WEBVTT - OpenAI Unveils First Custom AI Chip With Broadcom 0:00:00.080 --> 0:00:15.080 Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is live 0:00:15.200 --> 0:00:18.680 from the heart of Silicon Valley with ed La though 0:00:18.840 --> 0:00:19.880 in Van Francisco. 0:00:22.880 --> 0:00:24.400 This is Bloomberg Tech coming up. 0:00:24.440 --> 0:00:28.320 Open Ai unveils its first custom AI chip called Kalapino, 0:00:28.640 --> 0:00:32.520 developed in partnership with Broadcom, plus s k Highnex's plane 0:00:32.560 --> 0:00:36.159 to raise twenty nine billion dollars in a landmark US listing, 0:00:36.479 --> 0:00:40.199 racing to increased capacity to meet memory chip demand, and 0:00:40.240 --> 0:00:44.120 Cerebrus reports quarterly earnings for the first time since going public. 0:00:44.200 --> 0:00:49.479 CEO Andrew Feldman joins us later this hour. Let's get 0:00:49.520 --> 0:00:52.680 to our top story custom Silicon open ai is tackling 0:00:52.680 --> 0:00:56.680 one of AI's biggest challenges, the supply and costs of computing. 0:00:56.920 --> 0:01:00.920 The company unveiled Kalapino Intelligence process. So that's the wafer, 0:01:01.320 --> 0:01:03.920 a custom AI chip developed with broad coomp which the 0:01:03.960 --> 0:01:07.560 companies say carries a fifty percent lower cost versus a 0:01:07.640 --> 0:01:12.720 typical AIGPU. The focus faster, cheaper AI inference shares a 0:01:12.760 --> 0:01:16.360 broad Open pretty higher round two percent. We're off session highs, 0:01:16.360 --> 0:01:18.520 but still up a percentage point or so. In a 0:01:18.600 --> 0:01:21.600 market where tech at the index level kind of flat. 0:01:21.800 --> 0:01:25.200 This is outperformance. Blinbo Tech editor Seth Figgerman is here 0:01:25.240 --> 0:01:28.080 with the details. This is the big picture, right, open 0:01:28.120 --> 0:01:32.160 ai is compute constrained, heavily reliant on Nvidia. 0:01:32.240 --> 0:01:34.640 They want to diversify. What do we need to know? 0:01:35.800 --> 0:01:38.280 Yeah, that's very I think, you know, Opening I wants 0:01:38.319 --> 0:01:42.400 to own more of the infrastructure stack, have more flexibility here, 0:01:42.440 --> 0:01:44.760 and then I think importantly, use the expertise that it's 0:01:44.760 --> 0:01:47.360 developed into what makes air models run better and more 0:01:47.360 --> 0:01:49.600 efficiently and apply that to the hardware and the same 0:01:49.640 --> 0:01:52.440 way that I think we're seeing Google do with its TPUs, 0:01:52.480 --> 0:01:54.160 and the hope would be that over time they can 0:01:54.160 --> 0:01:57.040 build a chip that's more cost efficient and also better 0:01:57.120 --> 0:01:59.520 performing and eventually cuts down its own costs. 0:02:00.600 --> 0:02:04.760 Bloomberg News spoke to both Hocktan the CEO of Broadcom, 0:02:05.200 --> 0:02:07.920 and open ai is hardware chief. I guess you know, 0:02:08.000 --> 0:02:09.920 from raw Comm's perspective, there was quite a lot of 0:02:09.919 --> 0:02:12.240 fighting talk. Right, they see a lot of demand for 0:02:12.280 --> 0:02:15.520 a six or custom silicon. But what was interesting to 0:02:15.520 --> 0:02:18.200 me is hock Tans saying to us they see a 0:02:18.200 --> 0:02:21.360 world where every frontier lab goes to have a custom chip. 0:02:21.800 --> 0:02:23.960 Why yeah, I mean. 0:02:24.360 --> 0:02:26.240 It gets back to what we said about Openny a 0:02:26.280 --> 0:02:29.000 minie ago, I think giving them the versatility to kind 0:02:29.040 --> 0:02:32.360 of control their infrastructure stack and build customer hardware that 0:02:32.400 --> 0:02:34.799 meets their needs. And now I think Hocktan's com and 0:02:34.840 --> 0:02:37.560 he's saying there's not that many of these frontier model developers. 