months ahead. 0:18:33.280 --> 0:18:36.480 Bloomberg's executive editive for Tech, Peter Alstrom. Great to catch 0:18:36.560 --> 0:18:38.000 up with you, Thank you so much. Now coming up 0:18:38.040 --> 0:18:38.199 on the. 0:18:38.200 --> 0:18:43.360 Program, Love Musks Grock image generation tool faces backlash after 0:18:43.440 --> 0:18:47.800 reports that's generated thousands of undressed women and children. 0:18:48.119 --> 0:18:50.880 Go more on that data. Next, this is Bloomberg Tech. 0:19:00.760 --> 0:19:04.680 Elon Musk's AI startup Xai is burning cash quickly. The 0:19:04.760 --> 0:19:08.160 company spent almost eight billion dollars in the first nine 0:19:08.200 --> 0:19:10.600 months of the year. That's according to internal docs that 0:19:10.720 --> 0:19:15.000 Bloomberg's viewed. Though revenue has nearly doubled quarter over quarter 0:19:15.440 --> 0:19:18.840 to one hundred and seven million dollars, Xai's losses have 0:19:18.960 --> 0:19:21.879 been mounting on costs to build data centers, recruit talent 0:19:22.280 --> 0:19:25.440 developed software and to build AI that is self sufficient, 0:19:25.520 --> 0:19:29.320 and according to Bloomberg's reporting, the plan is to use 0:19:29.400 --> 0:19:33.320 that AI to eventually power Tesla's humanoid. 0:19:33.000 --> 0:19:34.480 Robots in the Optimus program. 0:19:35.040 --> 0:19:38.800 Right Sticking with XAI, the company has limited access to 0:19:38.920 --> 0:19:43.600 Grock's image generation tool for most users, following widespread criticism 0:19:44.040 --> 0:19:49.040 that the feature was producing thousands of explicit images of 0:19:49.160 --> 0:19:53.639 women and children. Bloomberg reporter Cecilia Deannastasio broke, the story 0:19:53.760 --> 0:19:57.760 has been tracking the developments, Cecilia, that we just summarized 0:19:58.560 --> 0:20:02.040 the issue, right think, let's start with the methodology, how 0:20:02.080 --> 0:20:04.440 we went about looking at the data, how we got 0:20:04.520 --> 0:20:05.960 the data, and we'll go from that. 0:20:07.520 --> 0:20:10.320 Thank you for having me. Bloomberg worked as a researcher 0:20:10.359 --> 0:20:13.840 who scraped thousands of images produced by groc and published 0:20:13.880 --> 0:20:16.760 to the platform X between the period January fifth and 0:20:16.880 --> 0:20:20.800 January sixth. The researcher analyzes images to determine what percent 0:20:20.840 --> 0:20:23.840 of them were sexualized and neudifying, and found that every 0:20:23.960 --> 0:20:27.400 hour during that twenty four hour period, X was publishing 0:20:27.440 --> 0:20:31.440 about six thousand, seven hundred images identified as such, and 0:20:31.920 --> 0:20:35.639 just for a comparison, websites that are dedafied dedicated to 0:20:35.880 --> 0:20:38.879 publishing deep faked images AI generated images of women who 0:20:38.880 --> 0:20:43.360 are sexualized neudified. Those top five websites out there published 0:20:43.400 --> 0:20:47.560 together eighty images per hour. So GROC is one of 0:20:47.600 --> 0:20:49.800 the biggest deep fake producers on the Internet today. 0:20:51.640 --> 0:20:53.239 Where are these images showing up? 0:20:53.680 --> 0:20:57.040 You know, to the uninitiated people may not know the 0:20:57.119 --> 0:21:00.399 sort of interconnects between XAI the AI company, and x 0:21:00.800 --> 0:21:03.639 the social media platform, for example, but they are now 0:21:03.760 --> 0:21:07.080 kind of one entity as they combine last year. 0:21:08.840 --> 0:21:11.879 Well, that connection is exactly why x has become one 0:21:11.920 --> 0:21:14.560 of the biggest deep fake websites on the internet. Anybody 0:21:14.720 --> 0:21:17.280 can well not anybody now, now you have to pay 0:21:17.400 --> 0:21:19.400 to do this, But over the last couple of days, 0:21:19.840 --> 0:21:23.280 anybody could just at rock on the X platform and say, 0:21:23.480 --> 0:21:25.520 put this woman in a bikini under an image that 0:21:25.600 --> 0:21:28.200 a woman had post on the internet. And a lot 0:21:28.240 --> 0:21:30.760 of women women primarily are the victims of this, took 0:21:30.840 --> 0:21:33.639 to complaining to GROC directly and arguing with GROC in 0:21:33.680 --> 0:21:34.320 their comments. 0:21:34.560 --> 0:21:36.240 GROC would apologize for posting. 0:21:36.000 --> 0:21:38.760 These images, but then continue posting them and not take 0:21:38.840 --> 0:21:42.040 down Many of the images that the women felt were violating. 0:21:42.440 --> 0:21:46.880 Succeed That did X or XAI respond to our reporting, 0:21:47.040 --> 0:21:47.920 engage with us on it. 0:21:49.280 --> 0:21:52.040 X has not responded to our numerous requests for comment 0:21:52.160 --> 0:21:55.440 over the last couple of days. Elon Musk himself, in 0:21:55.600 --> 0:21:58.359 a reply to a post on X said anybody using 0:21:58.440 --> 0:22:00.840 groc to make illegal con tent will suffer the same 0:22:00.880 --> 0:22:04.320