WEBVTT - Amazon Soars, Apple Struggles, AI Faces New Scrutiny 0:00:02.480 --> 0:00:14.440 Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is live 0:00:14.560 --> 0:00:18.239 from the heart of Silicon Valley with Ed Lulow in 0:00:18.440 --> 0:00:20.160 San Francisco. 0:00:22.239 --> 0:00:23.800 This is Bloomberg Tech coming up. 0:00:23.880 --> 0:00:27.159 Tim Cook's final Apple earning sco as CEO is a 0:00:27.240 --> 0:00:31.720 story of supply crunches, demand misreads, and a tough outlook. Plus, 0:00:31.720 --> 0:00:36.400 Amazon shares roaring on AI progress, big spending and accelerating 0:00:36.479 --> 0:00:40.720 cloud growth. And after open Aiyes breaches, it's now anthropics term, 0:00:40.760 --> 0:00:44.640 with the company saying it's AI models breached three organizations 0:00:45.040 --> 0:00:50.080 during cybersecurity tests. We have the details technology earnings and 0:00:50.159 --> 0:00:53.880 a very big reaction in stocks. Apple is on track 0:00:53.960 --> 0:00:58.440 for its biggest drop since March of twenty twenty. Amazon 0:00:58.720 --> 0:01:02.200 is on track for its biggest jump since April of 0:01:02.240 --> 0:01:03.120 twenty twelve. 0:01:03.280 --> 0:01:04.160 Let's get the stories. 0:01:04.200 --> 0:01:07.280 Bloom Bax Ryan Versedica's with us. Let's start with Apple. 0:01:07.560 --> 0:01:10.240 Everything we need to know about Tim Cook's final quarter 0:01:10.560 --> 0:01:11.600 as CEO of Apple. 0:01:13.160 --> 0:01:16.720 Well, I think the major takeaway from Apple's report was 0:01:16.840 --> 0:01:20.800 how much the increase in memory prices is really starting 0:01:20.840 --> 0:01:23.360 to weigh on the company. They talked a lot about 0:01:23.400 --> 0:01:26.880 component shortages. People have been very concerned that the memory 0:01:26.920 --> 0:01:29.920 issue is going to really eat into their margins, potentially 0:01:29.959 --> 0:01:32.400 eat into demand. But as we talked about a couple 0:01:32.400 --> 0:01:35.840 of days ago, the stock hadn't really been reflecting a 0:01:35.880 --> 0:01:38.440 lot of these concerns going into the report. It had 0:01:38.520 --> 0:01:41.240 a very strong month going into it. I think this 0:01:41.400 --> 0:01:44.679 as part of the overall ANTIAI sentiment. Now it does 0:01:44.720 --> 0:01:47.080 seem like that has flipped again and people are sort 0:01:47.080 --> 0:01:50.160 of reassessing some of the issues that Apple is facing, 0:01:50.280 --> 0:01:52.920 even as they continue to maybe take a new look 0:01:53.160 --> 0:01:55.639 at some of the AI names that had been selling 0:01:55.680 --> 0:01:57.040 off prior to this. 0:01:57.760 --> 0:01:59.520 Apple was saying revenue growth in the set time of 0:02:00.240 --> 0:02:02.720 to eleven percent, the street was looking at twelve and 0:02:02.760 --> 0:02:06.240 they also admitted they misjudged demand. They didn't even place 0:02:06.280 --> 0:02:09.760 the right amount of orders for chips on the processors side. Amazon, 0:02:09.919 --> 0:02:13.560 there is growth in aws and there's clearly AI demand. 0:02:13.600 --> 0:02:14.320 What do we need to know? 0:02:15.080 --> 0:02:17.240 Yeah, they also talked about how their chip business is 0:02:17.240 --> 0:02:19.440 starting to see pretty strong revenue. 0:02:19.440 --> 0:02:19.639 There. 0:02:19.880 --> 0:02:22.600 Overall, a very strong report out of Amazon. I think 0:02:22.600 --> 0:02:25.880 it really highlights the divide we've seen this earning season 0:02:25.880 --> 0:02:28.400 between the companies that are able to do a lot 0:02:28.400 --> 0:02:31.200 of AI related spending, but make the case that they 0:02:31.240 --> 0:02:35.440 are justifying that with improved growth, with improved prospects all 0:02:35.480 --> 0:02:38.520 you'd put Amazon and Microsoft into one bucket. Both of 0:02:38.520 --> 0:02:41.760 those companies really had powerful reaction to the results. I 0:02:41.760 --> 0:02:45.640 think people were sort of reassessing their outlooks following the 0:02:45.680 --> 0:02:48.120 strong results