WEBVTT - Oracle’s Cloud Growth; Debate Around AI Risks 0:00:02.520 --> 0:00:04.210 Bloomberg Audio Studios. 0:00:04.690 --> 0:00:06.210 Podcasts. Radio. 0:00:06.650 --> 0:00:07.050 News. 0:00:13.600 --> 0:00:17.140 Bloomberg Tech is live from the heart of Silicon Valley 0:00:17.660 --> 0:00:19.939 with Ed Ludlow in San Francisco. 0:00:22.730 --> 0:00:23.820 This is Bloomberg Tech. 0:00:23.860 --> 0:00:26.960 Coming up, Oracle's latest results help ease concerns around its 0:00:27.100 --> 0:00:31.740 massive AI infrastructure build-out, plus the debate around AI's societal 0:00:31.800 --> 0:00:34.820 risks grow louder, with the Pentagon's Emile Michael now pushing 0:00:34.900 --> 0:00:38.000 back against those fears. And we'll discuss that debate, as 0:00:38.020 --> 0:00:40.339 well as the data center build-out backlash, and much more. 0:00:40.620 --> 0:00:44.309 with the CEO of venture capital firm Eclipse, Lior Susan. 0:00:44.810 --> 0:00:46.990 Let's get right to Oracle. It's our main technology and 0:00:47.030 --> 0:00:49.769 earnings story. And actually, the stock right now completely flat. 0:00:50.090 --> 0:00:51.790 After the open, it hit a session high of more 0:00:51.810 --> 0:00:54.840 than 8%, then went to negative territory. I want to 0:00:54.860 --> 0:00:57.100 go through some of the big numbers. Cloud infrastructure in 0:00:57.120 --> 0:00:58.820 the quarter, $ 7. 4 billion, a gain of 121% year on year. 0:00:58.840 --> 0:01:00.260 The backlog, as measured by RPO, $ 664 billion. But CapEx 0:01:00.380 --> 0:01:10.190 in that quarter that they reported... $ 28. 5 billion. Was that 0:01:10.209 --> 0:01:13.110 the focus? They raised the full-year outlook for both revenue 0:01:13.350 --> 0:01:15.970 and adjusted EPS. Let's get to Bloomberg's Brody Ford, who 0:01:16.010 --> 0:01:17.050 leads our coverage of Oracle. 0:01:17.230 --> 0:01:17.789 What's going on? 0:01:19.010 --> 0:01:22.110 Seeing the share price gain ease this morning, I would 0:01:22.160 --> 0:01:24.820 think it has to be the big spending in the quarter. 0:01:24.840 --> 0:01:27.600 That's the CapEx number. Right, the CapEx in the quarter. 0:01:27.620 --> 0:01:29.819 It came in higher than expected, and the margins went 0:01:29.900 --> 0:01:33.700 down because cloud infrastructure is a much less profitable business 0:01:33.720 --> 0:01:35.380 than their traditional database business. 0:01:35.720 --> 0:01:38.300 But overall, it was a very clean quarter. 0:01:38.380 --> 0:01:41.140 It pretty much showed that they are delivering the capacity 0:01:41.160 --> 0:01:44.660 they promised. Revenue is ramping up. In general, these were 0:01:44.680 --> 0:01:46.900 the figures you wanted to see if you're an investor. 0:01:47.360 --> 0:01:49.180 I'm going to get through some of the numbers, then 0:01:49.200 --> 0:01:50.860 we'll put the numbers to one side. And that is 0:01:51.220 --> 0:01:54.070 debt on the balance sheet is $ 120 billion. Cash is 0:01:54.090 --> 0:01:57.650 about $ 36 billion. So in a leverage context, they're at 3.4x, which, 0:01:57.990 --> 0:01:59.990 I don't know, negative free cash flow is better than 0:02:00.170 --> 0:02:01.190 expected in the quarter, right? 0:02:01.770 --> 0:02:02.410 It's funny, right? 0:02:02.450 --> 0:02:05.270 I mean, negative free cash flow doesn't sound good, but 0:02:05.310 --> 0:02:06.950 it wasn't as negative as people expected. 0:02:06.970 --> 0:02:10.780 Yes, not as bad as expected. Let's put the numbers 0:02:10.800 --> 0:02:13.180 to one side. What is the story for Oracle? They're 0:02:13.200 --> 0:02:16.660 trying to build data centers fast and get that capacity online. 