WEBVTT - Taiwan Eyes China AI Chip Sales Curbs; OpenAI Files for IPO 0:00:02.520 --> 0:00:13.200 Bloomberg Audio Studios, Podcasts, radio news. Bloomberg Tech is a 0:00:13.280 --> 0:00:17.040 live from coast to coast with Caroline Hide in New 0:00:17.120 --> 0:00:19.599 York and v Lo Loow in San Francisco. 0:00:21.600 --> 0:00:23.240 This is Bloomberg Tech coming up. 0:00:23.280 --> 0:00:26.600 Taiwan, weihs some of its toughest restrictions yet on AI 0:00:26.680 --> 0:00:29.600 chip sales to China. Will have the details plus the 0:00:29.720 --> 0:00:32.800 AIIPO raise. It heats up with Open Ai joining its 0:00:32.880 --> 0:00:35.720 rivals with plans for a potential public listing later this year. 0:00:36.400 --> 0:00:39.360 And Apple lays the foundation for the AI era, but 0:00:39.400 --> 0:00:41.479 how soon will it come? As it hints that the 0:00:41.479 --> 0:00:44.960 company's upcoming faudaal iPhone is on deck. But first I 0:00:45.120 --> 0:00:48.040 check on those markets. We're off by eight ten percent. 0:00:48.080 --> 0:00:51.239 We have been optimistic to start trade and then suddenly 0:00:51.280 --> 0:00:54.279 we've turned quite a leg lower into negative territory. And 0:00:54.320 --> 0:00:56.880 it is tech chip stocks in particular the dragoslower, but 0:00:56.920 --> 0:00:59.840 it's also the key player that is Apple, key player 0:00:59.880 --> 0:01:01.560 that is in Vidia that drag us down. From our 0:01:01.560 --> 0:01:04.000 points perspective, I shine light what was happening in terms 0:01:04.000 --> 0:01:07.319 of hardware. The Philly Semiconductor Index, the socks as we 0:01:07.360 --> 0:01:09.520 call it, one point eight percent lower on the day 0:01:09.560 --> 0:01:12.840 after having reprieved some of the Friday sell off. Yesterday's 0:01:12.840 --> 0:01:15.960 trade Apple off again, the Juggernaut off by two point 0:01:16.040 --> 0:01:18.600 seven percent. That's important in terms of points. It's important 0:01:18.600 --> 0:01:20.320 to the overall index, and it drags the S and 0:01:20.400 --> 0:01:22.479 P five hundred lower as well as one of those 0:01:22.480 --> 0:01:25.520 some anxiety as to when really we will get Siri 0:01:25.720 --> 0:01:29.000 AI in our hands. But first we also talk Taiwan 0:01:29.280 --> 0:01:31.479 and it is weighing some of its toughest restrictions yet 0:01:31.520 --> 0:01:34.319 on AI chip sales to China, a move we'll bring 0:01:34.360 --> 0:01:37.039 it closer in line with Washington and mid mounting US 0:01:37.120 --> 0:01:40.840 concern over technology leakage and national security. For more and 0:01:40.880 --> 0:01:42.720 what it means for the AI race and the US 0:01:42.840 --> 0:01:46.240 China tech tensions. We're joined by Blue Meg's global tech editor, 0:01:46.360 --> 0:01:47.160 Peter Elstrom. 0:01:47.560 --> 0:01:49.000 Is it significant this move? 0:01:51.080 --> 0:01:54.880 It's certainly significant. The most sophisticated chips in the world 0:01:54.880 --> 0:01:58.320 are made in Taiwan by TSMC. Of course, in what 0:01:58.360 --> 0:02:01.840 we understand from our source is that the Taiwanese government 0:02:02.200 --> 0:02:05.120 is now weighing these restrictions on making it a crime 0:02:05.240 --> 0:02:08.520 to export advanced AI chips to China. Now, as you mentioned, 0:02:08.520 --> 0:02:11.919 this is in line with US restrictions already, where in 0:02:12.000 --> 0:02:14.880 Nvidia can't sell its most advanced AI chips to China, 0:02:14.880 --> 0:02:17.480 for example. But why is this important for Taiwan. It 0:02:17.560 --> 0:02:20.079 is because they hold that place as the manufacturer of 0:02:20.160 --> 0:02:24.000 these key chips. But also Taiwanese authorities will be able 0:02:24.040 --> 0:02:26.720 to use more tools to be able to enforce these 0:02:26.720 --> 0:02:30.440 restrictions within Taiwan. They did detain their first person because 0:02:30.440 --> 0:02:33.240 of these export restrictions a while ago, but they had 0:02:33.280 --> 0:02:36.480 to do it because of falsifying documents, not because it's 0:02:36.520 --> 0:02:39.440 a crime right now to smuggle chips out of the country. 