0:02:37.560 --> 0:02:39.040 So I think agree between the lines we're kind of 0:02:39.040 --> 0:02:41.120 talking about. Is anthropic, I'm going to move in this 0:02:41.200 --> 0:02:44.040 direction and that's certainly possibility. And to your other point, 0:02:44.280 --> 0:02:48.000 I think Hocktan sounds quite bullish generally on the prospects 0:02:48.040 --> 0:02:50.320 for this chip and chips like it going forward. He 0:02:50.400 --> 0:02:53.080 sees no limit to the demand for this kind of 0:02:53.120 --> 0:02:57.200 infrastructure and is predicting that the demand and roll out 0:02:57.280 --> 0:02:59.760 for this chip next year will exceed his prior estimates 0:02:59.800 --> 0:03:03.160 for one point three gigawads to supply on. 0:03:03.080 --> 0:03:06.240 The open AI side, it's the latest example of them 0:03:06.320 --> 0:03:09.400 being prepared to spend. Did we get any sense from 0:03:09.440 --> 0:03:13.679 our conversation with the company about how this custom chip 0:03:13.720 --> 0:03:17.120 will rank in their overall plan for compute, whether it's leasing, 0:03:17.200 --> 0:03:19.320 data center, buying and video chips, et cetera. 0:03:20.080 --> 0:03:21.800 Yeah, you know, they remain pretty tight. Lop DOWNE what 0:03:21.800 --> 0:03:23.760 the financing for this will look like. We have previously 0:03:23.760 --> 0:03:26.280 reported that Opening I planned to spend tens of billions 0:03:26.280 --> 0:03:28.760 of dollars on these chips, with Broadcom part of its 0:03:28.919 --> 0:03:33.119 hundreds of billions of dollars in infrastructure commitments. Seeing Broadcom 0:03:33.120 --> 0:03:35.840 has previously said that they were standing up a chip 0:03:35.880 --> 0:03:38.840 financing vehicle that could help Opening I's efforts, but it's 0:03:38.880 --> 0:03:41.880 unclear that's the primary financing mechanism for what Opening I 0:03:41.920 --> 0:03:42.440 will do here. 0:03:43.280 --> 0:03:46.360 Bloomberg's AI editor Seth Figemann, thank you very much. Indeed, 0:03:46.720 --> 0:03:49.360 s k Heinez plans to raise twenty nine point four 0:03:49.440 --> 0:03:53.000 billion dollars in its US market debut, a deal large 0:03:53.120 --> 0:03:55.720 enough to rank among the top five share sales of 0:03:55.760 --> 0:03:58.240 all time. The listing comes at a time when markets 0:03:58.280 --> 0:04:01.600 are hungry for memory chips and semicindut to companies are 0:04:01.720 --> 0:04:05.280 rushing for capital to expand capacity meet that demand. The 0:04:05.440 --> 0:04:08.080 BOS Peter Elstrom, who leads our coverage of Asia tech, 0:04:08.400 --> 0:04:12.120 joins us that's part of it, right, Expand memory fab 0:04:12.160 --> 0:04:15.240 capacity raise a lot of money. There's also this idea 0:04:15.360 --> 0:04:20.159 of like valuation and prestige of an ADR situation where 0:04:20.160 --> 0:04:21.960 you get exposure to American investors. 0:04:25.000 --> 0:04:25.680 Yeah, that's right. 0:04:26.040 --> 0:04:28.640 Sk Heinez is a company we don't talk about that 0:04:28.800 --> 0:04:32.200 often on the show. Obviously a leading memory chip company, 0:04:32.200 --> 0:04:34.640 but they often play second fiddle to Samsung, even in 0:04:34.680 --> 0:04:37.039 South Korea. But what we've seen from them over the 0:04:37.080 --> 0:04:39.479 past three years is really they've come from quite far 0:04:39.560 --> 0:04:44.120 behind to essentially the same market cap size as Samsung 0:04:44.160 --> 0:04:46.440 at this point. And that's largely because they've been so 0:04:46.520 --> 0:04:50.680 successful in these memory chips for AI applications that HBM 0:04:50.760 --> 0:04:54.559 chips that we've talked about before. They especially HBM chips 0:04:54.560 --> 0:04:57.599 are DRAM chips stacked on top of each other, and 0:04:57.600 --> 0:05:00.760 their technology for doing that stacking has been very good. 0:05:00.760 --> 0:05:02.200 They got Nvidia's blessing. 0:05:02.400 --> 0:05:04.960 They've been able to jump ahead of Samsung in that 0:05:05.000 --> 0:05:06.640 market quite effectively. 