WEBVTT - Nvidia Makes $3.5 Billion Bet on MediaTek 0:00:02.520 --> 0:00:04.210 Bloomberg Audio Studios. 0:00:04.690 --> 0:00:05.370 Podcasts. 0:00:05.690 --> 0:00:07.050 Radio. News. 0:00:13.630 --> 0:00:17.180 Bloomberg Tech is live from the heart of Silicon Valley 0:00:17.700 --> 0:00:19.959 with Ed Ludlow in San Francisco. 0:00:23.140 --> 0:00:24.259 This is Bloomberg Tech. 0:00:24.360 --> 0:00:28.460 Coming up, NVIDIA makes a $ 3. 5 billion investment in MediaTek, 0:00:28.740 --> 0:00:31.159 deepening its collaboration with the Taiwanese chipmaker. 0:00:31.460 --> 0:00:33.270 We'll hear from both CEOs. 0:00:33.630 --> 0:00:36.970 Plus, today is Tim Cook's last day as CEO of 0:00:37.010 --> 0:00:41.010 Apple before John Ternus takes the reins tomorrow. And all 0:00:41.050 --> 0:00:43.790 eyes on Anthropic as it prepares to file publicly in 0:00:43.810 --> 0:00:46.640 the coming weeks. This is other companies either own U.S. 0:00:46.680 --> 0:00:49.420 listing plans in the second half of the year. 0:00:50.260 --> 0:00:51.980 Okay, let's get straight to our top story. 0:00:52.040 --> 0:00:56.350 NVIDIA's making another multi-billion dollar bet on the AI ecosystem, 0:00:56.390 --> 0:01:02.230 investing $ 3. 5 billion in Taiwan's MediaTek. The partnership brings MediaTek's 0:01:02.290 --> 0:01:07.780 custom AI chips into NVIDIA's architecture, plugging so-called XPUs into 0:01:07.840 --> 0:01:10.640 NVIDIA high-speed networking and memory technology. 0:01:10.680 --> 0:01:11.840 The deal could also. 0:01:11.660 --> 0:01:16.059 Revive concerns about circular financing as NVIDIA invests across its 0:01:16.140 --> 0:01:19.200 own ecosystem. I sat down with NVIDIA CEO Jensen Wang 0:01:19.550 --> 0:01:22.069 and MediaTek CEO Rick Tsai earlier today. 0:01:22.890 --> 0:01:26.709 This is not circular because obviously they do their own 0:01:26.730 --> 0:01:31.550 business and we do our own business. And MediaTek's already 0:01:31.670 --> 0:01:38.270 incredibly profitable, incredibly successful. They are the world's best SOC maker. 0:01:38.590 --> 0:01:41.050 They make more SOCs than any company in the world. 0:01:42.130 --> 0:01:44.090 It's a great pleasure for us to invest in a 0:01:44.130 --> 0:01:45.550 technology company in Taiwan. 0:01:45.870 --> 0:01:48.210 It's a great pleasure for us to invest in MediaTek. 0:01:48.750 --> 0:01:51.410 This is going to turn out, of course, the investment 0:01:51.590 --> 0:01:54.090 is grounded on a deep partnership that's going to be 0:01:54.220 --> 0:01:58.100 multiple years. This is a part, the roadmap that we're 0:01:58.160 --> 0:02:02.260 executing on right now is a, it's not one or 0:02:02.340 --> 0:02:05.860 two years, it's a decade long. And so this partnership 0:02:06.020 --> 0:02:08.680 is extremely important to us, and we want to make 0:02:08.720 --> 0:02:12.820 sure that we demonstrate every opportunity, take every opportunity to 0:02:12.880 --> 0:02:16.109 demonstrate the importance of it, the seriousness by which we 0:02:16.150 --> 0:02:18.709 take it, the confidence we have in the partnership, and 0:02:18.730 --> 0:02:23.170 the investment that we're making here demonstrates that. But ultimately, 0:02:23.310 --> 0:02:26.110 this is going to be a phenomenal investment for us, 0:02:26.450 --> 0:02:28.170 and it's going to generate incredible returns. 0:02:28.230 --> 0:02:29.090 I have every confidence. 0:02:29.560 --> 0:02:32.260 Rick, in July, MediaTek said that it would use its 0:02:33.300 --> 0:02:37.680 bond program, $ 5 billion of dollar bonds, basically to support 0:02:37.720 --> 0:02:41.359 and grow supply chain. Earlier in our conversation, you talked 0:02:41.380 --> 0:02:44.940 about the benefit of working with NVIDIA on MVLink Fusion 0:02:44.980 --> 0:02:47.720 is speed to market. What do you plan to do 0:02:48.020 --> 0:02:51.300 with the proceeds? And when you talk about supply chain, 0:02:51.639 --> 0:02:53.880 do you need to do more on Taiwanese suppliers or 0:02:54.040 --> 0:02:56.780 US suppliers in the context of this relationship? 