0:02:39.480 --> 0:02:41.360 So it gives them a few more tools to be 0:02:41.360 --> 0:02:44.200 able to enforce these restrictions. It's a sign that President 0:02:44.280 --> 0:02:46.760 LII there in Taiwan is taking a much harder line 0:02:46.800 --> 0:02:47.480 against China. 0:02:48.080 --> 0:02:51.040 And what does that mean from relationships between the US 0:02:51.120 --> 0:02:53.880 and China. What does that mean in terms of Montbelio's 0:02:54.000 --> 0:02:56.880 conversation coming from the leadership in China against Taiwan. 0:02:58.040 --> 0:03:01.079 Well, you're alluding to the complex city of the situation, 0:03:01.200 --> 0:03:03.880 and we just saw President Trump go and visit Shi 0:03:04.800 --> 0:03:07.320 in China. They talked about these very issues. There was 0:03:07.360 --> 0:03:10.959 a lot of discussion about many different things, including trade 0:03:11.000 --> 0:03:14.160 and technology in particular. So it's a moving playing field 0:03:14.160 --> 0:03:16.480 at this point. And the Trump administration has signaled that 0:03:16.520 --> 0:03:19.720 it would allow Nvidia to sell the H two hundred chips, 0:03:19.720 --> 0:03:22.680 for example, into China. China is ensure that it wants that. 0:03:23.240 --> 0:03:25.400 It's very you know, they also would like to develop 0:03:25.440 --> 0:03:29.240 their own domestic semiconductor industry. They want to support Huawei 0:03:29.320 --> 0:03:31.360 and other companies like that that are trying to build 0:03:31.400 --> 0:03:34.160 in this area. So it's a complicated task. It certainly 0:03:34.240 --> 0:03:37.440 does mean that there are tensions that remain between the 0:03:37.440 --> 0:03:40.640 two countries when it comes to these kinds of technologies, 0:03:40.720 --> 0:03:42.880 and there's not a resolution in the near future. 0:03:43.560 --> 0:03:46.840 We just saw in fact, China preparing two trillion and 0:03:46.960 --> 0:03:51.000 one of spending to really focus on its own domestic 0:03:51.120 --> 0:03:53.640 AI build out, whether that be data centers. In many ways, 0:03:53.640 --> 0:03:55.320 that's going to be a focus on its own chips. 0:03:55.480 --> 0:03:57.440 But b to we also got that sort of interesting 0:03:57.480 --> 0:04:01.360 news that the Pentagon is really once again reamping some 0:04:01.400 --> 0:04:04.400 of the concerns about Chinese companies, tech companies in their 0:04:04.480 --> 0:04:07.360 relationship with the military. All of the spells that that 0:04:07.440 --> 0:04:08.760 antagonism just remains. 0:04:10.400 --> 0:04:12.920 Yeah, the first story you mentioned is a very important 0:04:12.960 --> 0:04:16.159 one again a people familiar story that we wrote about 0:04:16.240 --> 0:04:19.560 how China is weighing this plan to invest two trillion 0:04:19.640 --> 0:04:22.320 one roughly three hundred billion dollars into data centers in 0:04:22.360 --> 0:04:24.560 the country. This is kind of the missing piece in 0:04:24.600 --> 0:04:27.640 the Chinese AI industry at this point, as we saw 0:04:27.680 --> 0:04:30.359 with deep seak Ali Baba's work with Gwen in particular. 0:04:30.560 --> 0:04:33.600 They've been quite successful at developing advanced models, but they've 0:04:33.720 --> 0:04:36.920 made nowhere near the kinds of investments in infrastructure and 0:04:37.000 --> 0:04:39.640 data centers that we've seen in the US. And now 0:04:39.920 --> 0:04:42.359 three hundred billion dollars works out to about sixty billion 0:04:42.400 --> 0:04:44.640 dollars a year over five years. It's not nearly as 0:04:44.720 --> 0:04:47.400 much as the four hyperscalers are spending. They're aiming for 0:04:47.440 --> 0:04:50.000 seven hundred and twenty five billion dollars this year, so 0:04:50.040 --> 0:04:52.760 it's sort of a ten to one ratio. But these 0:04:52.839 --> 0:04:55.480 data centers are much less expensive in China, and it's 0:04:55.520 --> 0:04:58.160 a sign that they're