WEBVTT - SpaceX IPO Multiple Times Oversubscribed 0:00:02.560 --> 0:00:13.560 Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is alive 0:00:13.600 --> 0:00:17.400 from coast to coast with Caroline Hide in New York 0:00:17.720 --> 0:00:20.040 and ever though in sentences. 0:00:19.600 --> 0:00:24.560 Go this is Bloomberg Tech coming up. Wall Street can't 0:00:24.560 --> 0:00:28.280 get enough of SpaceX. With demand from big institutional investors 0:00:28.440 --> 0:00:32.480 and the biggest IPO in history, way over subscribed, class. 0:00:32.159 --> 0:00:35.519 Google backstops and thropping data centers, a Silicon valley raises 0:00:35.600 --> 0:00:38.839 to build AI infrastructure with ever more intertwined deals. 0:00:38.840 --> 0:00:41.400 An Oracle reports after the closing bell, it's a race 0:00:41.520 --> 0:00:44.720 between building data centers and booking AI cloud. 0:00:44.479 --> 0:00:49.280 Revenues, AI AI AI and space SpaceX's over subscribed IPO 0:00:49.440 --> 0:00:52.840 is where we have to start ed because the geographical 0:00:53.200 --> 0:00:54.920 reach of the level of demand. 0:00:55.280 --> 0:00:58.080 We've been mesmerized by this record breaking. 0:00:57.920 --> 0:00:59.960 Yeah, it's out of this world. I don't apologize for 0:01:00.080 --> 0:01:01.960 that for one bit. The state of players this right. 0:01:02.160 --> 0:01:05.319 The order book for institution investors closes four pm Eastern today, 0:01:05.560 --> 0:01:09.200 and as we've reported, there are several long only asset 0:01:09.280 --> 0:01:11.959 managers basically that want ten billion dollars worth of shares. 0:01:12.000 --> 0:01:13.399 It's seventy five billion dollars worth. 0:01:13.520 --> 0:01:15.760 So somebody is going to miss out. Now the retail 0:01:15.800 --> 0:01:19.360 investor can still place orders I think through Thursday on 0:01:19.400 --> 0:01:22.360 whatever platforms are available, but they're not guaranteed to get 0:01:22.360 --> 0:01:24.559 hold of those shares either. So that's the state of play. 0:01:24.640 --> 0:01:26.679 And believe it or not, there is a roadshow happening 0:01:26.680 --> 0:01:27.479 in the background, and. 0:01:27.440 --> 0:01:29.800 We're learning ever more on that roadshow. That's the entire point. 0:01:29.840 --> 0:01:32.840 We're understanding the transparency, the business model. We're learning much 0:01:32.840 --> 0:01:34.840 about those orbital data centers. 0:01:34.920 --> 0:01:36.640 Yeah, I think the focus in the pitch has still 0:01:36.640 --> 0:01:39.360 been let us explain orbital data centers. That brings us 0:01:39.400 --> 0:01:42.479 to today's big number, two hundred and fifty billion. That's 0:01:42.520 --> 0:01:46.120 a total amount of SpaceX IPO orders we've reported this morning, 0:01:46.200 --> 0:01:50.120 one to five billion of which is coming from Saudi Qate, 0:01:50.320 --> 0:01:53.760 other Middle East funds, sovereign funds. That's according to sources. 0:01:54.000 --> 0:01:57.000 That's the absolute latest. Joining us now to talk all 0:01:57.040 --> 0:02:01.240 things SpaceX. It's IPO. Also the general landscape Peter Singlehurst, 0:02:01.440 --> 0:02:03.520 head of private Companies are Bailey giff and we just know, 0:02:03.920 --> 0:02:06.200 you know SpaceX is a really important investment for you 0:02:06.320 --> 0:02:10.760 guys prior to the offering. Let's start there. You know, 0:02:10.919 --> 0:02:15.200 what does this the biggest IPO in history represent to 0:02:15.320 --> 0:02:19.080 you and to the firm and to I guess support 0:02:19.120 --> 0:02:22.000 the thesis when you first made the investment way back when. 0:02:24.320 --> 0:02:27.400 I think that the SpaceX IPO needs to be seen 0:02:27.680 --> 0:02:30.840 as the culmination of a trend which has been playing 0:02:30.880 --> 0:02:34.840 out now for fifteen years or longer, of companies staying 0:02:35.000 --> 0:02:37.400 private for longer. And this is something that we started 0:02:37.440 --> 0:02:40.880 to see in twenty twelve when we first started investing 0:02:41.440 --> 0:02:45.440 in private companies. Now, we didn't think the companies would 