0:05:06.920 --> 0:05:08.360 Samsung has struggled a little bit. 0:05:08.440 --> 0:05:10.360 So now what we're seeing with this twenty nine billion 0:05:10.400 --> 0:05:14.080 dollar fundraising is really them announcing to the US, hey, 0:05:14.120 --> 0:05:16.720 that they're ready to raise this kind of capital. They 0:05:16.720 --> 0:05:19.520 want more exposure in the US and yeah, they'd like 0:05:19.600 --> 0:05:21.719 more of that pe multiple that you see with the 0:05:21.839 --> 0:05:24.880 US tech companies too. They've traded at a discount to that. 0:05:24.960 --> 0:05:27.320 If they get some of that ADR exposure, that may 0:05:27.320 --> 0:05:31.120 give them a little more opportunity to increase that valuation, and. 0:05:31.080 --> 0:05:33.640 Just on the mechanics, Peter, so it will be ADRs 0:05:33.800 --> 0:05:36.320 I believe, on the Nasdaq. So we have a timeline 0:05:36.360 --> 0:05:38.400 for this. And how much of a surprise was it 0:05:38.640 --> 0:05:40.080 that sk plowed ahead with this. 0:05:42.440 --> 0:05:45.960 Well, it's certainly they've talked about wanting to raise money 0:05:45.960 --> 0:05:48.680 in the US for a while, so I think in 0:05:48.760 --> 0:05:51.599 terms of direction, this has been expected. I think in 0:05:51.680 --> 0:05:53.680 terms of magnitude, it's a bit of a surprise though 0:05:53.680 --> 0:05:55.280 it's a lot of money. As you mentioned, it could 0:05:55.279 --> 0:05:59.600 be one of the top five offerings of its kind. Ever, 0:06:00.520 --> 0:06:03.480 they plan to start trading these ADRs on July tenth. 0:06:03.520 --> 0:06:05.240 They've talked about that. They still have a couple of 0:06:05.240 --> 0:06:07.360 steps to go before they formally do that, but that's 0:06:07.400 --> 0:06:09.440 the plan. At this point, they'd be able to raise 0:06:09.480 --> 0:06:11.400 that money and then, as you say, they really want 0:06:11.440 --> 0:06:14.320 to spend this money and trying to solve this choke 0:06:14.360 --> 0:06:16.919 point that they've had in the AI industry where memory 0:06:16.960 --> 0:06:19.520 chips are in short supply. Prices have been soaring, which 0:06:19.560 --> 0:06:22.799 is driving profits for a bunch of companies, including Micron, 0:06:22.800 --> 0:06:24.960 as we'll see after the bell today. But they want 0:06:25.000 --> 0:06:27.360 to be able to build additional capacity in South Korea. 0:06:27.560 --> 0:06:30.159 They're also expanding in the United States, where they have 0:06:30.200 --> 0:06:31.240 some facilities too. 0:06:32.320 --> 0:06:35.120 Bloomberg Tech Executive Editor details from Thank you very Much. 0:06:35.120 --> 0:06:38.680 The AI hardware story and memory in particular are set 0:06:38.720 --> 0:06:41.840 to dominate market conversation today. We've got the perfect person 0:06:41.880 --> 0:06:44.400 to speak about it. Selin Wu is a portfolio manager 0:06:44.600 --> 0:06:47.839 on Lazard Asset Management's Global Robotics and Automation team. She 0:06:47.920 --> 0:06:51.440 manages the firms tech in Ai Focus ETF, called TECHI 0:06:51.560 --> 0:06:56.240 or TEKY and actively managed ETF where sk Heinex and 0:06:56.279 --> 0:06:58.040 its career and shares are the top holding. 0:06:58.040 --> 0:06:59.880 Welcome to the program, Thanks for having me. 0:07:00.720 --> 0:07:04.000 The situation with the memory makers is fascinating. 0:07:05.040 --> 0:07:06.600 Demand is exceeding supply. 0:07:07.440 --> 0:07:11.640 Historically, that's an enviable position to be in, and the 0:07:11.680 --> 0:07:14.560 shares of those companies probably reflect that. How do you 0:07:14.600 --> 0:07:18.360 see investor appetite to be exposed to the memory trade? 0:07:18.680 --> 0:07:21.640 Yeah, I mean the shares have been fantastic, But even 0:07:21.720 --> 0:07:24.840 after that, I think memory stocks still love for a 0:07:25.040 --> 0:07:28.840 very compelling investment case. And here's why. I don't think 0:07:28.880 --> 0:07:33.560 it's necessarily just driven by temporary supply bottlenecks. Instead, I 0:07:33.600 --> 0:07:37.480 think there's something structured going on in the cycle. For example, 0:07:37.960 --> 0:07:42.080 memory's strate peagic value is rising and changing to a 0:07:42.160 --> 0:07:46.920 primary driver for performance. Think about how cloudsher providers, one 0:07:46.960 --> 0:07:51.040 of their largest clients. They're still adding more memory despite 0:07:51.040 --> 0:07:54.480 the price of inflation because this is still the most 0:07:54.520 --> 0:07:59.120 cost efficient way to maximize their system level performance. And 0:07:59.360 --> 0:08:02.600 when of the when token demand, excuse me, the token 0:08:02.640 --> 0:08:05.320 demand is rising on the back of the inference as 0:08:05.360 --> 0:08:08.440 well as agent the AI. I think the bottlenecks in 0:08:08.560 --> 0:08:12.000 terms of the capacity and bendwidth are becoming even more challenging, 0:08:12.440 --> 0:08:14.720 and this is a place where memory can come in 0:08:15.000 --> 0:08:16.880 and help navigate those challenges. 