0:02:58.500 --> 0:02:59.490 I sent the, uh. 0:03:00.610 --> 0:03:06.310 Suddenly, well, back in the end of July, we were 0:03:06.350 --> 0:03:14.050 not able to discuss the strategic investment part of the bond. So, 0:03:14.750 --> 0:03:18.750 but the proceeds, I think, will be for multiple purposes, 0:03:18.970 --> 0:03:22.590 supply chain definitely being one of them, and the supply 0:03:22.610 --> 0:03:27.330 chain is a global, I mean, wherever needed to enhance 0:03:27.650 --> 0:03:31.389 and enable the customers to But on the other hand, 0:03:31.430 --> 0:03:38.290 I think, as Jensen said so eloquently, the investment that 0:03:38.410 --> 0:03:42.510 Miratech is a tech company, it's a very strong tech company. 0:03:42.780 --> 0:03:49.330 And we will definitely invest if needed. if anything, more into. 0:03:49.110 --> 0:03:53.160 The technology in our capability to design, in our capability 0:03:53.220 --> 0:03:57.910 to make the ASIC chip in the IP for interconnects 0:03:58.750 --> 0:03:59.890 in advanced packaging. 0:04:00.490 --> 0:04:01.010 I think the. 0:04:02.610 --> 0:04:07.750 The key here is really the collaboration agreements that we 0:04:07.790 --> 0:04:12.470 have built and expanded with NVIDIA as described in our 0:04:12.510 --> 0:04:16.720 press release. Those are the key things and the proceeds 0:04:16.800 --> 0:04:20.659 will basically go in to make those things happen better 0:04:21.180 --> 0:04:21.740 and further. 0:04:22.720 --> 0:04:25.219 Jenson, on the earnings call last week, you talked about, 0:04:25.460 --> 0:04:30.880 with confidence, right, that the GPU-based architecture has superior economics. 0:04:31.080 --> 0:04:33.320 You were talking about that in the context of the 0:04:33.360 --> 0:04:37.360 Frontier Labs, largely because you were asked about custom silicon. 0:04:37.800 --> 0:04:40.299 You know, the simple terms of it is that MVLink 0:04:40.339 --> 0:04:43.690 Fusion just makes it easy to take XPU and just 0:04:43.710 --> 0:04:47.530 drop it into the data center. That would show that, 0:04:47.570 --> 0:04:51.390 you know, Nvidia does see XPU coming into the market. 0:04:52.190 --> 0:04:55.010 Just the difference of your thinking this week to what 0:04:55.029 --> 0:04:55.970 you outlined last week. 0:04:57.440 --> 0:05:01.000 I see XPUs in the market. It's not coming into 0:05:01.020 --> 0:05:04.920 the market. It's in the market. Look, an XPU is 0:05:04.960 --> 0:05:12.440 a specialized chip. NVIDIA's GPU is a general-purpose accelerator. We 0:05:12.480 --> 0:05:19.690 accelerate the entire lifecycle of AI, from data processing to 0:05:19.730 --> 0:05:26.680 pre-training to post-training to agentic AI inferencing. We accelerate every 0:05:26.740 --> 0:05:29.840 single AI model in the world from closed models, every 0:05:29.980 --> 0:05:33.860 closed model, every open model, video models and language models, 0:05:34.020 --> 0:05:38.580 biological models and physics models and robotics models. We literally 0:05:39.140 --> 0:05:45.560 accelerate every single area of AI. That explains why it 0:05:45.670 --> 0:05:48.050 is that we are the most fungible and the most 0:05:48.150 --> 0:05:52.549 durable and therefore the most rentable GPU, the most rentable 0:05:52.589 --> 0:05:54.050 computer infrastructure in the world. 0:05:54.430 --> 0:05:57.010 We're in every cloud, we're on-prem, we're at the edge. 0:05:57.690 --> 0:06:01.830 And so we're able to address markets that, frankly, nobody 0:06:01.930 --> 0:06:05.330 else can because of the architecture of our company and 0:06:05.770 --> 0:06:09.890 the entire full-stack AI factory platform that we have. 0:06:12.870 --> 0:06:15.000 Let's get more on the story with Bloomberg's Ian King, 0:06:15.010 --> 0:06:18.280 who helped me break it this morning. It's really interesting. 