from their cloud business, even in the face 0:02:48.160 --> 0:02:50.600 of higher spending. On the other side of that equation, 0:02:50.680 --> 0:02:53.440 you had Alphabet and Meta maybe didn't do as good 0:02:53.480 --> 0:02:55.760 a job justifying how much money they're spending, and you 0:02:55.800 --> 0:02:57.120 saw those stocks come back. 0:02:58.040 --> 0:03:00.680 Amazon did raise capex by twenty billion dollars in part 0:03:00.720 --> 0:03:03.400 because the cost environment's higher, but a lot of AI 0:03:03.480 --> 0:03:06.320 momentumly bits. Ryan Vstellica, Thank you very much to our 0:03:06.360 --> 0:03:10.919 next guest. Amazon set of numbers define game changer. Jordan 0:03:11.000 --> 0:03:14.239 Clin Managing Directs, a sector specialist for tech, media and 0:03:14.280 --> 0:03:17.120 Telecom at Missouri Securities, is with us. You see it 0:03:17.160 --> 0:03:20.079 as a game changer. What was it about Amazon's numbers 0:03:20.080 --> 0:03:20.679 that you liked? 0:03:22.639 --> 0:03:25.480 Well, yeah, I definitely think it was a game changer, 0:03:25.680 --> 0:03:30.120 just because the acceleration in AWS growth on the top 0:03:30.160 --> 0:03:33.560 line at thirty seven percent, I think has really convinced 0:03:33.680 --> 0:03:38.320 more people that AI is you know, improving their overall business, 0:03:39.360 --> 0:03:42.600 and then the disclosure they gave in terms of how 0:03:42.640 --> 0:03:47.600 AI is impacting them on the silicon side. You know, 0:03:47.640 --> 0:03:50.960 with their Gratron and Trainium chip, they're really well positioned. 0:03:51.080 --> 0:03:55.240 So all in all, you know, the bookings growth, the backlog, 0:03:55.400 --> 0:03:58.080 and the commentary form Andy Jassey about what's to come, 0:03:58.600 --> 0:04:03.600 I think set this this to continue to outperform going forward. 0:04:03.600 --> 0:04:04.040 For sure. 0:04:04.880 --> 0:04:07.840 What Amazon said is it's chip business where it rents 0:04:07.880 --> 0:04:11.240 compute capacity, but based on their in house silicon right 0:04:11.280 --> 0:04:13.800 has a revenue run rate twenty five billion dollars. 0:04:14.240 --> 0:04:15.320 Dwell on that a little bit. 0:04:15.360 --> 0:04:17.640 Are you kind of extrapolating out and saying, wow, that's 0:04:17.680 --> 0:04:21.440 another line of business where Amazon can get some upside. 0:04:22.720 --> 0:04:26.240 Yeah, I think it's. It's it's showing how much benefit 0:04:26.279 --> 0:04:29.680 that they get from having their own custom chip. Not 0:04:29.720 --> 0:04:33.120 only does it enable them to get supply right because 0:04:33.120 --> 0:04:37.880 it's so difficult to get pint graphic processors from Nvidia, 0:04:38.720 --> 0:04:42.560 it also enables them to offer that to their customers, 0:04:42.839 --> 0:04:46.800 probably at a lower cost, So you win on both 0:04:46.839 --> 0:04:50.479 sides supply and price or cost. And you saw that 0:04:50.560 --> 0:04:51.800 because their margins are. 0:04:51.680 --> 0:04:52.359 Going up. 0:04:53.680 --> 0:04:56.960 Very healthily, and I think That's also encouraging to investors 0:04:57.360 --> 0:04:59.880 because all these companies are going to be continuing to spend. 0:05:00.000 --> 0:05:03.800 They need more compute, more capacity. But what the worry 0:05:03.839 --> 0:05:06.960 would be is they see compression on the margins or 0:05:06.960 --> 0:05:10.839 the cash flow. And even though Amazon's cash flow was 0:05:10.920 --> 0:05:13.800 negative for the last twelve months, their margins in their 0:05:13.800 --> 0:05:16.440 AWS business are increasingly going higher. 0:05:17.040 --> 0:05:18.839 I think some would point out on the chip business 0:05:18.880 --> 0:05:21.520 as well, that they're not surrendering any margin to Nvidia 0:05:21.640 --> 0:05:23.720 if they go with the in house silicon John, let 0:05:23.720 --> 0:05:25.839 me ask you this, do you have any Amazon Echo 0:05:25.880 --> 0:05:27.520 devices in your household? 