0:02:16.720 --> 0:02:17.620 Is it as simple as that? 0:02:17.780 --> 0:02:20.910 Yeah, they're trying to transition from a traditional software business 0:02:20.990 --> 0:02:24.450 to an AI data center for developer. And the question is, 0:02:24.630 --> 0:02:26.510 is that a good business to be in? And they've 0:02:26.550 --> 0:02:29.670 become a big kind of focus on this question of, 0:02:30.230 --> 0:02:32.750 You know, is it worth going negative free cash flow 0:02:32.790 --> 0:02:34.359 for a couple of years based on a couple of 0:02:34.419 --> 0:02:37.680 huge AI contracts? And so folks are watching each quarter 0:02:37.720 --> 0:02:40.139 very closely for any tidbits on that. 0:02:40.460 --> 0:02:43.049 I spoke to Sam Altman, the OpenAI CEO, last week 0:02:43.110 --> 0:02:46.130 when Astra was released. And it was very interesting to 0:02:46.169 --> 0:02:50.130 see Oracle disclose that. one of the data central sites 0:02:50.160 --> 0:02:52.380 that they have online was very key to the training 0:02:52.400 --> 0:02:52.800 of Astra. 0:02:52.840 --> 0:02:53.560 What did they say? 0:02:53.860 --> 0:02:54.679 Very interesting. 0:02:54.700 --> 0:02:58.240 Yeah, their flagship data center with OpenAI out in West Texas, 0:02:58.300 --> 0:02:58.880 they said that. 0:02:58.919 --> 0:02:59.559 This is Abilene? 0:02:59.620 --> 0:03:02.780 In Abilene, right. They said that OpenAI's newest model was 0:03:02.820 --> 0:03:03.280 trained there. 0:03:03.760 --> 0:03:07.290 And it just shows that providing the infrastructure here, you know, 0:03:07.389 --> 0:03:10.450 some might deride it as a commodity business, but it's 0:03:10.510 --> 0:03:12.370 essential to some of the big AI labs. 0:03:13.530 --> 0:03:14.890 Let's end on the four-year outlook. 0:03:14.910 --> 0:03:21.650 So they're saying at least $ 90 billion of revenue... Adjusted EPS, $ 8. 10, which, 0:03:21.790 --> 0:03:24.680 you know, that's a bit of a hike. But how 0:03:24.740 --> 0:03:27.120 much of this will focus on the conversion of what 0:03:27.290 --> 0:03:31.060 the remaining performance obligation backlog into revenue? 0:03:31.139 --> 0:03:33.100 How closely does the street look at that? 0:03:33.540 --> 0:03:35.600 Yeah, the big question is how quick can you turn 0:03:35.640 --> 0:03:38.900 that RPO into revenue, especially in light of some of 0:03:38.940 --> 0:03:41.440 the data center pushback that you mentioned? They've had a 0:03:41.480 --> 0:03:44.620 couple of big sites with some tough headlines around NIMBYism. 0:03:44.660 --> 0:03:45.540 Local pushback. 0:03:45.840 --> 0:03:48.620 It's a recurring question is, is that going to delay projects? 0:03:48.920 --> 0:03:51.340 Bloomberg's Brody Ford, who's had a very busy week in 0:03:51.360 --> 0:03:52.500 the technology earnings context. 0:03:52.520 --> 0:03:53.600 We're going to stay with Oracle. 0:03:53.690 --> 0:03:57.250 Joining us now is Stefan Slewinski, BNP Paribas Global Head 0:03:57.570 --> 0:04:00.490 of Software Research. He hasn't outperformed on the stock, but 0:04:00.530 --> 0:04:02.780 actually lowered his price target from $ 290 to $ 248. 0:04:02.710 --> 0:04:04.830 Let's start there, Stefan. Why maintain the call but lower 0:04:04.850 --> 0:04:05.410 the price target? 0:04:12.880 --> 0:04:13.400 Yeah, hi, Ed. 0:04:13.680 --> 0:04:14.140 Hi, Brody. 0:04:14.460 --> 0:04:17.400 Good to be with you today. So, as Brody was saying, look, 0:04:17.440 --> 0:04:19.420 it was a pretty good set of results across the board. 0:04:19.460 --> 0:04:22.560 There were some nitpicks, which we can go into. But, 0:04:23.020 --> 0:04:24.740 you know, we did lower our price target a bit today. 0:04:24.760 --> 0:04:26.420 You know, we have seen a pullback in some of 0:04:26.440 --> 0:04:28.520 the multiples in software over the last couple of days. 0:04:29.060 --> 0:04:31.830 So from a relative standpoint, we've taken the price target down. 0:04:32.089 --> 0:04:33.430 But we still see a lot of upside here. 0:04:33.470 --> 0:04:34.289 This is a company that's. 0:04:34.170 --> 0:04:37.710 Going to be growing earnings at 25%, 30% over the 0:04:37.750 --> 0:04:41.140 coming couple of years, trading on 20 times earnings currently. 