going to compete on this infrastructure 0:04:58.240 --> 0:05:00.479 level as well as the model level, So that's a 0:05:00.480 --> 0:05:01.560 big step for them. 0:05:01.600 --> 0:05:04.400 Fascinating play most Peter Elstrom, where the rap of all 0:05:04.440 --> 0:05:05.040 that is your news. 0:05:05.120 --> 0:05:07.320 We so appreciate it now as. 0:05:07.160 --> 0:05:11.640 AI becomes increasingly tied to both geopolitics, but investment strategy 0:05:11.960 --> 0:05:15.000 messas are looking beyond the initial infrastructure build out potentially 0:05:15.000 --> 0:05:17.880 and actually applications of AI. Let's discuss that with LATU 0:05:18.200 --> 0:05:20.680 Alliance spend seen head of thematic equities who sees AI 0:05:20.680 --> 0:05:25.120 adoption drawning as the focus shifts to long term competitive positioning. 0:05:25.560 --> 0:05:28.600 So at this moment, just to bring in the US 0:05:28.600 --> 0:05:31.760 and China perspective, how competitive is the US and the 0:05:31.800 --> 0:05:35.919 buildout of AI data centers, They're like, how solid is 0:05:35.920 --> 0:05:39.479 the supply chain if we're worried about US China relationships 0:05:39.480 --> 0:05:40.599 and what it means for Taiwan. 0:05:41.200 --> 0:05:45.120 Yes, I think supply chain is important, but if you 0:05:45.120 --> 0:05:47.720 think about US holistically, I do think a lot of 0:05:47.839 --> 0:05:51.200 leading edge technology comes out of US. And if you 0:05:51.279 --> 0:05:54.240 look at the US companies, the tech giants, when you 0:05:54.279 --> 0:05:57.760 look at the billions and hundreds of billions solid they're spending, 0:05:58.240 --> 0:06:00.960 they're actually quite strategic about it, and the pace of 0:06:01.040 --> 0:06:04.440 innovation and the innovation in the architecture itself is actually 0:06:04.520 --> 0:06:08.480 picking up pretty rapidly. So think about them. They're thinking 0:06:08.480 --> 0:06:12.240 about the cost protoken holistically, and more and more companies 0:06:12.320 --> 0:06:15.840 maybe are taking that holistic view and be more vertically 0:06:15.839 --> 0:06:19.240 integrated and then think about whether it be Google introduces 0:06:19.360 --> 0:06:23.480 proprietory chips, etc. So they are thinking about different architectures, 0:06:23.480 --> 0:06:26.880 how to lower the cost protoken, so that set them 0:06:26.880 --> 0:06:31.359 ahead in terms of competitive advantage. So because all the 0:06:31.440 --> 0:06:34.760 hyperscalers are all competing in it, and in a way 0:06:34.760 --> 0:06:37.240 it's an arms race, and that everybody wants to get 0:06:37.240 --> 0:06:40.479 ahead because that actually goes well for their long term 0:06:40.520 --> 0:06:44.760 competitive vantage and long term margin. So because the competition 0:06:45.120 --> 0:06:48.840 is so high that we do think the US at 0:06:48.920 --> 0:06:52.280 this moment when you look at architecture, certainly US is 0:06:52.480 --> 0:06:54.760 quite ahead of most people in the world. 0:06:54.960 --> 0:06:58.040 I think that vertical integration is being so clearly demonstrated 0:06:58.320 --> 0:07:01.640 by SpaceX at the moment that Elomusk is deciding, yes, 0:07:01.680 --> 0:07:03.920 I'm a space giant, I'm an AI giant, but also 0:07:03.960 --> 0:07:06.080 I'm a data center giant, and actually I'm selling some 0:07:06.120 --> 0:07:09.400 of my compute to the likes of Alphabet and Google 0:07:09.560 --> 0:07:10.560 or indeed anthropic. 0:07:11.360 --> 0:07:12.000 Is that the. 0:07:11.920 --> 0:07:14.239 Way in which you would like to see companies thinking 0:07:14.280 --> 0:07:16.720 themselves in a more vertically integrated manner, or do you 0:07:16.800 --> 0:07:19.400 want to see more specific applications of AI is that 0:07:19.440 --> 0:07:20.760 also attracted. 