0:05:28.240 --> 0:05:28.480 Yeah? 0:05:28.480 --> 0:05:29.000 I have one? 0:05:29.800 --> 0:05:30.160 Okay? 0:05:30.200 --> 0:05:31.839 And how many days a week would you say? You 0:05:31.880 --> 0:05:33.440 have an Amazon Prime package on your. 0:05:33.320 --> 0:05:37.080 Doorstep every other day at least? 0:05:37.480 --> 0:05:37.800 Okay? 0:05:37.960 --> 0:05:41.080 So what I found so interesting were the other data points, 0:05:41.120 --> 0:05:44.240 the non financial right where they say here is some 0:05:44.320 --> 0:05:48.280 evidence that people that use Alexa plus their consumer facing 0:05:48.279 --> 0:05:51.440 AI are more likely to sign up for Prime, are 0:05:51.440 --> 0:05:54.520 more likely to spend in a greater basket size on 0:05:54.560 --> 0:05:57.680 Amazon dot Com. Less talked about this morning. I mean, 0:05:57.680 --> 0:05:59.520 has anyone brought that up on your desk? 0:06:01.080 --> 0:06:01.520 Actually? 0:06:01.880 --> 0:06:04.159 No, I think people are so focused on the AI, 0:06:05.040 --> 0:06:07.640 the data center, the compute side of the business because 0:06:07.640 --> 0:06:11.960 it drives all the profit that they've kind of I 0:06:11.960 --> 0:06:15.320 wouldn't say overlooked, but they're not as concerned or worried 0:06:15.360 --> 0:06:19.120 about e commerce, where they're doing very well, and I 0:06:19.160 --> 0:06:21.200 think that's a good thing. But I agree like you 0:06:21.560 --> 0:06:24.480 would like to see them get more value or credit 0:06:24.520 --> 0:06:29.760 for their breadth, their strength, their traction, their brand and people. 0:06:29.800 --> 0:06:32.920 What they are saying is why does Amazon trade at 0:06:33.080 --> 0:06:35.640 kind of a mid to low twenties forward multiple when 0:06:35.640 --> 0:06:38.960 you look at Costco and Walmart look at their valuations. 0:06:40.240 --> 0:06:42.000 I appreciate you bringing it back to the stock. 0:06:42.080 --> 0:06:45.000 I mean, we're a household where my eco device competes 0:06:45.040 --> 0:06:47.080 with my baby is the alarm clock every morning, and 0:06:47.120 --> 0:06:50.960 we've not really moved beyond that. On the AI side, 0:06:51.120 --> 0:06:54.920 you heard Ryan talk about Apple was the AI safety 0:06:54.920 --> 0:06:58.640 trade in a way in Apple, this is a hell 0:06:58.680 --> 0:07:01.640 of a reaction in the stock. There's no sentimentality here 0:07:02.040 --> 0:07:04.800 for Tim Cook's final earning school. You talk about that 0:07:04.839 --> 0:07:07.160 in your note on the stock in the trade. What 0:07:07.200 --> 0:07:08.120 are you making of all of it? 0:07:10.280 --> 0:07:12.720 Well, I would tell people that the move, you know, 0:07:12.880 --> 0:07:17.160 down nine percent is probably more than the actual results 0:07:17.200 --> 0:07:20.440 would suggest, but that's because the stock has materially outperformed 0:07:20.960 --> 0:07:23.400 Tech and the mag seven over the last thirty days, 0:07:23.400 --> 0:07:25.960 so there was a ton of money hiding in Apple. 0:07:26.080 --> 0:07:30.480 Is like the anti safety AI trade that's going to 0:07:30.520 --> 0:07:33.280 come out and it's going back to Microsoft and Amazon 0:07:33.400 --> 0:07:35.800 and some others. But what I would make of it 0:07:35.880 --> 0:07:39.000 is is that they're going to probably continue to see 0:07:39.080 --> 0:07:42.720 higher risk and struggling a bit, not only in iPhone 0:07:42.760 --> 0:07:46.280 to get components in supply, but their services business missed 0:07:47.000 --> 0:07:49.440 and that's where they get a lot of profit and margin, 0:07:49.560 --> 0:07:52.000 and it sounds like that's going to remain at a 0:07:52.040 --> 0:07:54.080 lower growth rate as you look into the end of 0:07:54.120 --> 0:07:56.680 the year. And my concern, and this is why I 0:07:56.720 --> 0:07:59.600 wouldn't be chasing Apple even down nine percent, is that 0:08:00.120 --> 0:08:04.280 they're going into their iPhone eighteen launch cycle and I 0:08:04.320 --> 0:08:06.880 feel like the risk is that you go into a 0:08:06.920 --> 0:08:10.080 new product cycle and you can't get enough components. 