0:04:41.580 --> 0:04:43.180 If you do give them the credit and you look 0:04:43.279 --> 0:04:44.839 out towards 2029, 2030, you're looking at about a seven 0:04:44.860 --> 0:04:49.110 times P.E., So if they can keep delivering, they can 0:04:49.130 --> 0:04:51.969 keep bringing capacity online, you can keep seeing the OCI growth. 0:04:52.350 --> 0:04:55.349 And importantly, if those margins and cash flows come through, 0:04:55.390 --> 0:04:59.090 which is where the market's still questioning the strategy, then 0:04:59.330 --> 0:05:00.429 the equity should perform. 0:05:01.470 --> 0:05:03.430 Well, Stefan, let's go into those nitpicks. 0:05:03.470 --> 0:05:06.510 We're showing that the trading of the day, we're now 0:05:06.570 --> 0:05:09.090 down modestly four tenths of a percent. This is a 0:05:09.130 --> 0:05:11.930 stock that early in the session was up eight and 0:05:11.970 --> 0:05:14.789 a half percent. What's changed in the last two hours 0:05:15.250 --> 0:05:16.330 to the mind of investors? 0:05:18.020 --> 0:05:19.539 Yeah, I think there was even though it's a good 0:05:19.560 --> 0:05:21.789 set of results, there was no knockout punch. I think 0:05:21.820 --> 0:05:24.470 investors maybe were hoping for a bigger bounce. The implied 0:05:24.490 --> 0:05:26.930 volatility showed a 10 percent move on the day of results. 0:05:27.010 --> 0:05:28.990 And when we didn't get that, you know, I think 0:05:29.029 --> 0:05:32.730 some people maybe took their chips off the table. Also, 0:05:33.510 --> 0:05:38.100 you know, we're waiting for more discussion around capital requirements. 0:05:38.930 --> 0:05:41.450 Oracle have said that they need an extra $ 20 billion 0:05:41.490 --> 0:05:44.430 of capital this fiscal year. We think that probably comes 0:05:44.490 --> 0:05:47.290 beginning of next calendar year. They kind of punted any 0:05:47.330 --> 0:05:49.750 discussion around that to the investor day, which will come 0:05:49.790 --> 0:05:50.670 at the end of October. 0:05:51.029 --> 0:05:51.670 That'll be key. 0:05:51.750 --> 0:05:54.330 You have a new CFO, and we think that she 0:05:54.370 --> 0:05:57.920 will give commentary around what their capital requirements will be 0:05:58.010 --> 0:06:00.400 beyond this fiscal year. That could be a real catalyst 0:06:00.420 --> 0:06:01.900 for the stock, but we didn't get that last night. 0:06:03.420 --> 0:06:05.400 I think we should talk about leverage. I was talking 0:06:05.440 --> 0:06:07.690 about it with our guys at Bloomberg Intelligence this morning. 0:06:07.710 --> 0:06:09.890 You know, they have $ 120 billion of debt on the 0:06:09.930 --> 0:06:14.590 balance sheet. Cash and cash equips, $ 36 billion. A negative 0:06:14.610 --> 0:06:17.750 free cash flow was not as bad as expected. But 0:06:17.800 --> 0:06:21.760 everyone still seems pretty calm about that. If you compare 0:06:22.460 --> 0:06:25.140 that rate of leverage to other hyperscalers or the average 0:06:25.620 --> 0:06:27.359 unadjusted base in the S & P 500, how do 0:06:27.400 --> 0:06:29.060 you feel about it? I get you're looking at this 0:06:29.080 --> 0:06:30.390 from the equity side of the story. 0:06:32.100 --> 0:06:34.000 Yeah, but even on the debt side, we've seen over 0:06:34.020 --> 0:06:36.440 the last couple of weeks with the arrival of Astra 0:06:36.500 --> 0:06:39.550 from OpenAI, that's been well received. As you pointed out, 0:06:39.589 --> 0:06:42.450 that was trained at Abilene, Texas on Oracle infrastructure. 0:06:43.410 --> 0:06:45.650 OpenAI hitting $ 45. 