0:07:21.040 --> 0:07:24.440 I think it's both. Actually it's at this point of 0:07:24.520 --> 0:07:28.200 the cycle, it's certainly important to see more applications come out. 0:07:28.480 --> 0:07:31.800 But if we look back historically, it's actually quite interesting 0:07:31.960 --> 0:07:36.280 vertical integration and that type of cost atdvantage is very 0:07:36.320 --> 0:07:40.280 hard to replicate, and in this very compute intensive and 0:07:40.320 --> 0:07:44.200 capital intensive cycle, I think vertical integration is something that's 0:07:44.320 --> 0:07:48.400 very interesting to worth that's worth considering from an investment's perspective. 0:07:48.800 --> 0:07:51.960 But that said, if you look in the past cycle, 0:07:52.080 --> 0:07:55.720 what happened is we build a massive comp infrastructure and 0:07:55.760 --> 0:08:00.400 then the entire ecosystem that's capitalized ultimately monetized, and that's 0:08:00.440 --> 0:08:03.400 created the tech giants as we know today. So it's 0:08:03.440 --> 0:08:06.520 probably too early to tell, but I certainly think vertical 0:08:06.520 --> 0:08:09.480 integration seems to be a direction that a lot of 0:08:09.520 --> 0:08:12.840 these companies are considering and we're watching very closely. 0:08:13.040 --> 0:08:14.240 It could be quite interesting. 0:08:14.400 --> 0:08:17.160 Are you feeling comfortable with the capex still of a 0:08:17.160 --> 0:08:20.000 lot of these large companies and the industry writ large. 0:08:20.480 --> 0:08:23.000 I think this year we always thought that it was 0:08:23.080 --> 0:08:24.920 going to be this has to be the year of 0:08:24.960 --> 0:08:28.240 AI adoption and the fact that token is being utilized 0:08:28.240 --> 0:08:31.240 and adopted as such, a rapid pace is certainly a 0:08:31.280 --> 0:08:34.840 good sign that AI is being adopted, it's being used, 0:08:35.160 --> 0:08:37.840 So that makes us feel better about the amount of 0:08:37.840 --> 0:08:41.040 capex that's going into it, because ultimately we do want 0:08:41.080 --> 0:08:43.600 to spend and build a road to somewhere, and seeing 0:08:43.640 --> 0:08:47.440 the early signs of rapid adoption is certainly very very encouraging. 0:08:47.440 --> 0:08:50.439 You're all about themes, You're all about long term perspectives 0:08:50.440 --> 0:08:51.480 and not single names. 0:08:51.800 --> 0:08:54.200 But at this point, is there going to be better. 0:08:54.080 --> 0:08:57.360 Entry points to the market? Does it worry you quiv 0:08:57.400 --> 0:09:00.000 you calls for ants that just how high we are 0:09:00.360 --> 0:09:05.040 in general trading on big benchmarks, at big private market valuations, 0:09:05.040 --> 0:09:06.599 big public market valuations. 0:09:07.280 --> 0:09:11.240 I think that all companies are created equal, and I 0:09:11.280 --> 0:09:13.520 think we have to look at the company really one 0:09:13.520 --> 0:09:16.800 by one in terms of we really have to look out. 0:09:17.360 --> 0:09:21.679 And I think historically when we look back, sometimes the 0:09:21.720 --> 0:09:24.800 consensus street estimates they get it right the near term, 0:09:25.040 --> 0:09:27.559 but they miss the long term. And it's the long 0:09:27.679 --> 0:09:31.040 term that gets revised up and up that ultimately determines 0:09:31.080 --> 0:09:33.680 how much a stock should be valued today. So in 0:09:33.720 --> 0:09:36.360 a way, to look out actually allows you to make 0:09:36.400 --> 0:09:39.840 more money near term. So it's really about long term 0:09:39.880 --> 0:09:43.880 total addressable market and more importantly, the ability to hold 0:09:44.040 --> 0:09:46.640 your margin and get the long term profit. That's what 0:09:46.760 --> 0:09:50.200 ultimately determines whether a stock is valued fairly today. 0:09:50.600 --> 0:09:52.400 Thanks you, We love having you on the show. 0:09:52.520 --> 0:09:55.800 Thank you for joining us of Alliance Bernstein there. Meanwhile, 0:09:55.840 --> 0:09:59.079 coming up, open Ai does indeed join its AI rivals 0:09:59.080 --> 0:10:00.320 and the race of potential. 