0:08:18.120 --> 0:08:22.040 Does it matter that investors will have the opportunity to 0:08:22.120 --> 0:08:25.880 have US listed shares ADRs of sk How does that 0:08:25.960 --> 0:08:28.200 kind of change the mechanics of the market for you 0:08:28.280 --> 0:08:28.680 as well? 0:08:29.120 --> 0:08:31.840 Yeah, I mean a clearly the more exposure to more 0:08:32.440 --> 0:08:36.080 diverse type of shareholder base is clearly a very positive 0:08:36.080 --> 0:08:40.439 indication for any company in the world, especially when your fundamentals, 0:08:41.200 --> 0:08:44.439 when your fundamentals are improving. So yeah, super exciting us 0:08:44.480 --> 0:08:45.960 and looking forward to the development. 0:08:46.320 --> 0:08:48.880 A couple of months ago, sitting down with Jensen Wong 0:08:49.040 --> 0:08:51.880 and I said to him, do we even need the 0:08:51.920 --> 0:08:56.239 textbooks anymore? That would tell us historically memory is cyclical, 0:08:56.520 --> 0:09:00.000 it is boom and bust. All of the evidence suggests 0:09:00.040 --> 0:09:02.320 that in the context of HBM going into data center, 0:09:02.440 --> 0:09:04.240 it just doesn't behave the same way. 0:09:04.440 --> 0:09:05.719 What is your thesis on that? 0:09:05.960 --> 0:09:08.360 I mean, I agree with that. I mean, like I said, 0:09:08.400 --> 0:09:12.360 there's something structural going on. Yes, there is certain type 0:09:12.360 --> 0:09:16.200 of cyclicality going on indoor market. But at the end 0:09:16.240 --> 0:09:18.040 of at the end of the day, what you need 0:09:18.080 --> 0:09:21.800 to remembered is structure side of booth, demand and supply, 0:09:22.280 --> 0:09:23.840 and think about the supply side. 0:09:24.120 --> 0:09:25.920 Despite the fact that all the three. 0:09:25.679 --> 0:09:29.559 Major companies are scrambling to add new supply, there are 0:09:29.600 --> 0:09:34.720 three factors that you have to remember. Number one, manufacturing intensity. 0:09:34.760 --> 0:09:39.800 Manufacturing complexity is rising. Number two, capital intensity to open 0:09:39.800 --> 0:09:42.800 a new fab is getting more expensive, and number three 0:09:43.240 --> 0:09:47.480 the difficulty of technology and migration all together pointing out 0:09:47.480 --> 0:09:51.480 that overall supply demand balance will stay in favor of 0:09:51.520 --> 0:09:53.720 the memory stocks. I mean, I think that's really great. 0:09:53.960 --> 0:09:56.640 I'd love to talk about your actively managed ETF for 0:09:56.679 --> 0:09:59.319 a moment. T e k y Techy Techy I launched 0:09:59.400 --> 0:10:02.680 last April, write sixty million in assets currently. What I 0:10:02.679 --> 0:10:05.680 find so interesting about the composition of it is something 0:10:05.679 --> 0:10:08.000 we've talked about quite recently on the show. There are 0:10:08.000 --> 0:10:11.760 the capital expenditure deployers and there are the capital expenditure 0:10:11.880 --> 0:10:14.599 recipients in the same basket when you look at the 0:10:14.640 --> 0:10:19.119 top holdings. Is that a conscious, active decision on composition. 