0:02:45.520 --> 0:02:48.600 get this big and stay private this long, But here 0:02:48.639 --> 0:02:52.040 we are with, you know, SpaceX going public at something 0:02:52.080 --> 0:02:55.680 like a one point eight trillion dollar valuation. That's nine 0:02:55.800 --> 0:02:59.600 hundred times larger and more valuable than Tesla was when 0:02:59.600 --> 0:03:02.720 it went public in twenty twelve. So, on the one hand, 0:03:02.760 --> 0:03:05.239 this is a story of a truly exceptional company which 0:03:05.320 --> 0:03:09.800 has compounded its growth at a very high rate. On 0:03:09.840 --> 0:03:11.959 the other hand, it's a story of a bigger structural 0:03:12.000 --> 0:03:14.480 trend of companies staying private longer and more and more 0:03:14.480 --> 0:03:17.000 return to accruing within the high growth private. 0:03:16.720 --> 0:03:18.240 Market and peter to that end. 0:03:18.320 --> 0:03:21.240 When you think about Tesla after it's gone public, it 0:03:21.280 --> 0:03:25.000 was a volatile ride, but it's twenty five thousand percent 0:03:25.160 --> 0:03:29.120 higher than when it listed, And so will we see 0:03:29.280 --> 0:03:31.680 a level of returns do you think in the public 0:03:31.720 --> 0:03:34.360 market or does that have to be in some ways 0:03:34.400 --> 0:03:38.160 pushed against the meat and bones of returns going to 0:03:38.200 --> 0:03:40.000 have happened to private investors. 0:03:40.960 --> 0:03:43.680 I think it's mathematically it's very hard to see how 0:03:43.760 --> 0:03:47.360 you could see SpaceX delivering the same kind of returns 0:03:47.400 --> 0:03:50.000 as a public company as Tesla did. But I think 0:03:50.000 --> 0:03:53.800 what this speaks to is a requirement for investors to 0:03:53.920 --> 0:03:58.040 have exposure to growth in both the private and the 0:03:58.080 --> 0:04:01.840 public markets. Has been set up to almost divide these 0:04:02.200 --> 0:04:04.400 things and say there's kind of private growth and there's 0:04:04.440 --> 0:04:07.320 public growth, and these things are different, and we've taken 0:04:07.360 --> 0:04:09.640 a different approach. We've sort of taken the view that actually, 0:04:09.640 --> 0:04:11.400 if you want to do growth equity investing, you want 0:04:11.440 --> 0:04:13.400 to do it properly, you have to do it in 0:04:13.440 --> 0:04:15.680 the private markets, and you have to do it in 0:04:15.720 --> 0:04:19.440 the public markets. And what our clients are and beneficiaries 0:04:19.440 --> 0:04:21.920 who are predominantly pension funds, what they need and what 0:04:21.960 --> 0:04:24.480 they ask from us is that we give them exposures 0:04:24.520 --> 0:04:27.479 to the world's best growth stage companies starting in the 0:04:27.480 --> 0:04:30.000 private markets. Earning the returns that we can generate there, 0:04:30.400 --> 0:04:32.360 and then also only those in the public markets from 0:04:32.400 --> 0:04:34.240 within our public funds to make sure that they're still 0:04:34.320 --> 0:04:37.919 capturing that growth even once companies transition into the public markets. 0:04:38.120 --> 0:04:40.599 Peter, I think it's important to pose the question why 0:04:40.720 --> 0:04:43.640 is SpaceX going public? And when Elon Musk spoke to 0:04:43.720 --> 0:04:46.680 Jamie Diamond, he eventually got to the answer, which is 0:04:47.200 --> 0:04:51.039 they need capital for this growth phase. But what we 0:04:51.080 --> 0:04:53.840 are seeing outside of just this fixation on IPOs is 0:04:53.880 --> 0:04:58.200 a race for capital through equity. How comfortable do you 0:04:58.360 --> 0:05:02.359 feel as a firm at Bailey gifed, whatever mechanism it 0:05:02.480 --> 0:05:05.520 is raising money at that volume, but it basically then 0:05:05.600 --> 0:05:09.160 goes directly into capital expenditure. That's what's happening here. 0:05:11.600 --> 0:05:13.800 You want to invest in companies that are able to 0:05:13.839 --> 0:05:16.560 deploy capital or high rates of return. So I don't 0:05:16.560 --> 0:05:19.679 think there's anything wrong, per se in investing in capital 0:05:19.720 --> 0:05:22.520 intensive businesses. In