0:06:18.320 --> 0:06:20.940 For some time now, gentlemen have been asked about, what 0:06:20.980 --> 0:06:24.240 about all these custom chips that hyperscalers are working on, 0:06:24.279 --> 0:06:27.540 that the frontier labs are working on? And to this point, 0:06:27.560 --> 0:06:30.770 he's been a little bit dismissive of the threat. Today, 0:06:30.830 --> 0:06:35.510 with MediaTek, they found a mechanism for those same AI 0:06:35.570 --> 0:06:38.890 companies to drop an XPU or a custom chip into 0:06:38.910 --> 0:06:41.690 an existing data center to work with NVIDIA's technology. 0:06:42.180 --> 0:06:43.420 What else do we need to know here? 0:06:43.520 --> 0:06:45.640 Yeah, I mean, taking a step back and looking at 0:06:45.680 --> 0:06:49.839 this from a technology perspective, the way the data center 0:06:49.880 --> 0:06:52.720 works and how effective it is isn't just how good 0:06:52.740 --> 0:06:54.820 are the individual components, it's how good do they all 0:06:54.860 --> 0:06:57.220 link together. And what NVIDIA is saying is, hey, use 0:06:57.279 --> 0:07:00.960 our link, use this fabric, they call it, and to 0:07:01.000 --> 0:07:04.400 bind everything together and it'll all work better. So what 0:07:04.420 --> 0:07:06.280 he's trying to do is Even if you want to 0:07:06.320 --> 0:07:07.960 go it alone, even if you want to use your 0:07:08.040 --> 0:07:09.239 own stuff, use. 0:07:09.140 --> 0:07:11.170 It inside our framework. 0:07:11.230 --> 0:07:13.010 Use it inside our house and that'll help you. 0:07:13.030 --> 0:07:14.030 And of course that means. 0:07:15.040 --> 0:07:17.600 in terms of business, that look, hey, if companies want 0:07:17.640 --> 0:07:19.920 to do their own thing like Amazon did with that 0:07:19.940 --> 0:07:22.550 announcement we saw during earnings, they're going to do it 0:07:22.770 --> 0:07:25.010 and they're still going to be using at least some 0:07:25.050 --> 0:07:27.610 of his chips, so it works business-wise. 0:07:27.830 --> 0:07:29.390 We haven't said it to this point, so I'll say 0:07:29.410 --> 0:07:32.680 it now for the audience. This is NVIDIA buying convertible 0:07:32.750 --> 0:07:34.720 bonds that convert into equity. 0:07:34.720 --> 0:07:36.080 $ 3. 0:07:34.720 --> 0:07:39.980 5 billion, we confirmed, is NVIDIA's biggest non-US direct investment. 0:07:40.300 --> 0:07:42.760 So kind of interesting, but Probably talk a little bit 0:07:42.780 --> 0:07:44.140 more about MediaTek, right? 0:07:44.180 --> 0:07:47.120 What is MediaTek? What is it comparable to here in 0:07:47.140 --> 0:07:47.680 the U.S. 0:07:47.920 --> 0:07:49.540 Yeah, no, that's a very good question. This is a 0:07:49.610 --> 0:07:53.310 very important company that nobody really pays that much attention to. 0:07:53.690 --> 0:07:58.850 Think they're the Taiwanese equivalent of Qualcomm. They make smartphone processors, 0:07:58.910 --> 0:08:01.430 smartphone modems. You know, the real guts of that, they 0:08:01.450 --> 0:08:04.570 have a large presence in technology. the China market and 0:08:04.630 --> 0:08:08.010 large presence worldwide. And by numbers, they're one of the 0:08:08.050 --> 0:08:10.990 biggest companies. And that's really what they've done, and they're 0:08:11.010 --> 0:08:11.970 looking to expand. 0:08:12.010 --> 0:08:14.770 But in ASICs, they're more like Marvell or Broadcom, aren't they? 0:08:14.950 --> 0:08:16.380 That's right. This is what they're trying to do. 0:08:16.390 --> 0:08:19.560 They're trying to basically take that expertise that they've developed 0:08:19.640 --> 0:08:22.320 in smartphones and expand that into the data center, which 0:08:22.360 --> 0:08:25.020 is a more lucrative business. And what they're doing is 0:08:25.180 --> 0:08:27.860 trying to replicate what Broadcom does, what Marvell does, which 0:08:27.920 --> 0:08:30.870 is help, say, Google, say, Amazon. Hey, you've got this 0:08:30.930 --> 0:08:33.290 great idea for a chip. You've got these specifications. We'll 0:08:33.330 --> 0:08:34.350 get you all the way there. 