0:08:10.240 --> 0:08:11.000 Okay, what does. 0:08:10.920 --> 0:08:13.280 That mean for your ability to beat and raise numbers? 0:08:13.320 --> 0:08:14.560 I don't think it's good. 0:08:14.960 --> 0:08:16.760 So this is a people story as well, right, Like 0:08:16.840 --> 0:08:20.680 Tim Cook, a cool facet of his career has been 0:08:20.720 --> 0:08:22.840 his ability to manage the supply chain, and he made 0:08:22.920 --> 0:08:27.560 the admission that they misjudged demand, they miscalculated how many 0:08:27.600 --> 0:08:31.680 processes they should order in September in the launch cycle 0:08:31.720 --> 0:08:34.120 for the eighteen He hands the rains to turn Us, 0:08:34.120 --> 0:08:36.360 who is a product guy. I mean, how worrying has 0:08:36.360 --> 0:08:39.720 that been for people phoning you today. 0:08:39.920 --> 0:08:42.400 I don't actually think people are all that nervous or 0:08:42.440 --> 0:08:46.080 worried about the management transition. I think they're going to 0:08:46.080 --> 0:08:48.800 be status quo. They're going to focus on the supply chain. 0:08:48.840 --> 0:08:51.680 That's what they do best. I think they need to 0:08:52.480 --> 0:08:56.320 have a good, flashy launch with this foldable product. Actually, 0:08:56.400 --> 0:08:58.480 that could go a long way to helping turn Us, 0:08:59.200 --> 0:09:01.800 even though that was probably put into place well before 0:09:01.840 --> 0:09:04.360 he took the reins, but it would help, at least 0:09:04.360 --> 0:09:07.480 on the margin, convince people that they've got some new innovations, 0:09:07.520 --> 0:09:10.480 some exciting new products coming. But end of day, it's 0:09:10.520 --> 0:09:13.040 about how much your consumer is going to be willing 0:09:13.080 --> 0:09:14.600 to spend. They're going to raise the price of the 0:09:14.640 --> 0:09:18.280 iPhone probably by two hundred dollars, and they they're doing 0:09:18.320 --> 0:09:21.480 really well. Actually when max where they raise prices and 0:09:21.559 --> 0:09:24.320 they can't ship enough product to meet demand. But the 0:09:24.400 --> 0:09:27.679 question is going to be come early next year after 0:09:27.760 --> 0:09:30.640 the launch quarter you know, in the fall is behind 0:09:30.679 --> 0:09:34.000 them and they've raised prices. Is demand holding up as 0:09:34.080 --> 0:09:37.720 well given the higher costs? And we'll have to wait 0:09:37.720 --> 0:09:40.680 and find out. I think it's it's undetermined right now. 0:09:41.440 --> 0:09:43.880 Again, a decline of nine percent in the session for Zapple, 0:09:43.960 --> 0:09:46.280 on track for its biggest drop since March of twenty twenty. 0:09:46.520 --> 0:09:49.680 Jordan Klein, Managing Directors TMT sector Specialists in a Zoo 0:09:49.960 --> 0:09:52.000 Go and read his end of July. Note there's some 0:09:52.080 --> 0:09:54.719 relief in there, Thank you very much. Now, coming up, 0:09:55.240 --> 0:09:58.280 concerns about Chinese AI startup Moonshot. 0:09:57.920 --> 0:09:59.960 Grow following a Bloomberg report. 0:10:00.000 --> 0:10:03.240 Well, it has a computing power agreement with Ali Barber 0:10:03.760 --> 0:10:06.320 which uses or relies on Nvidia chips. We have the 0:10:06.360 --> 0:10:13.400 details next. This is Bloomberg Tech. Chinese AI champion Moonshot 0:10:13.600 --> 0:10:16.760 has a computing power agreement with Ali Barber for the 0:10:16.840 --> 0:10:21.360 use of around twenty thousand Nvidia chips, according to sources. 