fact, what you want as a company 0:05:22.520 --> 0:05:25.200 that can deploy large amounts of capital, but where you 0:05:25.240 --> 0:05:27.800 can earn high returns on that capital. And ultimately that's 0:05:27.839 --> 0:05:30.320 what separates a good business from a bad business. It's 0:05:30.360 --> 0:05:33.040 return on equity. And so when we're looking at a company, 0:05:33.040 --> 0:05:35.800 whether it's SpaceX or Anthropic, what any other company that 0:05:35.839 --> 0:05:38.760 we invest in privately or publicly. Ultimately, what we're asking 0:05:38.839 --> 0:05:41.880 is how do you get to high returns on equity? 0:05:41.920 --> 0:05:43.800 And it's building those thesis that then leads us to 0:05:43.880 --> 0:05:47.560 invest in companies. And in the case of SpaceX, increasingly 0:05:47.560 --> 0:05:50.279 that thesis is going to have to rely on AI. 0:05:50.920 --> 0:05:53.960 They've shown that they can invest capital or high rates 0:05:53.960 --> 0:05:57.440 of return in rockets in starlink, and of course the 0:05:57.480 --> 0:05:59.960 next leg of that is going to be in AI 0:06:00.040 --> 0:06:02.960 data center build out, quite possibly in space but that's. 0:06:02.800 --> 0:06:06.400 Where it becomes so fascinating, particularly Peter for Bailey Gifford, 0:06:06.440 --> 0:06:09.760 which in the private markets backspace X on a thesis 0:06:09.760 --> 0:06:12.840 of SPACE, Backtindthropic on a thesis of AI, and now 0:06:12.880 --> 0:06:16.200 they're all overlapping in terms of business models. What is 0:06:16.240 --> 0:06:19.520 your perspective of commoditization or a winner takes all or 0:06:19.640 --> 0:06:20.960 is there room for all. 0:06:20.839 --> 0:06:23.240 Of these giant AI players? Do we winning in. 0:06:23.240 --> 0:06:25.440 The technology as well as perhaps in the public markets. 0:06:26.800 --> 0:06:29.640 I think what your question gets to is this very 0:06:29.680 --> 0:06:35.560 important question of where does value accrue in the AI stack? Now, 0:06:35.600 --> 0:06:37.320 hopefully lots of value is going to accrue to the 0:06:37.440 --> 0:06:41.040 end customers. That has to happen. Historically, we've seen value 0:06:41.040 --> 0:06:44.560 accruing to the chip manufacturers, initially with Nvidio, but now 0:06:44.560 --> 0:06:47.480 increasing league so memory manufacturers. But I think what we're 0:06:47.480 --> 0:06:49.320 also starting to see is value a crew at the 0:06:49.320 --> 0:06:53.520 foundational model level. And I suppose with the Grock acquisition, 0:06:54.440 --> 0:06:56.840 SpaceX is making a bet not just on the foundational 0:06:56.839 --> 0:06:58.960 model level, but also on the infrastructure level. And I 0:06:59.000 --> 0:07:01.320 think what we're seeing with the deal that they did 0:07:01.360 --> 0:07:03.640 recently with Anthropic is that they have options in terms 0:07:03.680 --> 0:07:07.039 of how they can monetize in the AI transition, both 0:07:07.200 --> 0:07:11.600 through their own models, but also importantly through the infrastructure itself. 0:07:11.680 --> 0:07:14.160 I have so many questions about this. You know, let's 0:07:14.160 --> 0:07:17.200 be honest. The hedge that SpaceX has put in place 0:07:17.240 --> 0:07:19.840 in the interim is to become a hyperscaler and sell 0:07:19.880 --> 0:07:23.760 compute played a blinder with that. We got the design, 0:07:24.080 --> 0:07:26.600 or at least the renderings of Orbital Data Center. I 0:07:26.640 --> 0:07:27.800 think the team are going to put them up on 0:07:27.800 --> 0:07:30.840 the screen now in that presentation that Elon must make, 0:07:31.160 --> 0:07:34.000 like there's the body, there's the solar arrays, there's the radiator, 0:07:35.160 --> 0:07:38.320 which part of the thesis, Peter is most important to you, right, 0:07:38.680 --> 0:07:42.080 It is a long way from the tam of twenty 0:07:42.120 --> 0:07:45.160 six point five trillion that they're basically packaging it as 0:07:45.280 --> 0:07:49.120 enterprise AI and in the interim, this plan for orbital 0:07:49.200 --> 0:07:51.320 data center like it needs to work. That's what they're 0:07:51.360 --> 0:07:52.560 telling people on the road show. 0:07:53.800 --> 0:07:58.080 So there's absolutely no question that the orbital Data center 0:07:59.240 --> 0:08:04.840 strategic that they're making widens the range of outcomes for SpaceX. 