0:08:34.929 --> 0:08:37.470 Bloomberg's Ian King, who broke the story with me this morning, 0:08:37.490 --> 0:08:38.829 and it's one of the most read today on the 0:08:38.850 --> 0:08:41.790 Bloomberg platforms. Go and check out the details. NVIDIA's CEO, 0:08:41.830 --> 0:08:44.959 Jens Wang, also talks about how the specialist platforms are 0:08:45.020 --> 0:08:48.580 largely inference platforms. But as you heard, NVIDIA is across 0:08:48.740 --> 0:08:49.459 every phase. 0:08:49.780 --> 0:08:50.239 Listen to this. 0:08:51.340 --> 0:08:52.979 An XPU is a specialized chip. 0:08:54.130 --> 0:08:58.970 NVIDIA's GPU is a general purpose accelerator. 0:08:59.330 --> 0:09:01.640 We accelerate the. 0:09:01.620 --> 0:09:06.579 Entire lifecycle of AI from data processing, which data processing 0:09:06.600 --> 0:09:10.620 to pre-training to post-training to agentic AI inferencing. 0:09:11.340 --> 0:09:15.510 We accelerate every single AI model in the world. 0:09:17.860 --> 0:09:18.840 Let's get more on NVIDIA. 0:09:18.880 --> 0:09:22.800 Brad Warden, Senior Portfolio Manager at Nomura Asset Management International 0:09:22.840 --> 0:09:26.319 joins us. And kindly, Brad, in advance, you outlined the 0:09:26.360 --> 0:09:30.140 different phases of AI. You've heard from Jensen there about 0:09:30.220 --> 0:09:33.939 why they did this with MediaTek to allow XPUs to 0:09:33.960 --> 0:09:37.490 come into the ecosystem. What does that reflect and which 0:09:37.570 --> 0:09:39.329 phase of AI you think we're in right now? 0:09:40.510 --> 0:09:42.849 Right. Thanks, Ed. And thanks for having me this morning. 0:09:42.910 --> 0:09:45.480 I think You know, if you think about the phases 0:09:46.100 --> 0:09:48.979 of AI, right, we've built this foundation, right, which was, 0:09:49.000 --> 0:09:52.870 you know, I think we started thinking about GPUs, right, 0:09:52.990 --> 0:09:55.130 as the base of that and really saw that part 0:09:55.150 --> 0:09:57.830 of the ecosystem develop. And then a lot of things 0:09:57.870 --> 0:09:59.910 develop around that. And I think that's what you're talking about. 0:10:00.220 --> 0:10:03.670 seeing right now. So you've had this foundation that's gotten 0:10:03.790 --> 0:10:06.650 built for AI. Now we've kind of moved into this 0:10:06.690 --> 0:10:11.530 phase of monetization. And a lot of the questions, I think, 0:10:11.590 --> 0:10:14.750 which go back to what you were outlining earlier, just 0:10:14.790 --> 0:10:18.620 around financing and thinking about what sort of are the 0:10:18.760 --> 0:10:21.460 next steps of that. And I think if you really 0:10:21.620 --> 0:10:24.780 step back and think about the relationship that was announced 0:10:24.820 --> 0:10:30.080 that you talked about, really is this idea of basically 0:10:30.200 --> 0:10:35.900 expanding the ecosystem and the strategic aspect of what NVIDIA 0:10:35.980 --> 0:10:39.750 is doing. is making sure that they play across that ecosystem. 0:10:39.770 --> 0:10:42.500 And you've seen it in a number of different investments 0:10:42.540 --> 0:10:46.099 that they've made, but I think this one obviously extends 0:10:46.140 --> 0:10:46.570 some of that. 0:10:48.020 --> 0:10:50.780 Brad, I really value your perspective and reaction here because 0:10:50.820 --> 0:10:56.040 in the Nomura Transformational Technologies ETF, the top holdings are NVIDIA, AMD, TSMC, 0:10:56.140 --> 0:10:59.540 Seagate Land Research, but also Broadcom's in there as well, right, 0:10:59.740 --> 0:11:02.990 on the custom ASIC side. What's the biggest motivation right 0:11:03.020 --> 0:11:06.290 now that you're seeing from those pursuing custom silicon? The 0:11:06.470 --> 0:11:10.309 economics on a dollar per token basis or diversity of 0:11:10.350 --> 0:11:12.650 supply is a kind of like safety net. 0:11:13.760 --> 0:11:16.240 That's a fantastic point. 