0:10:19.679 --> 0:10:23.040 I think it's the composition from our bottom up stock 0:10:23.120 --> 0:10:26.560 selection and portfolio construction. But like you said, that's exactly 0:10:26.640 --> 0:10:30.000 where majority of the capital is invested in currently. On 0:10:30.120 --> 0:10:33.679 one hand, there are big spenders, big kip expenders that 0:10:33.840 --> 0:10:38.160 are trying to build new competitive positioning in this AI 0:10:38.160 --> 0:10:41.000 homage pace. But on the other hand, we find a 0:10:41.040 --> 0:10:44.200 lot of exciting opportunities on the companies that we see 0:10:44.280 --> 0:10:47.960 these capitals AI hardware supply chain, for example. I think 0:10:47.960 --> 0:10:51.000 that's why there's more to come. In fact, we just 0:10:51.080 --> 0:10:53.760 recently came from Asia where we sat down with a 0:10:53.840 --> 0:10:57.320 number of different companies into supply chain. Everyone is telling 0:10:57.400 --> 0:11:01.679 us how they are seeing an extended visibility from customers, 0:11:02.200 --> 0:11:07.040 how there are more conversations about ltas. Everything collectively is 0:11:07.160 --> 0:11:12.280 highlighting that demand continues to substantially exceed the supply and 0:11:12.360 --> 0:11:15.320 the AI demand outlook remains pretty robust as well. 0:11:15.480 --> 0:11:18.440 For in the ETF context, it's a very competitive market. 0:11:19.480 --> 0:11:22.480 How closely are you thinking about flows? And again, I 0:11:22.520 --> 0:11:24.280 think people give you a lot of credit Sealine for 0:11:24.679 --> 0:11:28.359 identifying sk as being just critical to the broader infrastructure 0:11:28.400 --> 0:11:31.520 build out right now, But you know what happens next 0:11:31.559 --> 0:11:34.160 to your mind? How do you see the world changing 0:11:34.920 --> 0:11:38.120 in the next I guess six months to twelve months. 0:11:38.679 --> 0:11:40.680 I mean even more than that. I mean that's why 0:11:40.679 --> 0:11:43.560 we focus on AI tech stack, and at the top 0:11:43.600 --> 0:11:46.800 of the step we have application layer and this is, 0:11:46.840 --> 0:11:50.360 for example, where we are going to expect really large 0:11:50.400 --> 0:11:54.080 opportunities from physical AI. We think physical AI is going 0:11:54.120 --> 0:11:59.200 to be multi trillion dollars long term opportunity. Essentially, AI 0:11:59.640 --> 0:12:03.400 is in because it's a primary driver for productivity gains, 0:12:03.600 --> 0:12:09.359 and historically productivity growths tend to translate to massive economic expansions. 0:12:09.720 --> 0:12:11.720 I think the same thing is going to happen for AI. 0:12:12.320 --> 0:12:15.480 All the innovation we're seeing today is eventually going to 0:12:15.559 --> 0:12:19.439 open up very significant new end markets in physical AI, 0:12:19.720 --> 0:12:23.719 such as fully autonomous transformation transportations as well as a 0:12:24.400 --> 0:12:28.800 humanoid robot. That's why we're really excited about physical AI 0:12:29.520 --> 0:12:32.320 and in the meantime, that being said, it will still 0:12:32.400 --> 0:12:34.480 take a couple of years for this to turn into 0:12:34.520 --> 0:12:37.920 actual corporate earnings. So here's what we focus on instead. 0:12:38.400 --> 0:12:43.000 Companies with vertically interiorted the manufacturing excellence that can speak 0:12:43.000 --> 0:12:46.640 to the market with scale advantage for example, or open 0:12:46.679 --> 0:12:51.320 source platform that can expedite the excelevation adoption and everything. 0:12:51.840 --> 0:12:56.280 And lastly, global technology companies that can massively benefit from 0:12:56.400 --> 0:12:59.520 the mass adoption of the application in it sulf something 0:12:59.559 --> 0:13:02.359 to watch is something we're still pretty excited about. 0:13:02.400 --> 0:13:06.040 Seline Wu from Lazard Asseid Management, first time on Bloomberg Tech. 0:13:06.080 --> 0:13:08.440 Really grateful to have you here, Thank you very much. Indeed, 0:13:08.760 --> 0:13:12.360 now coming up, Meta and Microsoft are leading the spending 0:13:12.400 --> 0:13:16.559 spree on future data center leases, adding tens of billions 0:13:16.559 --> 0:13:18.760 of dollars just last quarter alone. We'll get into the 0:13:18.760 --> 0:13:34.120 details next. This is Bloomberg Tech. Okay, today's big number, 0:13:34.240 --> 0:13:37.120 eight hundred and fifty billion dollars. That's how much the 0:13:37.120 --> 0:13:41.240 world's largest cloud computing companies have committed