0:08:05.240 --> 0:08:07.320 On the one hand, if it works, it increases the 0:08:07.320 --> 0:08:10.080 potential upsides for the business. On the other hand, if 0:08:10.120 --> 0:08:12.240 this doesn't work, it's going to increase the downside for 0:08:12.320 --> 0:08:15.280 this company. And investing ultimately is about ranges of outcomes. 0:08:15.320 --> 0:08:18.320 It's about probabilities, and it's about payoffs in those range 0:08:18.320 --> 0:08:21.800 of outcomes. What we've seen with SpaceX over the years 0:08:21.880 --> 0:08:26.680 is that they have continuously tested and validated a series 0:08:26.680 --> 0:08:30.280 of outlandish hypotheses. The very notion of the business starting 0:08:30.280 --> 0:08:32.640 off as a private rocket company was a self an 0:08:32.640 --> 0:08:36.040 outlandish hypothesis that they've validated. Then the idea that you 0:08:36.080 --> 0:08:39.600 could have reusable rockets was also an outlanded hypothesis, and 0:08:39.640 --> 0:08:43.480 they validated it. They did the same with starlink with 0:08:44.600 --> 0:08:47.840 satellite based broadband. They've done the same with rockets on 0:08:47.920 --> 0:08:51.599 the scale of Starship and the orbital data centillate is 0:08:51.640 --> 0:08:55.320 that is the next hypothesis that they are seeking to validate. 0:08:55.480 --> 0:08:58.160 But everybody should be totally aware of the risks that 0:08:58.200 --> 0:09:00.679 are involved in this. It is unproven. In the event 0:09:00.720 --> 0:09:03.640 that they prove it, the payoffs will be large. But 0:09:03.679 --> 0:09:05.480 as we've already touched on the amount of cattle that 0:09:05.559 --> 0:09:07.320 is going into this means that the event that they 0:09:07.320 --> 0:09:09.800 don't validate it, it's going to increase the scope of 0:09:09.840 --> 0:09:12.120 downside in the investments as well, and investors just need 0:09:12.160 --> 0:09:14.560 to understand the range of outcomes and the payoffs that 0:09:14.600 --> 0:09:15.120 go with that. 0:09:15.559 --> 0:09:18.880 Can I ask about payoffs, Peter, because I don't want 0:09:18.920 --> 0:09:20.760 to go into the granularity of how much SpaceX E 0:09:20.760 --> 0:09:22.280 suppose you have, et cetera, but how long do you 0:09:22.280 --> 0:09:25.320 think you'll hold it and how much do you think 0:09:25.600 --> 0:09:27.719 it's a warrior or an anxiety that all these other 0:09:27.720 --> 0:09:30.880 big public companies are selling equity into this market at 0:09:30.880 --> 0:09:33.880 the same time Alphabet trying to fund its own capex 0:09:34.000 --> 0:09:36.199 in the equity market, Meta might be doing as well, 0:09:36.240 --> 0:09:38.240 And does that take oxygen out the room. 0:09:39.800 --> 0:09:42.280 I think that's probably part of the thinking that's going 0:09:42.320 --> 0:09:44.880 on for these different companies trying to raise these large 0:09:44.880 --> 0:09:46.760 amounts of cattle. They're sort of trying to soak up 0:09:46.760 --> 0:09:51.680 what available castle there is. But to your first question, 0:09:52.920 --> 0:09:54.839 different funds within Bailey Gifford are going to be in 0:09:54.920 --> 0:09:58.400 very different positions. For those funds that have owned SpaceX 0:09:58.440 --> 0:10:01.200 since twenty eighteen, since it was a thirty billion dollar company, 0:10:01.320 --> 0:10:05.960 those funds have very very large exposure, large positions in SpaceX. 