0:10:21.679 --> 0:10:24.760 Those sources say that the hardware forms a substantial chunk 0:10:24.960 --> 0:10:28.080 that the computing capacity Moonshot uses for its Kimmi models, 0:10:28.120 --> 0:10:31.520 the latest of which rival US competitors. You'll remember, of course, 0:10:31.720 --> 0:10:35.240 that Washington had restricted Chinese access to some generations of 0:10:35.320 --> 0:10:38.920 Nvidia chip on national security grounds. Bloomberg's executive editor Peter 0:10:39.000 --> 0:10:41.720 Elstrich joins us, let's start with the basics of what 0:10:41.760 --> 0:10:42.680 we're reporting here. 0:10:44.640 --> 0:10:47.160 Yeah, well, just to back up one step, of course, 0:10:47.200 --> 0:10:51.280 Moonshot announced itself really to the Western world at least 0:10:51.640 --> 0:10:53.560 a couple of weeks ago, where it came out with 0:10:53.600 --> 0:10:57.160 this Kimmy K three model which has very strong performance, 0:10:57.960 --> 0:11:00.840 really near the leading edge of the frontier lane in 0:11:00.920 --> 0:11:03.880 the models at open Ai and Anthropic, And one of 0:11:03.880 --> 0:11:05.839 the questions since then has been how did they do it? 0:11:06.040 --> 0:11:09.040 We know that there's been breakthrough engineering at a number 0:11:09.040 --> 0:11:11.439 of the different Chinese companies. Of course, Deep Seek was 0:11:11.480 --> 0:11:13.760 really the phenomenon last year, but now the questions are 0:11:13.800 --> 0:11:16.440 around Moonshot and how did they actually pull this off? 0:11:16.679 --> 0:11:18.840 And we've seen a couple of bits and pieces of it, 0:11:18.880 --> 0:11:21.200 but this kind of helps us understand the broader picture. 0:11:21.520 --> 0:11:24.400 We know that Ali Baba has an investment in Moonshot, 0:11:24.520 --> 0:11:26.840 and what we were reporting in this story is that 0:11:26.880 --> 0:11:29.360 Ali Baba is also supplying a bunch of the computing 0:11:29.400 --> 0:11:32.200 power behind that model. We know that they're using twenty 0:11:32.200 --> 0:11:37.520 thousand of Nvidia chips. The Hopper generation of chips from Nvidia. 0:11:37.559 --> 0:11:39.440 Alibaba did get back to us and they said they're 0:11:39.480 --> 0:11:42.440 not that H two hundred chips which are the leading edge. 0:11:42.559 --> 0:11:44.840 Our sources have told us that they actually are those 0:11:45.040 --> 0:11:48.240 H two hundred chips, but they're definitely the Hopper generation 0:11:48.320 --> 0:11:50.680 the chips that they're providing, and Moonshot has been able 0:11:50.720 --> 0:11:52.920 to use that to build out this model. We also, 0:11:52.960 --> 0:11:56.000 of course had the White House Micratzios saying that they 0:11:56.000 --> 0:11:59.400 believe they're also accessing Grace Blackwell chips even more advanced 0:11:59.440 --> 0:11:59.880 chips from it. 0:12:01.360 --> 0:12:02.840 We're just showing that sort of. 0:12:04.520 --> 0:12:08.080 Communication from Alie Barbara and what they are denying and 0:12:08.120 --> 0:12:10.320 what they're not. I guess that, you know, it's worth 0:12:10.320 --> 0:12:12.720 a discussion, Peter, of what we don't know right. For 0:12:12.760 --> 0:12:16.800 a long time, we've been asking about, well, how many 0:12:17.280 --> 0:12:19.920 in video chips are actually going into China? Are Chinese 0:12:19.920 --> 0:12:23.640 firms using this technology outside of the United States and 0:12:23.679 --> 0:12:26.679 outside of China post some of those questions for us. 0:12:27.960 --> 0:12:30.680 Yeah, it's a complicated picture, and we do have this 0:12:30.800 --> 0:12:35.360 shifting regime of export regulations that's affected a bunch of this. 0:12:35.720 --> 0:12:38.600 In Video is not allowed to sell its most advanced chips, 0:12:38.600 --> 0:12:41.200 the Blackwell chips into China right now. What we have 0:12:41.280 --> 0:12:43.280 seen in the past is they were able to sell 0:12:43.320 --> 0:12:45.800 a number of the Hopper generation of chips, some of 0:12:45.800 --> 0:12:49.440 the earlier generations of chips. More recently, the US signed 0:12:49.480 --> 0:12:52.800 off on in video selling those H two hundreds into China. 