0:10:00.080 --> 0:10:02.040 Go public will have the details on that. S one 0:10:02.720 --> 0:10:03.600 as a Bloombg Tech. 0:10:15.320 --> 0:10:18.400 So open Ai has joined the AI rivals with plans 0:10:18.400 --> 0:10:21.400 for a public listing potentially later this year. The chatchypt 0:10:21.600 --> 0:10:24.680 make confidentially filed it's S one or the SEC, but 0:10:24.760 --> 0:10:27.160 it said the timing of an IPO hasn't been decided 0:10:27.200 --> 0:10:29.800 because there were still things it wants to do as 0:10:29.840 --> 0:10:33.680 a private company. Bloomberg's AI reporter Shurigafari joins us with 0:10:33.720 --> 0:10:34.200 a details. 0:10:34.240 --> 0:10:35.840 So what would those things potentially be? 0:10:36.080 --> 0:10:39.320 Why make it still unclear if they're going to come 0:10:39.360 --> 0:10:40.120 as soon as the fall? 0:10:41.559 --> 0:10:45.600 Yeah, the company said, this is really about optionality for them. 0:10:45.679 --> 0:10:47.320 Of course, there are a lot of factors when a 0:10:47.320 --> 0:10:49.920 company's deciding to go public. There's the market, how they 0:10:49.960 --> 0:10:52.959 stack up to their competitors, So these could all be reasons, 0:10:53.040 --> 0:10:55.080 right that a company has in mind when they're thinking 0:10:55.120 --> 0:11:01.720 about the exact timing of going forward with the filing. 0:11:00.080 --> 0:11:02.400 Forward with the fine, and going forward with the next 0:11:02.440 --> 0:11:04.200 phase of open AI. 0:11:04.280 --> 0:11:05.959 I thought it was really interesting the same. 0:11:05.800 --> 0:11:08.319 Time as we learn that they've filed an S one confidentially, 0:11:08.559 --> 0:11:12.000 they're also putting out a statement about the control of AI, 0:11:12.080 --> 0:11:14.720 about aligning it with humanity that was coming from Sam 0:11:14.720 --> 0:11:15.880 Altman and the chief scientist. 0:11:15.960 --> 0:11:18.920 What did you make of the blog? That's right? 0:11:18.960 --> 0:11:20.640 You know, there's been a lot of talk lately in 0:11:20.679 --> 0:11:23.480 the AI world about the idea of self improving AI. 0:11:23.640 --> 0:11:26.719 That these AI models are getting so good that researchers 0:11:26.760 --> 0:11:30.960 at the top labs are actually using them to program. 0:11:30.440 --> 0:11:31.840 The next model, and. 0:11:31.760 --> 0:11:35.160 That's caused both excitement but a lot of anxiety. We 0:11:35.240 --> 0:11:39.080 actually saw Inimthropic last week come out with a blog 0:11:39.080 --> 0:11:43.360 post even calling for a potential pause if this turns 0:11:43.360 --> 0:11:45.440 out to really be the case that these models are 0:11:45.480 --> 0:11:48.760 sort of improving so quickly and getting out of control. 0:11:49.360 --> 0:11:51.680 So I think, you know, Altman's post is in a 0:11:51.679 --> 0:11:55.320 way in conversation with that and acknowledging this idea that 0:11:55.360 --> 0:11:58.360 the AI is starting to automate itself and weighing some 0:11:58.480 --> 0:12:01.280 of both the opportunities it also risks there. 0:12:01.400 --> 0:12:03.920 And I think that's so interesting that, of course, again 0:12:04.080 --> 0:12:07.000 just by outlining that blog post, we're thinking about anthropic 0:12:07.280 --> 0:12:09.880 timing open AI timing. One puts a blog, the next 0:12:09.880 --> 0:12:12.800 pet's a blog. One's gone file confidentially the next one has. 0:12:13.000 --> 0:12:15.280 How much is their anticipation, how much is their worry 0:12:15.320 --> 0:12:16.800 that they're coming in the same sort of time. 0:12:17.960 --> 0:12:20.199 It does feel like a race on every level right 0:12:20.240 --> 0:12:25.080 now between these two sort of top AI firms, both 0:12:25.120 --> 0:12:28.920 at the technological level, with them releasing a new model 0:12:29.040 --> 0:12:31.520 sort of neck and neck every few months now even 0:12:31.559 --> 0:12:35.560 every few weeks, on the sort of ideological level about 0:12:35.600 --> 0:12:37.599 the direction that they're going to go in balancing the 0:12:37.679 --> 0:12:41.120 risks and the benefits, as well as just this IPO timeline. 