0:11:16.280 --> 0:11:19.690 I think if you look at each of the adoption 0:11:19.710 --> 0:11:22.420 that each of these big hyperscalers are doing in terms 0:11:22.460 --> 0:11:26.089 of looking at XPUs, looking at ways to really optimize 0:11:26.120 --> 0:11:31.350 that infrastructure, which fundamentally, if you look at technology, it's 0:11:31.390 --> 0:11:35.030 the cost per bit of output. So in this case, 0:11:35.110 --> 0:11:37.890 it's token per watt, if you want to think about that. 0:11:38.330 --> 0:11:43.030 And if you know your infrastructure and your operating requirements 0:11:43.100 --> 0:11:46.400 the best out of anybody, which these hyperscalers do, their 0:11:46.420 --> 0:11:48.420 ability to optimize that and do it in their own 0:11:48.559 --> 0:11:51.700 silicon environment I think is something that's very, very powerful. Now, 0:11:51.720 --> 0:11:54.219 there are a lot of reasons why you use a 0:11:54.280 --> 0:11:57.520 general compute, a GPU as well. 0:11:57.660 --> 0:11:59.500 And so both of those are valuable. 0:11:59.580 --> 0:12:04.079 So how you leverage that, again, goes to optimizing that infrastructure. 0:12:04.140 --> 0:12:06.500 And if we think about it too, recently we've been 0:12:06.540 --> 0:12:10.170 talking about open source versus closed source models and how 0:12:10.230 --> 0:12:13.690 those both develop. And I think the idea is that 0:12:13.710 --> 0:12:16.530 you optimize the usage of that. And that's what technology 0:12:16.570 --> 0:12:19.210 does is over time, we figure out ways to optimize. 0:12:19.250 --> 0:12:21.050 And so there's a place for all of these. 0:12:21.090 --> 0:12:24.010 And it goes back to the strategy of being able 0:12:24.410 --> 0:12:26.650 to touch all of those points. And I think that 0:12:26.690 --> 0:12:29.469 was the point Jensen was making in in your interview, 0:12:29.809 --> 0:12:32.670 was really like, look, our customers are asking for all 0:12:32.710 --> 0:12:36.380 these different architectures and capabilities. And if we can be 0:12:36.429 --> 0:12:38.530 a partner in that and work on that, that's an 0:12:38.630 --> 0:12:42.040 important strategic move on the part of NVIDIA to stay 0:12:42.440 --> 0:12:44.420 at the heart of that and to make sure that 0:12:44.540 --> 0:12:47.819 their advantage stays there. Plus, they get a view into 0:12:47.960 --> 0:12:51.880 how the whole ecosystem is developing at every touchpoint that 0:12:51.900 --> 0:12:54.980 they've been involved in. And that's critical from a competitive 0:12:55.020 --> 0:12:57.980 standpoint and very smart from a strategic perspective as well. 0:12:58.840 --> 0:13:00.679 So that ending there, interesting. 0:13:00.760 --> 0:13:03.400 Where do you sit on the concern or lack of 0:13:03.460 --> 0:13:06.489 concern in what is or is not circular financing in 0:13:06.890 --> 0:13:07.370 this case? 0:13:08.429 --> 0:13:10.230 I think it's something that's worth watching, right? 0:13:10.250 --> 0:13:13.150 I mean, we're watching very closely the amount of leverage 0:13:13.190 --> 0:13:16.140 that's being put on the cash flows, but we're watching 0:13:16.179 --> 0:13:17.219 the monetization, right? 0:13:17.240 --> 0:13:17.959 We're watching very. 0:13:17.860 --> 0:13:22.000 Carefully the revenues that are accruing from those types of financing. 0:13:22.080 --> 0:13:24.640 And I think the market too is clearly skeptical, right? 0:13:24.800 --> 0:13:28.030 Because we've seen really powerful earnings power out of a 0:13:28.050 --> 0:13:31.230 number of these companies, including the ones we've just talked about. 0:13:32.450 --> 0:13:36.250 But really, the valuations continue to fall, right? And part 0:13:36.270 --> 0:13:39.960 of that is that skepticism around that and the circularity. 0:13:40.000 --> 0:13:43.280 You watch the CDS. May I just jump in and ask? 0:13:43.300 --> 0:13:44.480 I'm sorry, because we'll run out of time. 0:13:44.500 --> 0:13:48.260 In this MediaTek case study, the argument that both CEOs 0:13:48.320 --> 0:13:52.440 made was, In a custom ASICs context, you can get 0:13:52.480 --> 0:13:55.709 it into the real world quicker. You've got your custom accelerator, 0:13:55.770 --> 0:13:59.090 you just drop it into an existing architecture. But the 0:13:59.150 --> 0:14:01.689 counterargument to that was, well, NVIDIA's been able to find 0:14:01.750 --> 0:14:05.240 its own demand, therefore. I'm trying to work out the 0:14:05.580 --> 0:14:07.440 real world mechanics of how that would go. 0:14:09.340 --> 0:14:12.220 Well, I'm not sure specifically. 