0:06:44.650 --> 0:06:48.170 Billion of ARR. Sentiment around OpenAI has improved a bit. Therefore, 0:06:48.230 --> 0:06:50.630 sentiment around Oracle has improved. So we've actually seen the 0:06:50.690 --> 0:06:53.080 CDSs come in a bit, the credit come in a 0:06:53.100 --> 0:06:55.900 bit for Oracle as well. there's been a bit more 0:06:55.940 --> 0:06:59.980 confidence importantly the company has already said that capex will 0:07:00.000 --> 0:07:02.970 be peaking this year and next year and therefore free 0:07:02.990 --> 0:07:06.690 cash flow will inflect higher really in fiscal 29 so 0:07:06.730 --> 0:07:09.360 we actually see them heading back towards free cash flow 0:07:09.400 --> 0:07:11.280 positive If you think about it, you look at all 0:07:11.300 --> 0:07:15.240 the hyperscalers, Microsoft, Amazon, Google, Meta, Microsoft has said they'll 0:07:15.260 --> 0:07:18.040 stay free cash flow positive. But for the others, we're 0:07:18.060 --> 0:07:21.040 heading down into negative free cash flow territory. For Oracle, 0:07:21.420 --> 0:07:23.300 there's light at the end of the tunnel. We should 0:07:23.340 --> 0:07:26.210 be inflecting back towards positive free cash flow in the 0:07:26.250 --> 0:07:29.230 next couple of years. But again, the market wants to 0:07:29.290 --> 0:07:30.980 see more proof before it believes it. 0:07:31.890 --> 0:07:35.170 So you took us to tokenomics. I find this fascinating. 0:07:36.210 --> 0:07:40.980 Astra was trained in the Abilene site. And if Astra 0:07:41.160 --> 0:07:43.980 is a commercial success, frankly, it's run from a token 0:07:44.000 --> 0:07:44.790 generation perspective. 0:07:44.820 --> 0:07:46.320 For Oracle, that's good, right? 0:07:48.430 --> 0:07:53.230 I think OpenAI with a $ 300 billion commitment to Oracle 0:07:53.310 --> 0:07:56.770 over the next four to five years, obviously, if they're 0:07:56.790 --> 0:07:59.210 being successful with their models and they're being successful as 0:07:59.250 --> 0:08:05.070 a business, that makes them a good customer to have. Also, 0:08:05.130 --> 0:08:10.270 it's a proof of concept of Oracle's infrastructure that it works. Also, 0:08:10.290 --> 0:08:12.600 you know, we are seeing Oracle in the quarter. 0:08:12.630 --> 0:08:14.960 They had $ 10 billion of prepayment. 0:08:15.440 --> 0:08:20.280 So customers signing new contracts, they added $ 26. 0:08:19.040 --> 0:08:20.360 Billion to their backlog. 0:08:20.740 --> 0:08:23.300 And a lot of that was customers paying up front 0:08:23.520 --> 0:08:27.010 for the CapEx. So again, that means that Oracle needs 0:08:27.030 --> 0:08:29.610 to spend less of their own capital. They need to 0:08:29.670 --> 0:08:33.350 raise less capital from the market in order to finance that. 0:08:33.710 --> 0:08:35.729 It also is a validation of what they're providing. They're 0:08:35.750 --> 0:08:37.880 not just providing balance sheet of a service. They're not 0:08:37.890 --> 0:08:41.439 just a bank. They are providing services around the GPUs 0:08:41.780 --> 0:08:45.000 that customers find valuable, and they're able to monetize that. 0:08:46.700 --> 0:08:48.760 Stefan, you're Global Head of Software Research, and there was 0:08:48.800 --> 0:08:51.569 another earnings out overnight, which is Adobe. They gave a 0:08:51.670 --> 0:08:54.309 forecast that You know, there was a range on the street, 0:08:54.350 --> 0:08:57.510 but it basically is disappointed. What's the Adobe story here 0:08:57.550 --> 0:08:57.959 to your mind? 0:08:59.429 --> 0:09:02.250 Yeah, Adobe, we're finally starting to see the cracks that 0:09:02.429 --> 0:09:04.390 we've been expecting over the last couple of years, right? 0:09:04.429 --> 0:09:09.500 So the RPO, the backlog, actually declined quarter of a quarter, 0:09:09.530 --> 0:09:13.200 which is very rare for Q3. The net new ARR 0:09:13.460 --> 0:09:16.319 was down 40% almost year over year. 0:09:16.980 --> 0:09:20.400 The core creative apps slowed to around 8% growth. So 0:09:20.800 --> 0:09:21.780 those concerns. 0:09:21.440 --> 0:09:23.140 That we've been worried about for the last three years 0:09:23.200 --> 0:09:25.859 are finally starting to come through in the forward-looking indicators. 0:09:27.059 --> 0:09:28.800 And so this is a tough market. 0:09:29.020 --> 0:09:33.970 Obviously, OpenAI just came out with the GPT images 2.5. 