0:12:41.160 --> 0:12:44.080 So I think it's one of the closest races certainly 0:12:44.080 --> 0:12:46.920 that I've ever covered in the tech industry, and extremely 0:12:46.960 --> 0:12:48.600 competitive right now on every level. 0:12:48.720 --> 0:12:51.640 And you cover it brilliantly, Thank you, Bloomberg Sharing KAfari 0:12:51.679 --> 0:12:54.200 on all things open ai. And let's stick with that 0:12:54.280 --> 0:12:58.040 IPO pipeline. The AI names that we've just talked about. 0:12:58.040 --> 0:13:00.840 If you add in SpaceX, that's three point six trillion 0:13:00.880 --> 0:13:02.680 dollars a market cap that might be coming to the 0:13:02.679 --> 0:13:05.520 market in the next few months. But according to Pitchbook research, 0:13:05.880 --> 0:13:09.280 open ai may end up being the most expensive bet. 0:13:09.600 --> 0:13:12.080 Harrison roll first is with us senior analyst of Pitchbook's 0:13:12.080 --> 0:13:14.880 private company coverage joining us. Now, look, we can't see 0:13:14.960 --> 0:13:17.200 the detail of the revenues and the numbers that's been 0:13:17.240 --> 0:13:19.960 filed confidentially with the SEC. But you take a great 0:13:19.960 --> 0:13:22.480 crack at it with an eye on open ai, your 0:13:22.559 --> 0:13:24.920 late stage company research, and you think that open ai 0:13:25.040 --> 0:13:27.880 is expensive in some of your frameworks. Can you talk 0:13:27.920 --> 0:13:28.679 us through at Harrison? 0:13:30.080 --> 0:13:32.320 Well, yeah, so open aiy has essentially built the most 0:13:32.360 --> 0:13:36.320 widely used AI product in history, but they rely on 0:13:36.400 --> 0:13:42.040 so much capital intensive compute that their profits aren't necessarily 0:13:42.040 --> 0:13:46.560 going to cover their costs. So when looking at open ai, 0:13:46.800 --> 0:13:50.000 we want to see their Microsoft terms, we want to 0:13:50.000 --> 0:13:52.679 see how their accounting works. We want to see how 0:13:52.720 --> 0:13:55.439 their enterprise revenue as a share total and their net 0:13:55.480 --> 0:13:59.440 revenue retention, the annual compute obligation, and the related party 0:13:59.600 --> 0:14:04.640 disclosed to really figure out exactly where they stand. As 0:14:04.679 --> 0:14:09.520 a true business as opposed to the hype that's surrounding them. 0:14:10.800 --> 0:14:15.000 The framework that I've created, this AI business quality essentially 0:14:15.040 --> 0:14:19.360 looks at opening eye on five different dimensions. Is revenue quality, 0:14:19.840 --> 0:14:25.480 computing independence, remote durability, governance, optionality, as well as just 0:14:25.520 --> 0:14:28.840 capital efficiency. When you look at a company who really 0:14:28.880 --> 0:14:33.520 doesn't who really has a strong AI mind share, but 0:14:33.680 --> 0:14:38.160 isn't focusing on the true business qualities that make a 0:14:38.200 --> 0:14:42.520 business successful in the public markets, it's it's something hard 0:14:42.560 --> 0:14:45.280 to evaluate right now, and that's why they score the 0:14:45.320 --> 0:14:48.680 lowest on my scoring system. 0:14:48.240 --> 0:14:50.360 And four point eight b one of the highest valuations 0:14:50.560 --> 0:14:51.240 four point. 0:14:51.000 --> 0:14:55.240 Eight out of ten on aib key AIBQ framework. 0:14:55.560 --> 0:14:58.520 Now I'm interested in that sort of other. 0:14:58.920 --> 0:15:02.640 Parties perspective because the Microsoft relationship is one that you 0:15:02.720 --> 0:15:04.960 drill down on and perhaps some of that revenue sharing 0:15:05.000 --> 0:15:06.680 is something investors should really be thinking about. 0:15:08.160 --> 0:15:13.560 Absolutely, the Microsoft share aspect is very unique. 0:15:13.120 --> 0:15:14.840 By capping how. 0:15:14.800 --> 0:15:19.680 Much they oh Microsoft, it frees up some sort of revenue. However, 0:15:20.200 --> 0:15:24.440 it's still unsure about what's going to happen after twenty thirty, 0:15:24.600 --> 0:15:27.920 and that uncertainty is going to make it very hard 0:15:27.960 --> 0:15:31.360 for investors to look past twenty thirty and see if 0:15:31.400 --> 0:15:32.800 they're going to succeed until. 