0:14:12.240 --> 0:14:15.350 I think what this does is it does definitely speed 0:14:15.590 --> 0:14:19.370 up the capability to pull that in from a customer standpoint. 0:14:19.390 --> 0:14:21.950 And I think that is the idea is the speed 0:14:22.010 --> 0:14:24.590 to market and the speed with which you can connect 0:14:24.630 --> 0:14:28.350 all these systems together is becoming more critical because we're 0:14:28.390 --> 0:14:30.650 moving at light speed, right, in terms of how this 0:14:30.710 --> 0:14:32.910 is happening. So as we do this, it's actually a 0:14:32.930 --> 0:14:33.650 very smart thing. 0:14:35.160 --> 0:14:35.860 I'm enjoying this. 0:14:35.940 --> 0:14:38.160 Some people wake up and talk about monetary policy and 0:14:38.200 --> 0:14:41.600 inflation expectations, and we wake up and talk about compute 0:14:41.620 --> 0:14:44.400 and architecture. Thank you very much indeed. Numero Asset Management 0:14:44.780 --> 0:14:49.010 International Senior Portfolio Manager, Brad Warden. Coming up, China laid 0:14:49.090 --> 0:14:52.390 out the terms for AI talks with the US and 0:14:52.850 --> 0:14:55.490 accused Anthropic of bad behavior. 0:14:55.950 --> 0:14:58.410 We've got more on that story next. This is Bloomberg Tech. 0:15:07.980 --> 0:15:10.370 It's time now for Talking Tech. I'm Yahaira Anna. And 0:15:10.470 --> 0:15:13.410 first up, shares of Shein on the gray market are 0:15:13.530 --> 0:15:16.390 under pressure ahead of the fast fashion giant's Hong Kong 0:15:16.410 --> 0:15:19.570 debut Tuesday. The stock fell as much as 17% in 0:15:20.270 --> 0:15:21.890 gray market trading after. 0:15:21.690 --> 0:15:21.910 Its $ 1. 0:15:21.810 --> 0:15:22.370 7 billion IPO. 0:15:22.390 --> 0:15:24.510 That values Shein at about $ 26 billion, well below its 0:15:24.530 --> 0:15:34.840 nearly $ 100 billion valuation back in 2022. Plus, SK Hynix 0:15:34.920 --> 0:15:39.020 is eyeing Japan as it races to meet booming AI demand. 0:15:39.060 --> 0:15:43.310 The memory chip giant is studying potential manufacturing sites across 0:15:43.370 --> 0:15:47.070 the country with its chairman, telling Bloomberg it wants locations 0:15:47.110 --> 0:15:50.510 with good power and good water. According to a source, 0:15:50.550 --> 0:15:54.410 a small acquisition or partnership with a Japanese company is 0:15:54.580 --> 0:15:58.860 also on the table. And ZDOT's AI fell short of 0:15:58.960 --> 0:16:03.020 expectations as China's AI battle heats up. First up, first 0:16:03.080 --> 0:16:08.300 half revenue grew nearly fivefold to $ 142 million, but still 0:16:08.370 --> 0:16:13.190 missed analyst estimates by about 30%. That's despite the company's 0:16:13.610 --> 0:16:18.590 market value surging more than 800% since its January listing. 0:16:18.810 --> 0:16:18.910 Ed? 0:16:19.860 --> 0:16:20.960 Okay, thank you, Yuhaira. 0:16:21.020 --> 0:16:24.640 China called for greater guidelines for AI models and accused 0:16:24.680 --> 0:16:28.360 Anthropic of bad behavior ahead of AI talks with the U.S. 0:16:28.400 --> 0:16:31.000 China said on Sunday the U.S. to prove its AI 0:16:31.080 --> 0:16:34.710 companies are subject to the same guidelines before any, quote, 0:16:34.730 --> 0:16:37.490 substantive talks with China could take place. 0:16:38.250 --> 0:16:40.370 Mike Shepard joins us with the latest. 0:16:40.430 --> 0:16:43.410 This is China trying to set the agenda, right, Shep, 0:16:43.890 --> 0:16:45.710 ahead of what are pretty key conversations. 