in future data 0:13:41.280 --> 0:13:44.800 center leases. Leading the charge in the AI spending spree, 0:13:45.040 --> 0:13:48.880 Meta and Microsoft both committing tens of billions more just 0:13:48.920 --> 0:13:52.480 last quarter. Bloombo's Brodie Ford has an astonishing story and 0:13:52.520 --> 0:13:53.040 a hell of a. 0:13:53.120 --> 0:13:54.640 Chart to show us. 0:13:54.920 --> 0:13:57.280 What I find interesting about this is that these are 0:13:57.400 --> 0:14:01.880 future leases, essentially leases planned or announced separate from what 0:14:02.040 --> 0:14:05.760 is already in operation, what they are already contractually committed to. 0:14:06.920 --> 0:14:09.400 It's a really good point, right, So we're talking about 0:14:09.480 --> 0:14:12.240 eight hundred and fifty billion on data centers, which in 0:14:12.280 --> 0:14:16.000 many cases aren't even built yet. It's they are working 0:14:16.080 --> 0:14:18.440 through their current projects they have, and these are the 0:14:18.440 --> 0:14:22.400 ones that they've committed to leasing once they're operational. And 0:14:22.440 --> 0:14:25.440 so a lot of people wonder about what's the long 0:14:25.520 --> 0:14:29.600 term spending trajectory on data centers? How long does this last? Well, 0:14:29.640 --> 0:14:32.400 there's eight hundred and fifty billion dollars of leases that 0:14:32.480 --> 0:14:34.760 are going to be starting in the coming years. 0:14:35.880 --> 0:14:37.640 So when I was reading the story, I was trying 0:14:37.680 --> 0:14:41.600 to understand the future costs. It's an amazing chart, by 0:14:41.600 --> 0:14:43.440 the way, we're just showing on the screen right now, 0:14:43.880 --> 0:14:47.440 like the level of capital and it's the commitment grows. 0:14:47.600 --> 0:14:53.200 But where did their current financial healths and balance sheet 0:14:53.320 --> 0:14:55.680 stand relative to those future commitments. 0:14:55.800 --> 0:14:56.920 Is there a concern about that? 0:14:58.600 --> 0:15:00.840 I think there's a lot of concern that seem to 0:15:01.040 --> 0:15:05.600 always ebb and flow around the ROI of all this spending, right. 0:15:05.640 --> 0:15:07.600 I mean, it feels like every quarter or two the 0:15:07.720 --> 0:15:10.880 general sentiment flips about whether the market's worried or not, 0:15:11.360 --> 0:15:15.520 but every hyperscaler seems to agree that they must spend 0:15:15.600 --> 0:15:18.520 an incredible amount to catch up an AI to have 0:15:18.640 --> 0:15:22.280 enough data center capacity. I think about Microsoft, who through 0:15:22.320 --> 0:15:25.360 a lot of twenty twenty five said we're nervous, we're 0:15:25.400 --> 0:15:27.560 actually going to put on ice a lot of these leases, 0:15:27.680 --> 0:15:30.280 and they've come to regret it pretty deeply. And I 0:15:30.320 --> 0:15:32.800 think a lot of the biggest tech companies, as we 0:15:32.840 --> 0:15:35.360 can see in this chart, are hoping to avoid that 0:15:35.440 --> 0:15:36.320 kind of scenario. 0:15:37.120 --> 0:15:39.760 Microsoft added forty one billion dollars in commitments to a 0:15:39.800 --> 0:15:42.640 total of almost one hundred and ninety seven billion dollars. 0:15:42.800 --> 0:15:44.560 Compare and can trust really quick Oracle. 0:15:45.920 --> 0:15:49.680 Oracle's very interesting because they signed a ton of leases, 0:15:49.840 --> 0:15:52.560 largely for open AI, and they've been flat for the 0:15:52.640 --> 0:15:55.880 last two quarters, which says to us that they're digesting 0:15:55.920 --> 0:15:58.880 a lot of capacity, contrasted with somebody like Meta, who 0:15:58.960 --> 0:16:00.720 is just signing as much much as they seem to 0:16:00.760 --> 0:16:01.440 be able. 0:16:01.160 --> 0:16:04.840 To Bloombanks, Brady Ford, thank you very much. Just get 0:16:04.880 --> 0:16:06.880 out to New York where bloom Bek Jihara and and 0:16:07.040 --> 0:16:07.760 is standing by Hi. 0:16:07.760 --> 0:16:11.440 Yahara, Hi ed It's time now for talking tech. First 0:16:11.480 --> 0:16:15.000 up Softbanks. Masayoshi Sun says he has no plans to 0:16:15.040 --> 0:16:18.240 