0:12:52.840 --> 0:12:55.280 That would be a pretty advanced chip for many of 0:12:55.280 --> 0:12:58.560 these companies to use, including Moonshot and deep Seek. But 0:12:58.800 --> 0:13:02.440 Beijing is pushed back on that idea that domestic companies 0:13:02.440 --> 0:13:05.280 should use those chips. It would prefer to continue building 0:13:05.280 --> 0:13:09.840 out its own semiconductor industry, particularly supporting companies like Huawei 0:13:09.840 --> 0:13:13.000 and Cambra Con in particular. So if you are using 0:13:13.120 --> 0:13:16.080 H two hundred chips, that would be it's possible from 0:13:16.080 --> 0:13:19.360 the US standpoint, within within China, but there would be 0:13:19.400 --> 0:13:21.520 a little bit of pushback from Beijing if you're using 0:13:21.640 --> 0:13:24.000 H two hundred ships that they would prefer you not use. 0:13:24.040 --> 0:13:27.080 At this point, just to flag one other loophole here, 0:13:27.800 --> 0:13:31.440 it is still entirely possible, entirely legal for Chinese companies 0:13:31.480 --> 0:13:34.880 to access all of Nvidia's chips outside of the country. 0:13:34.920 --> 0:13:36.960 They can go to Southeast Asia and they can get 0:13:37.040 --> 0:13:40.040 access to some of these chips, including the H two 0:13:40.120 --> 0:13:43.880 hundreds there. That's not forbidden in any case. This is 0:13:43.920 --> 0:13:45.880 a loophole that some of the China Hawks have wanted 0:13:45.880 --> 0:13:47.720 to close for a long period of time. But at 0:13:47.800 --> 0:13:50.640 least at this period of time, you can't still access there. 0:13:50.800 --> 0:13:52.360 We have a Bloomberg terminal client. 0:13:52.400 --> 0:13:53.920 Just send me an ib and make a point that's 0:13:53.960 --> 0:13:57.400 in the story that Alibaba's one of Moonshot's biggest investors, 0:13:57.760 --> 0:14:00.320 and of course they're also competitors like go Away. Read it, 0:14:00.320 --> 0:14:03.520 Bloomberg's Peter Elstrom, Thank you very much. Indeed, m Fropic 0:14:03.880 --> 0:14:06.760 says its AI models broke out to the closed testing 0:14:06.880 --> 0:14:11.240 environment and breached three outside organizations, and Forropic said it 0:14:11.280 --> 0:14:15.760 made the discovery after reviewing its own cybersecurity tests following 0:14:15.840 --> 0:14:18.559 open AI's announcement of a similar incident. Let's get the 0:14:18.640 --> 0:14:21.800 latest with Bloomboz Patrick Howe O'Neil one of the most 0:14:21.880 --> 0:14:26.080 read stories unsurprisingly on Bloomberg. Tell me everything that we're 0:14:26.120 --> 0:14:26.960 reporting right now. 0:14:28.800 --> 0:14:29.760 Yeah, happy to be here. 0:14:30.120 --> 0:14:33.640 So, as you mentioned, Anthropic began an internal review of 0:14:33.760 --> 0:14:38.160 their cybersecurity tests after the hugging face incident was disclosed 0:14:39.080 --> 0:14:42.040 a little over a week ago. And discovered that in 0:14:42.160 --> 0:14:46.720 three instances, their models, which were conducting their own internal tests, 0:14:46.800 --> 0:14:51.160 had broke out and had hacked three external organizations. 0:14:51.920 --> 0:14:54.440 I think it raises a ton of really interesting. 0:14:54.120 --> 0:14:59.440 And important questions about these AI companies' abilities to control, understand, 0:14:59.480 --> 0:15:03.360 and monitor these models, and especially the fact that it took, 0:15:03.440 --> 0:15:06.640 in at least one case, three months for Anthropic to 0:15:06.760 --> 0:15:09.800 discover what had happened, which just raises questions about the 0:15:09.840 --> 0:15:11.240 ability to really see what's going on. 0:15:12.040 --> 0:15:12.400 Patrick. 0:15:12.400 --> 0:15:14.400 In the Anthropic case, we don't know who the three 0:15:14.440 --> 0:15:16.480 companies are, right, we do know that there was a 0:15:16.520 --> 0:15:19.200 connection to the Internet escaping the sandboxed environment. 0:15:19.440 --> 0:15:22.680 Go over all that real quick, That's right. 