0:16:46.970 --> 0:16:47.590 Very much so. 0:16:47.690 --> 0:16:49.730 And this is a conversation, of course, that would set 0:16:49.790 --> 0:16:53.440 the stage for a dialogue on artificial intelligence and AI 0:16:53.480 --> 0:16:57.560 policy between President Donald Trump and Chinese leader Xi Jinping 0:16:58.180 --> 0:17:01.840 September 24th here in Washington. And Ed, before the two 0:17:01.880 --> 0:17:04.900 leaders sit down, though. They need to sort out exactly 0:17:05.040 --> 0:17:07.399 what will be on the table before them. And there 0:17:07.420 --> 0:17:10.680 needs to be this dialogue between lower-level. 0:17:10.060 --> 0:17:12.420 Officials from the two countries. 0:17:12.700 --> 0:17:15.380 So far, we have seen no date or agenda set 0:17:15.500 --> 0:17:18.510 for such a conversation to take place. And we see 0:17:18.550 --> 0:17:21.570 here in this post from a social media account that's 0:17:21.640 --> 0:17:26.609 affiliated with the Chinese state broadcaster CCTV a version of 0:17:26.730 --> 0:17:30.670 what it would like. Before any conversation can happen on AI, 0:17:30.710 --> 0:17:33.690 the U.S. really does, according to China, need to meet 0:17:33.750 --> 0:17:38.470 these terms that it is setting. American AI developers need 0:17:38.530 --> 0:17:41.430 to be following a common set of rules when it 0:17:41.470 --> 0:17:44.090 comes to disclosure, when it comes to audits, and when 0:17:44.109 --> 0:17:47.280 it comes to safety. And interestingly, Ed, it also singled 0:17:47.359 --> 0:17:50.080 out Anthropic as perhaps its main villain in the whole 0:17:50.160 --> 0:17:50.560 opera here. 0:17:51.280 --> 0:17:53.939 So that is interesting. I mean, they are singling out 0:17:54.060 --> 0:17:57.520 a single U.S. company or entity. Do we know why? 0:17:59.220 --> 0:18:02.800 Well, you know, the post itself was titled, Anthropic Has 0:18:02.840 --> 0:18:06.600 Contracted the American Disease, and then used anthropic really as 0:18:06.640 --> 0:18:11.660 a way of explaining all of its issues. Anthropic, though, 0:18:11.720 --> 0:18:14.460 why them and not OpenAI, which, of course, has had 0:18:14.500 --> 0:18:16.980 its own safety issues that we have talked about extensively 0:18:17.040 --> 0:18:21.000 on this show with the hack on Hugging Face. Anthropic, though, 0:18:21.060 --> 0:18:23.320 has really taken a position that is very much in 0:18:23.340 --> 0:18:27.010 the view of officials from Beijing as anti-China. Dario Amadei 0:18:27.050 --> 0:18:31.150 has supported the export controls that past administrations have pushed 0:18:31.490 --> 0:18:35.670 to prevent Beijing and Chinese companies from acquiring advanced AI chips. 0:18:35.970 --> 0:18:38.379 And he has also warned against the prospect of China 0:18:38.460 --> 0:18:43.560 actually taking the lead in artificial intelligence development and software. 0:18:43.600 --> 0:18:45.840 And the reason is that they see it as possibly 0:18:45.880 --> 0:18:48.770 being used as a tool by Beijing. And of course, 0:18:48.790 --> 0:18:52.210 we're seeing China now react against against Anthropic. 0:18:53.040 --> 0:18:55.000 Bloomberg's Mike Sheppard with the reporting. Thank you very much. 0:18:56.020 --> 0:18:59.690 Tomorrow, Apple gets a new CEO. But how will that 0:18:59.730 --> 0:19:03.649 affect the iPhone makers' priorities? Bloomberg's Mark Gurman, who leads 0:19:03.710 --> 0:19:06.290 our coverage of Apple and consumer tech, has been looking 0:19:06.330 --> 0:19:09.450 into that and what John Turner brings to the table. 0:19:09.490 --> 0:19:10.300 Mark is with us now. 0:19:10.340 --> 0:19:12.919 This was the subject of this weekend's Power On. but 0:19:12.960 --> 0:19:17.440 a really detailed piece in Bloomberg Businessweek. And let's start there. Okay, 0:19:17.460 --> 0:19:20.140 so today is Tim Cook's last day. Tomorrow is day 0:19:20.220 --> 0:19:23.310 one of John Ternus' CEO. What changes overnight? 0:19:25.040 --> 0:19:28.640 I would say pretty much nothing changes overnight of any 0:19:29.160 --> 0:19:33.399 major significance. This is going to be a gradual multi-year 0:19:33.840 --> 0:19:36.760 changing of the guard. Tim Cook is not going anywhere. 