0:09:33.970 --> 0:09:36.130 Every day, it feels like there's a new competitor coming 0:09:36.170 --> 0:09:39.970 into the market more aggressively. So on the headlines, Adobe 0:09:40.010 --> 0:09:43.550 is still doing OK, 12% revenue growth, 42% operating margins, 0:09:43.630 --> 0:09:46.170 trading on 12 times earnings, looks cheap. But we know 0:09:46.190 --> 0:09:48.350 that revenue growth is going to slow here as they 0:09:48.410 --> 0:09:49.970 really push into freemium. 0:09:50.429 --> 0:09:52.820 So they're really prioritizing getting free users. 0:09:52.840 --> 0:09:55.440 They now have over a billion users in total, but 0:09:55.460 --> 0:09:59.929 that's coming at the expense of near-term ARR or near-term 0:09:59.990 --> 0:10:00.690 revenue growth. 0:10:01.510 --> 0:10:02.630 You have a new CEO in place. 0:10:02.840 --> 0:10:04.390 Let's see if he'll change the strategy. 0:10:04.580 --> 0:10:07.020 Obviously, he's not new to the company. 0:10:07.059 --> 0:10:08.760 He's been with the company for a very long time. 0:10:08.800 --> 0:10:11.319 And so the question is, what is going to change 0:10:11.340 --> 0:10:13.200 the trajectory of revenue growth for Adobe? 0:10:14.850 --> 0:10:17.490 Stefan Slawinski of BNP Paribas. I've really enjoyed our time 0:10:17.510 --> 0:10:19.500 with you this morning covering both those earnings. Thank you 0:10:19.510 --> 0:10:22.500 very much indeed. Now coming up, the top Pentagon official 0:10:22.540 --> 0:10:24.280 is pushing back on fears. 0:10:24.300 --> 0:10:25.559 Of AI doomerism. 0:10:25.700 --> 0:10:28.760 We've got more on what is the only debate in 0:10:28.800 --> 0:10:31.059 Silicon Valley and around the technology world right now. 0:10:31.179 --> 0:10:47.530 Thank you. This is Bloomberg Tech. Bridgewater Associates co-CIO. 0:10:46.770 --> 0:10:49.870 Greg Jensen is the latest to add his voice to 0:10:49.890 --> 0:10:53.230 the growing calls for slowing down of the AI industry. 0:10:53.309 --> 0:10:56.410 Speaking on Bloomberg's Odd Lots podcast, Jensen, who's an early 0:10:56.450 --> 0:11:00.380 investor in both open AI and anthropic, issued a pressing 0:11:00.400 --> 0:11:04.180 warning about the risks and consequences of the current AI boom, 0:11:04.260 --> 0:11:07.520 likening it to the early days of the COVID pandemic 0:11:07.559 --> 0:11:09.920 when the world took too long to react. 0:11:11.260 --> 0:11:14.470 Unfortunately, this is like what it was like in February 2020. 0:11:14.470 --> 0:11:19.070 Until the AI starts killing people, unfortunately, history would suggest 0:11:19.290 --> 0:11:20.250 we're not going to do anything. 0:11:21.070 --> 0:11:23.250 But we are going to face that. That's going to happen. 0:11:23.830 --> 0:11:27.130 And it'd be much better if we started dealing with 0:11:27.240 --> 0:11:29.140 it before then. 0:11:30.740 --> 0:11:33.520 And you can do it, right? It's also not hopeless. 0:11:35.240 --> 0:11:38.340 The debate around AI doomerism seems to be everywhere right now, 0:11:38.400 --> 0:11:41.100 but a top Pentagon official is pushing back on the idea. 0:11:41.140 --> 0:11:45.100 Defense Department Chief Technology Officer Emil Michael says AI harms 0:11:45.380 --> 0:11:49.530 can be mitigated by the free market, industry-wide collaboration, and 0:11:49.590 --> 0:11:52.990 government engagement. Bloomberg's tech editor Mike Shepard in Washington, D.C., 0:11:53.390 --> 0:11:55.730 joins us now. There are so many different headlines and 0:11:55.770 --> 0:11:59.010 stories for us to pass, but what have we been 0:11:59.070 --> 0:12:02.330 hearing from the administration on this, from President Trump and 0:12:02.390 --> 0:12:05.660 President Trump's team about safety concerns? 