0:10:06.240 --> 0:10:08.600 Now it might make sense post lock up for those 0:10:08.600 --> 0:10:11.040 funds to start selling down, even if they want to 0:10:11.080 --> 0:10:16.160 maintain a meaningful exposure, because ultimately, we are beholding to 0:10:16.160 --> 0:10:18.560 our clients and we have to provide them with a 0:10:18.640 --> 0:10:21.560 level of diversification within their funds. And then, of course 0:10:21.600 --> 0:10:24.800 funds that don't own it, funds that are solely public funds, 0:10:24.960 --> 0:10:26.720 they then faced with a question of whether to buy 0:10:26.800 --> 0:10:29.760 it for their funds. So it might well be that 0:10:29.800 --> 0:10:32.840 you see different funds within Bailey Gifford doing different things 0:10:32.960 --> 0:10:34.800 over the coming months, and that will be a function 0:10:34.840 --> 0:10:37.360 of the history and the portfolio context. I think there's 0:10:37.520 --> 0:10:41.440 universe agreement that SpaceX has been an exceptional company. The 0:10:41.480 --> 0:10:44.240 real question from here is what is the right price 0:10:44.679 --> 0:10:48.360 and what is the right position size in SpaceX. 0:10:48.320 --> 0:10:51.360 From thirty billion to potentially one point eight trillion this 0:10:51.400 --> 0:10:54.600 week Peter Senglehurst A Bailey Gifford A joy to have 0:10:54.720 --> 0:10:57.880 you on about all things SpaceX and the IPO landscape. 0:10:57.960 --> 0:11:01.400 More broadly, let's get though, also to political tensions. They 0:11:01.440 --> 0:11:04.720 are continuing to whipsaw markets. We are down a percentage 0:11:04.760 --> 0:11:06.640 point again on the NASA one hundred s and p 0:11:06.800 --> 0:11:09.880 is under pressure. You're seeing a really hardware of by 0:11:09.880 --> 0:11:12.480 two percent if you're looking at the semiconductor index. President 0:11:12.480 --> 0:11:15.680 Trump is saying that Iran would pay the price. 0:11:15.840 --> 0:11:17.400 For delaying peace negotiations. 0:11:17.480 --> 0:11:19.280 Let's get you up to speakably the most Tyler Kendall, 0:11:19.320 --> 0:11:20.400 the latest sim of the White House. 0:11:20.440 --> 0:11:22.960 What do we need to know, hey, Caroline, Well, at 0:11:23.000 --> 0:11:26.320 this point, President Trump is renewing his threat, really just 0:11:26.480 --> 0:11:30.040 underscoring that this White House has mounting frustration with the 0:11:30.120 --> 0:11:33.760 ongoing negotiations, as the US has repeatedly maintained that they 0:11:33.760 --> 0:11:38.400 were trying to prioritize a diplomatic solution to end the conflict. Now, 0:11:38.400 --> 0:11:42.280 President Trump's remarks aren't totally clear if this means that 0:11:42.280 --> 0:11:45.400 we're going to see an end to the ceasefire agreement 0:11:45.520 --> 0:11:47.920 after we saw the worst flare up in fighting between 0:11:48.080 --> 0:11:51.200 the sides just overnight with both the US and Iran 0:11:51.280 --> 0:11:56.320 exchanging strikes after an American military helicopter was shot down. 0:11:56.600 --> 0:12:00.920 After the strike from the US then on Iranian military assets, 0:12:01.000 --> 0:12:04.200 we saw Iran put forward some strikes and attempts to 0:12:04.280 --> 0:12:07.720 hit American military assets. It's really been escalating from here. 0:12:07.760 --> 0:12:11.000 But our own analysts at Bloomberg Economics, perhaps this is 0:12:11.040 --> 0:12:14.079 a bid to escalate in a bid to de escalate. 0:12:14.360 --> 0:12:17.480 In one positive sign for the negotiation front, Irani and 0:12:17.520 --> 0:12:20.400 state media reported within the last hour that a Katari 0:12:20.559 --> 0:12:23.520 delegation has landed in Tehran in a bid to keep 0:12:23.559 --> 0:12:26.240 diplomacy on track. But Ed and Caroline, I want to 0:12:26.360 --> 0:12:29.440 highlight this renewed risk that we are seeing moments ago 0:12:29.760 --> 0:12:33.040 flashing across the Bloomberg terminal. India is now condemning an 0:12:33.040 --> 0:12:37.320 apparent attack on a commercial vessel near the Strait of 0:12:37.320 --> 0:12:40.600 her Moves off the coast of Oman. So definitely still 0:12:40.760 --> 0:12:43.760 very high intentions contributing to the situation that we're seeing 0:12:43.840 --> 0:12:44.400 on the ground. 