retire anytime soon as he looks to capitalize on the 0:16:18.400 --> 0:16:22.560 AI boom. Speaking at an annual shareholder meeting, the sixty 0:16:22.680 --> 0:16:26.160 eight year old visionary fired back at skeptics, saying, quote, 0:16:26.320 --> 0:16:29.480 the AI revolution has only just begun. Calling it a 0:16:29.520 --> 0:16:33.800 bubble is an insult plus and the latest showdown between 0:16:33.880 --> 0:16:37.080 Chinese big tech and national security. Ali Baba has sued 0:16:37.120 --> 0:16:41.320 the US Defense Department demanding removal from its blacklists. The 0:16:41.360 --> 0:16:46.360 e commerce giant called a designation arbitrary and unjustified, following 0:16:46.360 --> 0:16:49.480 a US crackdown earlier this month that accused several of 0:16:49.560 --> 0:16:54.760 China's top companies of aiding Beijing's military, and TikTok parent 0:16:54.840 --> 0:16:57.960 by Dance is in early talks to secure a record 0:16:58.160 --> 0:17:01.280 twenty billion dollars global loan. Well the exact use of 0:17:01.320 --> 0:17:05.440 the proceeds remains unclear. The massive borrowing push comes as 0:17:05.480 --> 0:17:09.360 the company wis boosting data center and AI capital. 0:17:08.920 --> 0:17:11.720 Spending ed Thank you very much, you Hira. 0:17:11.880 --> 0:17:15.159 Now coming up on the program defense startup, Hadrian discusses 0:17:15.240 --> 0:17:18.359 new funding at a seven point five billion dollar valuation 0:17:18.440 --> 0:17:22.160 that would be four times its past or latest value. 0:17:22.160 --> 0:17:24.000 We're going to discuss that next. This is Bloomberg Tech. 0:17:38.520 --> 0:17:42.679 Can AI replace human effort in the realm of scientific discovery? 0:17:42.840 --> 0:17:44.480 In the latest episode of The Circuit. 0:17:44.240 --> 0:17:47.639 Bloomberg's Emily Chang spoke to Jennifer Dowdner, the inventor of 0:17:47.680 --> 0:17:52.639 the groundbreaking crispher gene editing technology, about the future of biology. 0:17:52.640 --> 0:17:53.119 Listened to this. 0:17:54.000 --> 0:17:57.199 Biology is complex. We're not going to be able to 0:17:57.320 --> 0:18:00.560 simulate our way to an understanding of the human. We're 0:18:00.560 --> 0:18:03.239 not going to be able to avoid the need for 0:18:03.359 --> 0:18:06.399 certain types of testing. I do think there are opportunities 0:18:06.400 --> 0:18:10.760 to increase the efficiency in which we make discoveries about 0:18:11.040 --> 0:18:14.400 the way our bodies work and the way they interact 0:18:14.400 --> 0:18:16.560 with drugs that will be effective, and I think AI 0:18:16.680 --> 0:18:18.600 will be helpful there. But it's going to come down 0:18:18.640 --> 0:18:21.439 to training models on the right kind of data, and 0:18:21.480 --> 0:18:25.479 a big need for better and more data for training 0:18:25.480 --> 0:18:26.680 models if we want to achieve that. 0:18:27.160 --> 0:18:30.359 An Open AI executive recently suggested that if a discovery 0:18:30.359 --> 0:18:33.560 happens on chat GPT, let's say a drug discovery that 0:18:33.680 --> 0:18:35.640 open AI should get a cut of sales. 0:18:36.080 --> 0:18:36.760 What do you think of that? 0:18:37.920 --> 0:18:42.400 Good luck expand how are chatbots going to change drug discovery? 0:18:42.440 --> 0:18:43.840 I'm not sure the answer to that yet. 0:18:43.880 --> 0:18:44.320 I don't know. 0:18:44.440 --> 0:18:47.040 Lots of people are, of course, very very hopeful, so 0:18:47.040 --> 0:18:50.679 I'm very helpeful about it. But I think that innovation 0:18:51.040 --> 0:18:54.280 is still really in the domain of human beings right now. 0:18:54.520 --> 0:18:58.840 I'm not seeing chatbots in our own experience innovating. They 0:18:58.840 --> 0:19:01.320 can be helpful with summer rising data, that can be 0:19:01.320 --> 0:19:04.240 helpful in writing reports and things of that nature, but 0:19:04.359 --> 0:19:07.320 I'm not seeing chatbots coming up with a brand new 0:19:07.359 --> 0:19:09.600 idea for something that nobody else ever thought of. 0:19:09.880 --> 0:19:12.000 So you're saying AI can't innovate. 