0:12:07.040 --> 0:12:09.360 Well, Ed, let's start with Emil Michael. He was asked 0:12:09.420 --> 0:12:13.980 yesterday specifically about that high profile and now viral tweet 0:12:14.020 --> 0:12:17.439 from the anthropic researcher who resigned earlier this week over 0:12:17.500 --> 0:12:21.880 safety concerns. And Michael, who has led negotiations in the past, 0:12:22.320 --> 0:12:25.819 with Anthropic over extra safeguards that the company wanted on 0:12:25.900 --> 0:12:29.560 military use of its tools, culminating in that big blowup 0:12:29.600 --> 0:12:32.480 that we have covered so extensively here. Michael was having 0:12:32.580 --> 0:12:36.829 none of it on this talk about existential risks from AI. 0:12:37.160 --> 0:12:40.410 And he really tried to play down the Doomer talk. 0:12:40.890 --> 0:12:43.530 And we also heard from the president himself last night. 0:12:44.410 --> 0:12:48.570 stepping off Air Force One. He was asked by our 0:12:48.670 --> 0:12:52.390 colleague Jeff Mason, what do you make of the existential risks? 0:12:52.690 --> 0:12:55.670 And the president said, I don't have any concerns. And 0:12:55.720 --> 0:12:59.260 he instead pointed to his main concern with AI as 0:12:59.320 --> 0:13:02.460 being that the U.S. needs to beat China in the 0:13:02.540 --> 0:13:06.059 race to dominate this technology. And of course, the contrast 0:13:06.240 --> 0:13:10.240 is that many lawmakers on Capitol Hill are seeking to 0:13:10.300 --> 0:13:13.670 take some sort of action, ranging from a ban on superintelligence, 0:13:13.990 --> 0:13:16.770 to even a kill switch for AI. Unclear whether we 0:13:16.809 --> 0:13:19.510 would see actually any tangible progress on those. 0:13:20.720 --> 0:13:23.560 I wrote to Anthropic and basically asked them to weigh 0:13:23.660 --> 0:13:26.760 in directly with us because you have an employee that 0:13:26.800 --> 0:13:30.590 resigned and then a current employee who amplified it, and 0:13:30.610 --> 0:13:31.330 they haven't responded. 0:13:31.350 --> 0:13:32.450 I just wanted to put it out there. 0:13:33.150 --> 0:13:35.610 Let's talk about what Anthropic did say, which was they 0:13:35.670 --> 0:13:39.530 published a report yesterday that laid out a pretty wide 0:13:39.590 --> 0:13:44.309 range of findings, including another claim by the company that 0:13:44.390 --> 0:13:47.720 Chinese rivals have been misusing its models publicly. 0:13:47.860 --> 0:13:50.860 There are other security concerns. Shep, go through them. 0:13:52.140 --> 0:13:54.120 And this is yet another layer. And of course, it 0:13:54.230 --> 0:13:58.070 speaks to the president's desire to lead China in AI. 0:13:58.110 --> 0:14:00.530 And what Anthropic is saying in this report, which covered 0:14:00.550 --> 0:14:03.950 a wide range of topics. But the most interesting finding 0:14:03.990 --> 0:14:06.690 to our view was that it was claiming that Moonshot, 0:14:07.910 --> 0:14:10.910 Moonshot AI from China, and we all know it from 0:14:10.970 --> 0:14:14.890 its Kimmy K3 breakthrough model from July, that Moonshot had 0:14:15.010 --> 0:14:20.940 routed as many as 300,000 user requests through Claude, through 0:14:21.000 --> 0:14:24.120 Anthropic's Claude model, and then tried to present the findings 0:14:24.420 --> 0:14:27.600 as its own in a way to gain an edge 0:14:27.860 --> 0:14:31.940 in the AI race. it used as many as 5,000 0:14:31.940 --> 0:14:35.480 fraudulent accounts to do so, Anthropic said. And that's important 0:14:35.500 --> 0:14:39.790 because Anthropic has really barred users in China from turning 0:14:39.830 --> 0:14:44.630 to its technology. It sees China as a risk geopolitically. 0:14:44.670 --> 0:14:48.230 It's something that Dario Amadei, the chief executive of Anthropic, 0:14:48.250 --> 0:14:52.110 has articulated. And he is concerned that his company's products 0:14:52.250 --> 0:14:56.530 not be used in China. And they have This only 0:14:56.720 --> 0:15:00.380 adds to the claims from Anthropic and other AI companies 0:15:00.440 --> 0:15:04.040 in the U.S. that China has resorted to unfair practices, 0:15:04.100 --> 0:15:10.300 including distillation, where Chinese rivals have systematically extracted data from 0:15:10.340 --> 0:15:13.560 their American rivals to build this new vanguard of chatbots. 