0:15:32.520 --> 0:15:35.680 Then, can you talk a little bit about what you 0:15:35.720 --> 0:15:39.280 anticipate for governance structures coming from these companies. We're going 0:15:39.320 --> 0:15:40.560 to dig into it in a bit at the end 0:15:40.600 --> 0:15:43.800 of the show about questions people have about SpaceX's governance 0:15:44.200 --> 0:15:47.920 and the control muscaz how much you're starting to see 0:15:47.920 --> 0:15:49.160 how much control might be. 0:15:49.160 --> 0:15:49.840 Had by open Ai. 0:15:49.880 --> 0:15:52.520 For example, samum and we understand doesn't own any significant 0:15:52.520 --> 0:15:55.000 portion of equity. 0:15:55.160 --> 0:16:00.440 No, he's interesting from going from a nonprofit to this 0:16:01.360 --> 0:16:05.960 just new PBC structures. He's essentially trying to find a 0:16:06.000 --> 0:16:09.280 way to be the good guy. But at the same 0:16:09.360 --> 0:16:12.760 time he's already dealt with such backlash from his from 0:16:12.800 --> 0:16:16.320 the board and other parties involved in other investors where 0:16:16.760 --> 0:16:18.880 they are even unsure whether or not they want. 0:16:18.720 --> 0:16:22.680 To deal with this sort of these sort of issues. 0:16:22.800 --> 0:16:25.960 However, Sam Alvins still finds a way to make it happen, 0:16:26.640 --> 0:16:30.720 and now people are really relying and betting on his 0:16:30.760 --> 0:16:33.760 future and his aspirations to bring Open the Eye to. 0:16:33.800 --> 0:16:38.280 Number one, number one in a race we keep talking about. 0:16:38.280 --> 0:16:40.560 It is feeling like some sort of race, but it 0:16:40.640 --> 0:16:43.760 is about mind share. It is about oxygen in the room, 0:16:43.800 --> 0:16:45.560 but it is about money that's going to be able 0:16:45.560 --> 0:16:47.680 to be allocated to these companies when they go public. 0:16:48.080 --> 0:16:51.200 Is there a worry that they file confidentially lost and 0:16:51.200 --> 0:16:53.000 we're getting SpaceX as soon as this week. 0:16:55.040 --> 0:16:59.320 I think by filing last actually gives open Ad opportunity 0:16:59.560 --> 0:17:04.040 to learn from their competitors. Opening I can actually see 0:17:04.440 --> 0:17:07.959 where Anthropic is going wrong, exactly how they can adjust 0:17:08.400 --> 0:17:09.040 the gorong. 0:17:09.119 --> 0:17:10.080 Where are they going wrong? 0:17:11.359 --> 0:17:13.480 You know, Opening I could be going wrong with their 0:17:13.520 --> 0:17:17.879 gross margins. They don't really break down exactly the tier 0:17:18.040 --> 0:17:20.960 and the compute costs associated to it, and we'll really 0:17:20.960 --> 0:17:24.359 get a sense of the hyperscalar involvement with Opening Eye 0:17:24.480 --> 0:17:26.719 and just the overall reliance on them to take them 0:17:26.760 --> 0:17:29.400 to the next step. And then in relation to SpaceX, 0:17:29.680 --> 0:17:33.119 we're going to get a sliver of the AI aspect 0:17:33.280 --> 0:17:37.760 from XAI. However, that's not going to tell us enough 0:17:38.040 --> 0:17:40.720 of what it takes to be a leading frontier AI model. 0:17:41.200 --> 0:17:43.119 So the public markets are going to be the ones 0:17:43.160 --> 0:17:46.720 to determine how much it truly costs to run an 0:17:46.760 --> 0:17:51.639 AI business. And they have SpaceX as a somewhat of 0:17:51.640 --> 0:17:54.800 a comparison, even though they're they're a space company. 0:17:55.560 --> 0:17:58.920 Harrison wilfirs a pitch book. Thanks for bringing the analysis. 0:17:59.000 --> 0:18:02.719 We appreciate it. Coming up, Apple and bails a series 0:18:02.760 --> 0:18:06.399 of updates at WWDC, but says it isn't able to 0:18:06.480 --> 0:18:09.240 launch them in the EU due to a standoff with 0:18:09.280 --> 0:18:11.159 the blocks anti trust watchdock more on that. 0:18:11.240 --> 0:18:12.600 Next, this is Blomberg Tech. 0:18:26.359 --> 0:18:30.240 Apple used its Worldwide Developers Conference yesterday to lay the 0:18:30.280 --> 0:18:33.360 foundation for the AI era, unveiling a range of products 0:18:33.400 --> 0:18:38.359 including iOS twenty seven SIRIAI. Within that the company still 0:18:38.440 --> 0:18:41.919 has to overcome, though investor skepticism about this AI strategy. 