0:19:36.820 --> 0:19:39.550 He's still going to be around. There are questions about 0:19:39.609 --> 0:19:42.070 how much he's going to be around. I'm led to 0:19:42.109 --> 0:19:45.250 believe he will be very present and intimately involved with 0:19:45.430 --> 0:19:48.770 Apple in the several years to come, serving as an 0:19:48.830 --> 0:19:52.420 advisor to John Ternes, the new CEO, but also having 0:19:52.440 --> 0:19:58.080 a major impact on things like supply chain, operations, procurement, 0:19:58.700 --> 0:20:03.139 but most notably working with the Trump administration and the 0:20:03.180 --> 0:20:06.040 government in China and other governments around the world where 0:20:06.080 --> 0:20:08.720 he owns those hands-on relationships. 0:20:09.840 --> 0:20:11.080 Timing is so interesting. 0:20:11.240 --> 0:20:14.270 You know, next week, next Wednesday is the Apple full 0:20:14.410 --> 0:20:18.629 iPhone event. How much has John Ternus had a hand 0:20:19.390 --> 0:20:22.240 in this generation of iPhone that's likely to be announced 0:20:23.000 --> 0:20:23.980 when we get them next week? 0:20:25.490 --> 0:20:28.250 Well, my answer is the same for the last five generations. 0:20:28.350 --> 0:20:31.670 John Ternus has been leading the development of the iPhone 0:20:31.750 --> 0:20:34.920 since the release of the iPhone 12 in 2020. 0:20:34.920 --> 0:20:36.060 He spearheaded that. 0:20:36.380 --> 0:20:39.399 That was his final test before being named to the 0:20:39.460 --> 0:20:43.510 CEO role earlier this year or being named to the 0:20:43.530 --> 0:20:45.150 senior VP of hardware engineering role. 0:20:45.570 --> 0:20:47.310 In 2021, that. 0:20:47.310 --> 0:20:49.610 iPhone was a big success at the height of COVID 0:20:49.910 --> 0:20:52.290 with 5G and the new design. And then they rolled 0:20:52.369 --> 0:20:56.520 out subsequent iPhone after iPhone after iPhone under Ternus. And 0:20:56.580 --> 0:20:59.860 the timing of the transition is not a coincidence. Typically, 0:20:59.900 --> 0:21:04.960 they like to do these big transitions either in October, December, 0:21:05.240 --> 0:21:07.980 or at the very beginning of the year. September 1 0:21:07.980 --> 0:21:10.820 was picked for a very specific reason to have John 0:21:10.859 --> 0:21:14.159 Ternus be the one to introduce the the foldable iPhone, 0:21:14.780 --> 0:21:18.020 the smart home display, the new MacBook Pro with a touchscreen, 0:21:18.460 --> 0:21:20.820 basically be able to present him as sort of this 0:21:20.900 --> 0:21:24.160 new product development maestro as Apple rolls out its biggest 0:21:24.200 --> 0:21:25.659 new product roadmap in its history. 0:21:26.720 --> 0:21:28.770 So the thing that will change really overnight in the 0:21:28.810 --> 0:21:32.330 internal directory at Apple is some job titles. Very quickly, 0:21:32.450 --> 0:21:35.010 just recap what we've reported on how John Ternus is 0:21:35.070 --> 0:21:40.560 expected to be different in his leadership, his style, I 0:21:40.600 --> 0:21:41.000 think there's. 0:21:40.880 --> 0:21:42.600 Going to be a lot of continuity there, right? 0:21:42.680 --> 0:21:45.600 I mean, he has taken every opportunity to praise Cook's 0:21:45.660 --> 0:21:49.440 leadership style. The big difference is that over the past 0:21:49.440 --> 0:21:53.730 15 years, decision-making at Apple has been very collaborative, where 0:21:53.810 --> 0:21:57.250 Tim Cook has let the executive team make the decisions, 0:21:57.369 --> 0:21:59.450 and maybe he would make the final decision, but it 0:21:59.490 --> 0:22:04.010 was a consensus-driven leadership process. At Apple, under John Ternes, 0:22:04.390 --> 0:22:07.080 I expect far more decisiveness and singular decision-making. 0:22:08.320 --> 0:22:10.760 Bloomberg's Mark Gurman with two must-reads from the weekend, the 0:22:10.820 --> 0:22:14.359 Power On newsletter and this piece in Businessweek about the transition. 0:22:14.400 --> 0:22:19.020