0:12:45.559 --> 0:12:48.040 Goodvoks Tyler Kendall, Thank you very much. So coming up, 0:12:48.080 --> 0:12:51.400 Google steps up to backstop a massive thirty five billion 0:12:51.440 --> 0:12:54.360 dollar financing deal from Propic. We had the details. Next, 0:12:54.480 --> 0:13:03.240 this is Bloomberg Tech. Super Micro is targeting public markets 0:13:03.240 --> 0:13:05.920 for a massive seven billion dollar equity raise. Sales to 0:13:06.000 --> 0:13:09.800 super Micro servers fitted with video chips have surged for 0:13:09.960 --> 0:13:13.160 AI workloads. The server manufacturer is moving quickly to fund 0:13:13.200 --> 0:13:16.560 a staggering thirty nine billion dollars in orders, using the 0:13:16.600 --> 0:13:19.800 fresh cash injection to pay for the equipment needed to 0:13:19.840 --> 0:13:23.760 make the servers. That's greasing the wheels. Google is backstopping 0:13:23.800 --> 0:13:27.319 a massive thirty five billion dollar financing deal from Fropic, 0:13:27.600 --> 0:13:30.480 The creator of Claud, is leasing AI chips across five 0:13:30.760 --> 0:13:33.760 different data centers with help from the long established tech 0:13:33.800 --> 0:13:37.800 giant Doomscott Carpenter joins us now to break down the mechanics, like, 0:13:38.280 --> 0:13:41.839 you know, to our audience, what does backstopping mean? I 0:13:41.880 --> 0:13:44.559 think what we're saying is guaranteeing the funds for it 0:13:44.920 --> 0:13:46.360 and the event that something goes wrong. 0:13:46.440 --> 0:13:50.360 But but just go with the basics, right, So, first 0:13:50.400 --> 0:13:55.240 of all, Broadcom is providing a huge guarantee on the 0:13:55.440 --> 0:14:00.439 chips themselves, as the biggest part of the thirty five years. Yes, 0:14:00.480 --> 0:14:03.120 these are Google's TPUs that are going to be involved, 0:14:03.600 --> 0:14:06.319 So Broadcom is backstopping the debt itself. 0:14:07.080 --> 0:14:09.000 Now, the chips, when. 0:14:08.880 --> 0:14:12.440 They are delivered and they have to be manufactured, are 0:14:12.520 --> 0:14:15.400 going to be used in these five data centers that 0:14:15.600 --> 0:14:20.080 we identify in the story. The leases on those five 0:14:20.200 --> 0:14:24.200 data centers are backstopped by Google. So you could think 0:14:24.200 --> 0:14:28.560 of it as two different forms of guarantees being involved 0:14:29.040 --> 0:14:32.240 in this to put together this deal. There's the Broadcom 0:14:32.280 --> 0:14:34.080 one and there's the Google ones underneath. 0:14:34.280 --> 0:14:37.880 So in a way, Broadcom's saying almost Alphabet's going to 0:14:37.880 --> 0:14:40.560 get its money from Anthropic for buying the chips. 0:14:40.800 --> 0:14:41.280 Is that right? 0:14:41.400 --> 0:14:44.080 Meanwhile, who's getting the money for the leases and who 0:14:44.160 --> 0:14:46.480 therefore is Alphabet saying like, you're good for the money, 0:14:46.480 --> 0:14:48.920 don't worry. Is that the people actually constructing the data 0:14:48.920 --> 0:14:50.160 centers owning the land. 0:14:50.600 --> 0:14:54.880 It's the leases are to Fluid Stack, which is a 0:14:54.920 --> 0:14:58.680 company that Anthropic has said it's been working with to 0:14:58.800 --> 0:15:03.080 develop these data seen. So you see how it's it's complicated, right, 0:15:03.480 --> 0:15:04.400 there's many. 0:15:04.400 --> 0:15:05.560 See how it's complicated. 0:15:05.640 --> 0:15:08.480 Yeah, yeah, I mean to pull off a deal of 0:15:08.520 --> 0:15:12.040 this magnitude, which I think is the largest private credit 0:15:12.040 --> 0:15:16.160 deal in history, definitely the largest chip deal. There's a 0:15:16.160 --> 0:15:18.560 lot of moving pieces. One of the key things is 0:15:18.560 --> 0:15:20.840 that these chips are not I mean, they need to 0:15:20.920 --> 0:15:21.720 be created. 0:15:21.760 --> 0:15:27.200 They don't exist right now. But yeah, there's a lot 0:15:27.280 --> 0:15:27.960 that goes into this. 0:15:28.600 --> 0:15:33.880 Who's the manufacturer of questions around Intel? VISs TSMC absolutely fascinating. 0:15:33.920 --> 0:15:36.440 Scott Carpenter. He broke it down so clearly, we so 0:15:36.480 --> 0:15:39.400 appreciate it. Meanwhile, let's turn our attention to soft Bank. 0:15:39.520 --> 0:15:42.080 It's attempt to leverage its massive AI bets. It's hitting 0:15:42.200 --> 0:15:44.680 a bit of a wall. Sources told Bloomberg that talks 0:15:44.680 --> 0:15:46.960 are stalled with potential creditors to raise at least six 0:15:46.960 --> 0:15:50.200 billion dollars from a margin loan backed by its opening 0:15:50.240 --> 0:15:53.360 eye steak. So it's unclear why the pores comes just 0:15:53.400 --> 0:15:56.200 weeks after soft Bank slashed it's fundraising target from ten 0:15:56.240 --> 0:15:59.080 billion dollars, and soft Bank shares have tumbled nearly ten percent. 