0:15:22.800 --> 0:15:25.840 So there's an important distinction to be drawn here. In 0:15:25.920 --> 0:15:28.720 the recent open ai case, which is obviously drawing a 0:15:28.760 --> 0:15:32.880 lot of comparisons, open AI's model discovered novel vulnerabilities and 0:15:32.920 --> 0:15:36.000 hacked its way out of its environment. In the Anthropic case, 0:15:36.160 --> 0:15:40.520 there was a human misconfiguration that allowed Internet access for 0:15:40.600 --> 0:15:43.520 the model, and therefore the model thought that everything it 0:15:43.560 --> 0:15:45.760 had accessed, at least at the beginning was part of 0:15:45.800 --> 0:15:48.120 the test. And this is something that Anthropic is really 0:15:48.160 --> 0:15:52.880 insistent on because they're positioning this as primarily a human 0:15:53.000 --> 0:15:56.000 error or human series of errors, as opposed to an air. 0:15:57.360 --> 0:15:59.960 I invited anthropic on the show. They said no. 0:16:00.080 --> 0:16:05.360 Bloomberg's Patrick how O'Neil, thank you very much. Indeed, troubled 0:16:05.360 --> 0:16:09.760 hedge funds, Situational Awareness explored selling stakes in private companies 0:16:09.800 --> 0:16:12.440 to raise cash. Is It faced a wave of margin 0:16:12.520 --> 0:16:16.200 calls from lenders this week. Its assets plunged from forty 0:16:16.200 --> 0:16:19.000 five billion dollars to roughly ten billion dollars. Is AI 0:16:19.080 --> 0:16:22.400 stocks got hit in recent weeks. That's when Citadel's King 0:16:22.440 --> 0:16:24.960 Griffin swooped in to buy a bulk of the fund's 0:16:25.200 --> 0:16:29.160 public portfolio. You're seeing pictures of Leo Bold Ashan Brenner, 0:16:29.360 --> 0:16:34.320 founder of Awareness, coming from the website for our posterity. 0:16:34.800 --> 0:16:37.240 Let's get the details, Bloomberg Hedge Fund reporter Hema Palmer. 0:16:37.520 --> 0:16:41.680 It's with us extraordinary reporting. The new bit is trying 0:16:41.720 --> 0:16:45.520 to offload some private steak to continue to manage the situation. 0:16:46.320 --> 0:16:49.960 So what we were seeing is really incredibly dramatic unfolding 0:16:49.960 --> 0:16:52.680 of events over the past twenty four hours. And what 0:16:52.760 --> 0:16:57.200 we reported last night was the news that while Situational 0:16:57.240 --> 0:17:01.600 Awareness was in conversations with Citadel, Millennium Jane Street to 0:17:01.640 --> 0:17:05.280 offload their public book. They were also having conversations to 0:17:05.359 --> 0:17:09.680 potentially offload their private portfolio. This is a sizeable portfolio 0:17:09.680 --> 0:17:13.840 that includes a five billion dollar position in Anthropic and 0:17:13.920 --> 0:17:17.600 the fact that they were in talks with Sequoia and 0:17:17.760 --> 0:17:21.159 Green Oaks to potentially offload those stakes really shows how 0:17:21.240 --> 0:17:25.280 dramatic the situation was. Ultimately, they did not go ahead 0:17:25.320 --> 0:17:28.840 with that because they got the Citadel deal. But it 0:17:28.920 --> 0:17:31.119 is not a great sign when a fund is looking 0:17:31.200 --> 0:17:34.240 at trying to sell their public and their private book 0:17:34.320 --> 0:17:35.119 to meet margin. 0:17:35.160 --> 0:17:37.280 Calls Heimer, what's the state of play? 0:17:37.320 --> 0:17:39.560 It sounds like the firm is still there and has 0:17:39.600 --> 0:17:44.119 its private book. Has anyone heard from Leopold Ashton Brenner 0:17:44.320 --> 0:17:45.480 in the last twenty four hours. 0:17:45.560 --> 0:17:49.200 Yes, he's sent a letter to investors apologizing for the 0:17:49.280 --> 0:17:53.639 dramatic unfolding of events, saying that he is responsible, saying 0:17:53.760 --> 0:17:58.520 that the fund still has its ur date or since 0:17:58.560 --> 0:18:02.119 inception performance of up eighty percent, but keep in mind 0:18:02.640 --> 0:18:07.960 the monthly performance down nearly seventy percent. Also really important 0:18:08.359 --> 0:18:11.359 is thinking about when investors got in the LPs that 0:18:11.440 --> 0:18:14.959 got in early in the funds existence they're probably doing okay, 0:18:15.240 --> 0:18:16.919 but if you're an investor and you got in the 0:18:16.920 --> 0:18:19.359 past few months, the pain is pretty bad. 0:18:20.480 --> 0:18:24.159 Bloomberg's Hemma Palmer, thank you very much. Indeed, Tesla is 0:18:24.240 --> 0:18:28.119 considering separating its China business to pave the way for 0:18:28.160 --> 0:18:31.840 a potential