0:15:13.920 --> 0:15:18.760 China has rejected those claims. It's warned against U.S. retaliation 0:15:18.800 --> 0:15:22.840 or crackdown on Chinese companies. and threatened reprisals of its own. 0:15:22.880 --> 0:15:25.550 And of course, Ed, this all adds to tensions over 0:15:25.670 --> 0:15:29.770 AI as President Trump prepares to meet his Chinese counterpart, 0:15:30.410 --> 0:15:31.880 Xi Jinping, later this month. 0:15:32.700 --> 0:15:34.440 Bloomberg's Mike Shepard, thank you very much. 0:15:34.520 --> 0:15:37.060 As AI researchers continue to sound the alarm over the 0:15:37.100 --> 0:15:40.580 technology's risks, Anthropic's out with a new report detailing how 0:15:40.640 --> 0:15:43.200 its AI is being misused. And according to the company's 0:15:43.240 --> 0:15:48.090 threat intelligence team, users from Yemen, Iran, Russia, and China 0:15:48.490 --> 0:15:52.550 were flagged after having tapped Claude to develop missile systems, 0:15:52.610 --> 0:15:56.650 kamikaze drone swarms, and and even biological weapons research. And 0:15:56.690 --> 0:16:02.970 Fropik says it disrupted every operation in that report. 0:16:04.310 --> 0:16:05.210 Let's go to OpenAI. 0:16:05.450 --> 0:16:09.730 OpenAI is now considering slowing down the development of cutting-edge AI. 0:16:09.830 --> 0:16:13.270 And CEO Sam Altman is urging other AI companies to 0:16:13.310 --> 0:16:16.130 do the same. Bloomberg's Shreen Ghaffari broke the story with 0:16:16.200 --> 0:16:18.660 Rachel Metz and joins us now. And this is about 0:16:18.700 --> 0:16:22.060 an internal communication from Sam Altman to the OpenAI team. 0:16:22.890 --> 0:16:25.070 And the bit I want you to explain is he 0:16:25.170 --> 0:16:29.600 also kind of acknowledges that OpenAI can't do this in isolation. 0:16:31.770 --> 0:16:32.080 Right. 0:16:32.120 --> 0:16:35.380 So this was discussed in an all-hand staff meeting this week. 0:16:36.120 --> 0:16:38.920 And the sentiment of Altman's remarks is basically that the 0:16:39.020 --> 0:16:45.540 company is open to pacing the development of these cutting-edge 0:16:45.560 --> 0:16:48.020 AI systems, but it can't go it alone, that it 0:16:48.140 --> 0:16:50.020 is trying to get other companies on board. 0:16:50.300 --> 0:16:51.300 And, you know, we will see. 0:16:52.800 --> 0:16:54.400 How many could potentially join? 0:16:56.060 --> 0:16:58.600 Potentially joins the big the big thing. We don't have 0:16:58.640 --> 0:17:03.760 a sense from Anthropic specifically, right, that they have looked 0:17:03.800 --> 0:17:06.940 at the idea of slowing down work at the frontier. 0:17:08.820 --> 0:17:11.260 We do know that Anthropic has long been the company 0:17:11.280 --> 0:17:13.840 that has been very vocal about the existential problem. 0:17:14.160 --> 0:17:15.320 Risks of AI, right? 0:17:15.340 --> 0:17:19.660 You have CEO Dario Amadei has talked about, you know, 0:17:19.790 --> 0:17:23.489 it's actually been a sort of doomer term has been 0:17:23.510 --> 0:17:25.629 applied to him and something that he said he's not. 0:17:25.690 --> 0:17:29.770 But the company, I think Anthropic sort of philosophically has 0:17:29.810 --> 0:17:33.280 certainly been open to this idea, whether in practice in 0:17:33.320 --> 0:17:35.780 this moment they are ready to do it. And of course, 0:17:35.840 --> 0:17:38.939 ahead of a potential public listing that could be one 0:17:38.960 --> 0:17:41.560 of the biggest is the real question. 0:17:42.780 --> 0:17:45.860 Shereen very quickly on on the doomerism debate and that's 0:17:45.900 --> 0:17:50.180 probably putting it mildly from an anthropic employee that resigned 0:17:50.240 --> 0:17:54.669 an anthropic employee who amplified that statement what is anthropic 0:17:54.730 --> 0:17:58.650 actually said about uh everything that's being discussed largely online 0:17:58.670 --> 0:17:59.350 at the moment. 