0:15:59.120 --> 0:16:01.960 On the news today, A says the firm ways alternative 0:16:02.240 --> 0:16:04.400 funding options coming up. 0:16:04.840 --> 0:16:05.760 We are going to be speaking a. 0:16:05.840 --> 0:16:10.240 Saphia Noble, Professor, director of the Center of Resilience and 0:16:10.280 --> 0:16:13.800 Digital Justice, at the UCLA to discuss bias discrimination within 0:16:13.840 --> 0:16:14.520 this world of AI. 0:16:14.560 --> 0:16:16.720 We keep talking more on that next as a Blueberg tech. 0:16:24.040 --> 0:16:26.680 This week, as Open AI filed its S one confidentially, 0:16:26.760 --> 0:16:29.600 CEO Samaltman was also out with a sweeping long term 0:16:29.640 --> 0:16:33.160 vision for generative AIS alignment with humanity and warned that 0:16:33.440 --> 0:16:37.600 transformative technologies such as AI can concentrate power, stating Open 0:16:37.640 --> 0:16:40.400 AIS quote clear ride about the risks as it aims 0:16:40.440 --> 0:16:43.680 to build powerful systems that remain safe subject to human control. 0:16:43.720 --> 0:16:47.600 But critics have long voiced concerns about AI risks such 0:16:47.600 --> 0:16:51.280 as algorithmic bias in equality joining us now, Sofia Noble, 0:16:51.440 --> 0:16:55.800 UCLA professor author of the acclaimed book Algorithms of Oppression 0:16:56.800 --> 0:16:59.800 clear Ride, is that enough. Are we seeing some of 0:16:59.840 --> 0:17:04.200 the risks being digested and answered for within these models? 0:17:04.640 --> 0:17:06.960 I don't think so. I don't think we're anywhere near 0:17:08.160 --> 0:17:12.400 a call for or an ability to realize safe AI. 0:17:13.359 --> 0:17:17.160 What we see, in fact, are chapbot technologies and large 0:17:17.200 --> 0:17:21.000 language models that for the most part don't have markets. 0:17:21.359 --> 0:17:24.560 I mean, they were built for Corporate America to reduce 0:17:24.640 --> 0:17:27.760 labor costs, but corporate America is moving away from them 0:17:27.800 --> 0:17:31.320 because they're very expensive, they're not really reliable, they are 0:17:31.560 --> 0:17:35.160 incredibly negatively impactful on the environment. 0:17:35.400 --> 0:17:37.360 Moving away from them, you think corporate America is moving 0:17:37.359 --> 0:17:38.439 away from lagenguge porp. 0:17:38.520 --> 0:17:38.760 I do. 0:17:38.840 --> 0:17:42.200 We've been seeing studies where companies are saying that it's 0:17:42.280 --> 0:17:46.639 more expensive for them to use these chatbots because human 0:17:46.680 --> 0:17:50.200 beings have to check the efficacy and the reliability. There 0:17:50.200 --> 0:17:54.520 are so many errors, factual errors that are proliferating through 0:17:54.520 --> 0:17:58.600 these technologies. So if the technology itself is that flawed 0:17:59.320 --> 0:18:02.960 and it's being now on the public as some type 0:18:03.000 --> 0:18:06.119 of solution, I think we're in trouble. And of course 0:18:06.240 --> 0:18:10.760 we know that we have the racial bias, the gender bias, 0:18:11.040 --> 0:18:14.960 the kind of geographic and political concerns about what comes 0:18:14.960 --> 0:18:18.840 out of these technologies. I think that we are moving 0:18:18.880 --> 0:18:24.120 into very dangerous territory trying to bolster our society on 0:18:24.359 --> 0:18:26.280 large language models fascinatable. 0:18:26.560 --> 0:18:29.680 The large body of your work looked at the data 0:18:29.720 --> 0:18:33.560 issues for commercial search engines. Basically the net result is 0:18:34.119 --> 0:18:38.800 that the search engines, as per your books title, reinforce racism. Yes, 0:18:39.520 --> 0:18:43.840 what was the underlying issue in the search engine case study? 