0:18:42.080 --> 0:18:44.840 Bloomberg's Mark German joins us now with more details. 0:18:44.880 --> 0:18:46.600 Look, the skepticism is there. 0:18:46.600 --> 0:18:48.320 If you're looking at the share price, we're down again. 0:18:48.359 --> 0:18:51.480 It traded lower yesterday. Is there some anxiety about the 0:18:51.520 --> 0:18:54.159 pace at which SIRIAI will be unfolded? 0:18:54.160 --> 0:18:54.479 To us? 0:18:54.480 --> 0:18:54.680 All? 0:18:55.880 --> 0:18:58.320 Yeah, thank you for having me. Well, there are concerns 0:18:58.359 --> 0:19:01.440 about the pace of you. They said that it'll launch 0:19:01.440 --> 0:19:05.760 in English later this year. But Serie AI is real. 0:19:06.400 --> 0:19:08.760 If you're on the developer beta. If you're not me, 0:19:08.880 --> 0:19:10.960 I guess you're able to get off the wait lists 0:19:11.680 --> 0:19:14.399 and try these new serie features out. I think the 0:19:14.480 --> 0:19:18.000 response has been mostly positive at this point. As I 0:19:18.040 --> 0:19:20.920 said before the event, what Apple's doing here is they're 0:19:20.960 --> 0:19:25.240 going from completely subpar to completely adequate. We're not seeing 0:19:25.560 --> 0:19:29.440 Wizbang AI features that you're right now seeing from Google, 0:19:29.760 --> 0:19:34.200 Open Ai and Thropic per se, but you're seeing Siri work. 0:19:34.480 --> 0:19:36.920 And so Apple was finally able to do was take 0:19:37.000 --> 0:19:39.840 something that has so much promise for over a decade 0:19:40.200 --> 0:19:43.800 and finally make it work by recreating their underlying models, 0:19:44.160 --> 0:19:47.920 using technology from Google Gemini, using some new AI under 0:19:47.920 --> 0:19:51.399 the hood techniques, and actually presenting users with something they 0:19:51.440 --> 0:19:53.000 think they're going to be able to use on a 0:19:53.040 --> 0:19:56.240 daily basis. And so the concerns that should be out 0:19:56.240 --> 0:19:59.879 there really are what happens to Chat, GPT and Thropic 0:20:00.080 --> 0:20:02.480 on the Apple platform because a lot of people now 0:20:02.480 --> 0:20:05.920 are going to have a real LLM built in, and what. 0:20:05.960 --> 0:20:08.760 Happens to other rivals, maybe even Meta AI. Look, what's 0:20:08.840 --> 0:20:11.200 interesting is, as we've been speaking, there's been some breaking 0:20:11.240 --> 0:20:13.840 news that Meta Platform has been ordered by the EU 0:20:14.359 --> 0:20:17.800 to temporarily help policies that allegedly block AI firms from 0:20:17.840 --> 0:20:20.840 operating on the WhatsApp messaging service for businesses. 0:20:20.880 --> 0:20:21.840 This is the EU once. 0:20:21.680 --> 0:20:24.800 Again basically saying you need to make everything available to 0:20:24.840 --> 0:20:27.080 all your rivals at all times. This is something Apple's 0:20:27.080 --> 0:20:29.320 spoken about. They were sort of munslinging with the EU 0:20:29.359 --> 0:20:32.159 earlier today about whether Siri AI is going to get 0:20:32.160 --> 0:20:35.240 to the EU users slower and whether it's Apple's fault 0:20:35.280 --> 0:20:36.720 or the EU's fault. 0:20:37.440 --> 0:20:40.280 Yeah, I mean, you've seen other features be held back 0:20:40.840 --> 0:20:44.240 due to the negotiations between Apple and the EU, iPhone 0:20:44.240 --> 0:20:47.280 mirroring being one of them. There were a few other 0:20:47.320 --> 0:20:51.320 features on the operating systems last year, apple Intelligence being 0:20:51.359 --> 0:20:54.920 one of them that eventually expanded to the European Union. 0:20:55.000 --> 0:20:58.080 So I'm confident this will eventually work out. But yeah, 0:20:58.119 --> 0:21:00.280 you're seeing a public war of words play out between 0:21:00.280 --> 0:21:02.879