0:19:12.560 --> 0:19:14.480 I don't know if it can't innovate, I just don't 0:19:14.520 --> 0:19:15.440 think it is right now. 0:19:15.520 --> 0:19:17.280 What about after the AGI moment? 0:19:17.640 --> 0:19:20.320 Well, I never say never, so maybe that'll happen, but 0:19:20.359 --> 0:19:21.399 I'm not holding my breath. 0:19:21.760 --> 0:19:25.119 That was Bloomberg's Emily Chang speaking with Jennifer Dowdner, Nobel 0:19:25.160 --> 0:19:29.040 Laureate and founder of Innovative Genomics Institute. What's the four 0:19:29.040 --> 0:19:32.359 episode on Bloomberg Television tonight at six pm Eastern or 0:19:32.400 --> 0:19:38.280 on Bloomberg Originals at eight pm Eastern time. Defense manufacturing 0:19:38.359 --> 0:19:40.399 start up Hadrian has been in talks to more than 0:19:40.520 --> 0:19:44.640 quadruple its valuation in a new funding round, according to sources, 0:19:44.640 --> 0:19:47.800 Bloomberg's venture reporter of BEC and Torrents and I teamed 0:19:47.880 --> 0:19:48.720 up on this one. 0:19:48.960 --> 0:19:50.160 So let's go through the basics. 0:19:50.240 --> 0:19:53.199 Right, these are talks, they're at a certain stage, but 0:19:53.240 --> 0:19:55.320 they only raise money right at the beginning of this 0:19:55.400 --> 0:19:57.600 year at a certain valuation. What's the jump and what 0:19:57.640 --> 0:19:58.199 do we need to know? 0:19:58.720 --> 0:19:59.399 That's correct ed? 0:19:59.520 --> 0:20:02.280 So Hadrian raise money at a one point six billion 0:20:02.320 --> 0:20:05.679 dollar valuation just in January. They've now had conversations to 0:20:05.720 --> 0:20:08.040 raise up to a billion dollars in this latest round. 0:20:08.119 --> 0:20:10.400 That would more than quadruple their valuation, as you said, 0:20:10.400 --> 0:20:13.359 to seven point five billion dollars. The investor demand on 0:20:13.400 --> 0:20:16.560 this one is pretty significant. We reported that several existing 0:20:16.600 --> 0:20:20.680 investors are in talks for this one. We should note 0:20:20.680 --> 0:20:23.119 that when we reach out to comment, Hadrian said that 0:20:23.160 --> 0:20:26.640 the information provider was incorrect decline to provide further context. 0:20:26.960 --> 0:20:28.400 That said, we sitten by our reporting. 0:20:28.960 --> 0:20:31.919 This is an area that's extremely hot with investors right now, 0:20:32.080 --> 0:20:37.320 defense tech broadly. Hadrian makes facilities to speed up domestic 0:20:37.359 --> 0:20:40.080 manufacturing and that's been hugely of interest to a lot 0:20:40.080 --> 0:20:41.119 of folks and Silicon Valley. 0:20:42.680 --> 0:20:43.320 Thank you for that. 0:20:43.359 --> 0:20:45.879 I would also add that s fokes first and Hadrium 0:20:45.920 --> 0:20:49.000 said the reporting was incorrect but declined to add further context. 0:20:49.080 --> 0:20:50.760 I would have done that for you. Thank you very much. 0:20:51.040 --> 0:20:54.440 There is some reporting that we did about where they. 0:20:54.680 --> 0:20:57.960 Approached debt or a lack of approach to debt. 0:20:58.040 --> 0:21:03.280 Just explain that bites so building large automated factories often 0:21:03.359 --> 0:21:06.160 means that companies need to take on debt to take 0:21:06.640 --> 0:21:11.880 you know, on the risks of building these large facilities. 0:21:11.920 --> 0:21:15.480 Hadrian currently has four factories, the latest which operate, which 0:21:15.520 --> 0:21:18.560 opened in March off the back of a two point 0:21:18.560 --> 0:21:21.960 four billion dollar contract with the US Navy. So this amount, 0:21:22.280 --> 0:21:25.080 this up to one billion dollars is just equity that 0:21:25.119 --> 0:21:27.119 they would raise. It doesn't include any debt that they 0:21:27.200 --> 0:21:28.120 may be discussing. 0:21:28.920 --> 0:21:31.159 Rebecca, you did a great job explaining, you know what 0:21:31.200 --> 0:21:37.120 Hadrian does. It's basically an amazing factory or contract manufacturer. Generally, 0:21:37.160 --> 0:21:40.960 what is bench capital attitude right now to defense technology 0:21:40.960 --> 0:21:45.200