0:18:00.150 --> 0:18:02.730 Yeah so the company um did send a statement to 0:18:02.850 --> 0:18:07.440 us in uh in response to the heated discussion among 0:18:07.790 --> 0:18:10.450 staff recently. And they said, we have always been transparent 0:18:10.470 --> 0:18:15.149 that AI will bring enormous benefits and unprecedented risks. You know, 0:18:15.190 --> 0:18:16.690 it's a long statement, so I'm not going to read 0:18:16.750 --> 0:18:18.629 all of it. But in summary, they sort of called 0:18:18.690 --> 0:18:21.700 out their responsible scaling policy and how for a long time, 0:18:21.760 --> 0:18:26.520 Anthropic has, you know, had sort of framework in place 0:18:26.920 --> 0:18:30.360 to not release models if they feel they are overly dangerous. 0:18:31.690 --> 0:18:34.890 Bloomberg's Shireen Ghaffari, who again broke that story on OpenAI 0:18:34.990 --> 0:18:35.670 with Rachel Metz. 0:18:35.730 --> 0:18:37.490 Go and read it on all Bloomberg platforms. 0:18:38.369 --> 0:18:40.850 It's time now for Talking Tech. I'm Yahaira Anna. And 0:18:40.910 --> 0:18:44.879 first up, Moonshot AI is targeting a big jump in 0:18:44.970 --> 0:18:49.260 annualized revenue to $ 2 billion by the end of the 0:18:49.340 --> 0:18:52.919 year due to the breakout success of its Kimi K3 model. 0:18:53.260 --> 0:18:56.520 That's according to sources. It comes as Moonshot is in 0:18:56.540 --> 0:19:01.530 the process of raising funds at a $ 50 billion valuation. Plus, 0:19:01.690 --> 0:19:06.130 ZDOT's AI is looking to tap investors for about $ 5 0:19:05.630 --> 0:19:09.210 billion to fund more research and development. That's according to 0:19:09.280 --> 0:19:13.420 terms seen by Bloomberg. The firm, also known as Zipu, 0:19:13.700 --> 0:19:17.720 has been refreshing its model with stronger coding capabilities, releasing 0:19:17.760 --> 0:19:22.440 a new one just last month. And Micron is rewarding 0:19:22.480 --> 0:19:26.100 its employees in Taiwan after a huge jump in profits. 0:19:26.240 --> 0:19:30.129 Net income surged 15-fold in the third quarter, and unions 0:19:30.530 --> 0:19:33.250 have been pushing for workers to share. 0:19:33.090 --> 0:19:33.970 In those gains. 0:19:34.070 --> 0:19:39.000 Now employees will get nearly $ 32, 000 in cash bonuses, plus 0:19:39.060 --> 0:19:40.119 additional stock awards. 0:19:40.440 --> 0:19:40.600 Ed? 0:19:41.980 --> 0:19:42.200 Hi, Rhonda. 0:19:42.220 --> 0:19:42.820 Thank you very much. 0:19:42.960 --> 0:19:46.190 Apple may be late in the foldables game, but it's 0:19:46.270 --> 0:19:48.410 hoping the wait was worth it. Bloomberg's Chris Welch got 0:19:48.470 --> 0:19:51.280 hands-on with the new iPhone. Duo and joins us now. 0:19:51.320 --> 0:19:53.320 And we were together at Apple Park. And then you 0:19:53.359 --> 0:19:57.100 went down into the theater and you had time with it. 0:19:57.270 --> 0:19:57.609 I did. 0:19:57.720 --> 0:19:58.470 First impressions? 0:19:59.369 --> 0:20:00.949 I'd say it just feels different. You know, I've used 0:20:00.970 --> 0:20:03.290 all these foldables over the years. I'm sure they're late 0:20:03.310 --> 0:20:05.690 to the game, but that's what I'm learning from Samsung, 0:20:05.750 --> 0:20:08.070 learning from Google, learning from Apple, all these brands that 0:20:08.090 --> 0:20:10.370 have tried. and kind of now take their own approach. 0:20:10.890 --> 0:20:14.270 And so it just felt different. I mean, yeah, the shape, 0:20:14.690 --> 0:20:16.490 the squat horn factor's gonna take a lot of like, 0:20:17.150 --> 0:20:19.730 you know, it'll take time for people to adjust. It's 0:20:19.750 --> 0:20:22.889 shorter than your iPhone Max, it's a bit wider. But 0:20:22.910 --> 0:20:25.300 the great thing is like Apple has kind of like 0:20:25.320 --> 0:20:31.619 just made so many changes to iOS to like the 0:20:31.640 --> 0:20:35.760 controls on the right side. They kind of like moved 0:20:35.780 --> 0:20:38.399 everything so your thumb can access them easily. And so 0:20:38.460 --> 0:20:41.780