SpaceX merger, all according to a report in 0:18:31.840 --> 0:18:34.600 the Wall Street Journal. While the journal said some Tesla 0:18:34.680 --> 0:18:38.160 execs have been told to prepare for that move, CEO 0:18:38.240 --> 0:18:41.520 Elon Musk wrote on X the report was quote fake news, 0:18:42.119 --> 0:18:45.560 joining us now, Bloomberg Managing editor Break Trudel. Let's start 0:18:45.560 --> 0:18:48.960 with what is in the Wall Street Journal report. There 0:18:48.960 --> 0:18:54.159 are multiple lines of information about what's being considered summarized 0:18:54.200 --> 0:18:54.880 as best you can. 0:18:55.520 --> 0:18:58.840 Yeah, I think in the sort of broad scheme of things, 0:18:59.160 --> 0:19:01.600 the I've have been you know, for months now on 0:19:02.000 --> 0:19:03.879 is you know, SpaceX and Tesla. 0:19:04.000 --> 0:19:05.479 Are these companies going to combine? 0:19:05.480 --> 0:19:07.840 And one of the big questions has been how did 0:19:07.880 --> 0:19:11.719 these companies sort of come together? Considering the fact that 0:19:11.760 --> 0:19:15.520 SpaceX is this great, big, you know, US defense contractor, 0:19:15.960 --> 0:19:19.320 and Tesla has this you know, huge China business that 0:19:20.040 --> 0:19:22.840 is crucial to the company. Aren't those two things sort 0:19:22.840 --> 0:19:25.600 of in conflict? I think you know, what was you know, 0:19:25.800 --> 0:19:28.800 very interesting about this report last night was, Yes, it 0:19:28.880 --> 0:19:32.040 laid out some elements of the fact that, you know, 0:19:32.080 --> 0:19:34.880 these companies already are kind of separate. But I think 0:19:34.880 --> 0:19:37.640 that's pretty you know, well known and understood. I think 0:19:37.440 --> 0:19:41.760 the question is, you know, what exactly are advisors thinking 0:19:41.880 --> 0:19:45.280 or how is this possibly going to work out considering 0:19:45.320 --> 0:19:49.520 the fact that that the Chinese Communist Party probably is 0:19:49.560 --> 0:19:53.359 not particularly client for the Shanghai factory that's so crucial 0:19:53.400 --> 0:19:56.639 to Tesla sort of closing up to you know, the 0:19:56.680 --> 0:20:01.520 SpaceX business or at all sort of overlapping wa with SpaceX, 0:20:01.560 --> 0:20:04.520 which you know, Musk has been quite open about, you know, 0:20:04.560 --> 0:20:06.760 not being crazy about Starlink exactly. 0:20:06.800 --> 0:20:09.040 So you know, Musk posted more than once this is 0:20:09.119 --> 0:20:11.000 one of the posts and xx this has never even 0:20:11.040 --> 0:20:14.440 come up in a discussion ever, absurdly fake news. People 0:20:14.480 --> 0:20:19.480 should assume uses fake until proven otherwise. I think our 0:20:19.520 --> 0:20:23.439 colleagues in China were also able to make contact with 0:20:23.480 --> 0:20:25.800 Tesla China, who disputed the report. 0:20:25.880 --> 0:20:27.080 Right, that's correct. 0:20:27.119 --> 0:20:30.440 And you know, look, I think the fact that Musk 0:20:30.520 --> 0:20:32.480 has been sort of, you know, very public about the 0:20:32.520 --> 0:20:35.760 fact that Starlink is a no Goo in China has 0:20:35.800 --> 0:20:39.640 also talked quite openly about, you know, just how much China. 0:20:39.359 --> 0:20:40.959 Is competing with SpaceX. 0:20:41.119 --> 0:20:43.399 I think there are a lot of elements of the 0:20:43.480 --> 0:20:46.040 journals reporting that sort of you know, ring true. The 0:20:46.160 --> 0:20:48.720 question is, of course, you know, what's going on behind 0:20:48.720 --> 0:20:52.120 the scenes for these these companies to sort of game 0:20:52.160 --> 0:20:54.800 out what could be done. I think, you know, the 0:20:55.119 --> 0:20:57.920 journal you know, sort of brought some new things to 0:20:58.040 --> 0:21:01.960 light last night. Of course, Sir Muskin his campus pushing 0:21:02.000 --> 0:21:04.240 back and what was reported being. 0:21:04.160 --> 0:21:07.200 That's great, Trudell, thank you very much. Let's stick with muss. 0:21:07.240 --> 0:21:10.040