0:18:44.119 --> 0:18:47.480 And what is different or the same about the large 0:18:47.520 --> 0:18:50.560 language models that I think you're saying yield a similar 0:18:50.600 --> 0:18:51.560 result they do. 0:18:52.280 --> 0:18:56.480 So what we've seen over the last fifteen twenty years 0:18:56.640 --> 0:19:00.000 is that all of the discrimination that's in our society, 0:19:00.400 --> 0:19:04.520 all of the kind of stereotyping, all of the inequality, 0:19:05.160 --> 0:19:08.920 just gets packaged up and then used to train models. 0:19:09.240 --> 0:19:11.000 So it's within the data. 0:19:10.640 --> 0:19:13.480 Within the data, but it's also the people who are 0:19:13.480 --> 0:19:16.960 designing the models are really not aware. They don't understand 0:19:17.040 --> 0:19:21.320 the kind of social, historical, economic processes. These are software 0:19:21.320 --> 0:19:23.640 engineers who don't even think about they don't even ask 0:19:23.680 --> 0:19:26.480 the kinds of questions that let's say, a sociologist like 0:19:26.520 --> 0:19:30.240 I would ask. And so we have discriminatory data that 0:19:30.359 --> 0:19:34.680 is training models. But what's different now is that these 0:19:34.800 --> 0:19:40.880 models obfuscate the inequality. They appear to be factual and reliable. 0:19:41.160 --> 0:19:43.920 And if you don't know, if you don't have deep expertise, 0:19:44.200 --> 0:19:45.880 you're not going to know that the kinds of things 0:19:45.920 --> 0:19:49.640 that are being served up in these products are actually faulty. 0:19:49.840 --> 0:19:55.040 And imagine putting your whole business enterprise, your public institutions, 0:19:55.080 --> 0:19:59.240 your schools, your libraries, making that the backbone. I mean 0:19:59.320 --> 0:20:02.160 that is to me quite dangerous Sophia. 0:20:03.040 --> 0:20:05.800 We can talk at length about the risks and the problems. 0:20:06.200 --> 0:20:08.479 What about the solutions here because we saw but at 0:20:08.560 --> 0:20:10.400 least two years ago, I think it was when alphabet 0:20:10.480 --> 0:20:13.520 was struggling to ensure that some of the images AI 0:20:13.640 --> 0:20:18.080 generated images didn't overcompensate for some of the worries about 0:20:18.480 --> 0:20:20.480 racism and bias within the algorithm. 0:20:20.640 --> 0:20:23.879 So what have you been done that works. Let's not 0:20:23.920 --> 0:20:26.840 just beautify the problem, let's give us the solution. 0:20:27.560 --> 0:20:30.000 Well, I think that we don't want to give up 0:20:30.240 --> 0:20:37.040 what it means to have human expertise, human journalists, fact checkers, teachers, thinkers. 0:20:37.440 --> 0:20:39.560 This is our most powerful asset. 0:20:39.680 --> 0:20:43.280 These human beings are people, and we can't replace people 0:20:43.320 --> 0:20:46.320 with these kinds of machines. So to me, you know, 0:20:46.880 --> 0:20:50.240 having deep knowledge in the humanities and social sciences, these 0:20:50.240 --> 0:20:52.719 are the things that are going to really be important 0:20:52.760 --> 0:20:56.840 as we go forward in society. And we're over investing, 0:20:56.960 --> 0:20:59.199 I think, in the wrong things. We need to be 0:20:59.280 --> 0:21:04.960 investing in putting resources into pro social, pro rights respecting technology. 0:21:05.280 --> 0:21:09.320 There's a whole world of small language models and different 0:21:09.400 --> 0:21:13.680 kinds of very interesting kinds of technologies that women are 0:21:13.720 --> 0:21:17.040 thinking about that people of color are working on and 0:21:17.080 --> 0:21:19.800 these are the least invested in, but they are I 0:21:19.840 --> 0:21:23.800 think the kinds of technologies that are going to help 0:21:23.920 --> 0:21:25.160 us find a way forward. 0:21:25.240 --> 0:21:29.359 Sophea, How conscious of and open about are the companies 0:21:29.760 --> 0:21:32.879 on the issue, And you know, in research and writing 0:21:32.920 --> 0:21:34.720 your book, but your ongoing work, how much do they 0:21:34.720 --> 0:21:35.760 engage with you on it? 0:21:36.440 --> 0:21:40.320 The companies for the most part want to deny, deny 0:21:40.920 --> 0:21:44.560