WEBVTT - Apple’s Ternus Era Starts the iPhone Duo 0:00:02.540 --> 0:00:07.110 Bloomberg Audio Studios, podcasts, radio, news. 0:00:13.050 --> 0:00:16.580 Bloomberg Tech is live from the heart of Silicon Valley 0:00:17.120 --> 0:00:22.780 with Ed Ludlow in San Francisco. This is Bloomberg Tech. 0:00:22.840 --> 0:00:27.000 Coming up, it's what everyone's talking about. Apple's first foldable smartphone, 0:00:27.020 --> 0:00:31.020 the iPhone Duo, will break everything down from yesterday's product launch. Plus, 0:00:31.040 --> 0:00:34.020 the AI chip race is moving to inference. D-Matrix CEO 0:00:34.280 --> 0:00:37.010 Sid Sheff joins us on its new partnership with NVIDIA 0:00:37.229 --> 0:00:40.510 and the battle to make AI cheaper and faster. And 0:00:40.570 --> 0:00:44.830 Elon Musk's Boring Co. reaches a $ 23 billion valuation after 0:00:44.970 --> 0:00:49.640 its latest funding round. Let's get straight to our top story, 0:00:49.780 --> 0:00:50.980 Apple's showstopper. 0:00:51.890 --> 0:00:55.820 iPhone Duo brings an entirely new design while preserving everything 0:00:55.840 --> 0:00:58.760 you love and expect from iPhone. It gives you the 0:00:58.840 --> 0:01:02.620 largest display ever on an iPhone while still fitting easily 0:01:02.660 --> 0:01:07.310 in your pocket. It reflects extraordinarily complex engineering yet feels 0:01:07.510 --> 0:01:11.310 effortless to use. And it makes new experiences possible. 0:01:12.730 --> 0:01:15.429 That was Apple CEO John Turner making his first official 0:01:15.490 --> 0:01:18.150 keynote in the top seat, turning heads with the new 0:01:18.230 --> 0:01:22.270 foldable iPhone Duo. Shares really accelerating gains 90 minutes into 0:01:22.310 --> 0:01:25.690 the session, up 2.5% now. What's interesting is they were 0:01:25.709 --> 0:01:28.130 basically flat, bang in line with the average of an 0:01:28.290 --> 0:01:31.720 iPhone launch day market performance that we've seen going back 0:01:31.750 --> 0:01:34.800 at least a decade. Well, everyone's woken up and said, 0:01:34.840 --> 0:01:36.860 this is a big deal. Let's discuss. And who better 0:01:36.880 --> 0:01:39.220 to discuss it with than Mark Gurman, who leads our 0:01:39.260 --> 0:01:42.260 Apple coverage and basically broke most of the details prior 0:01:42.300 --> 0:01:45.880 to yesterday's event anyway. Start with the basics, iPhone Duo. 0:01:46.940 --> 0:01:49.260 We were spot on on the price. That's probably the 0:01:49.760 --> 0:01:53.520 big news headline, right? Just under $ 2, 000. But the specs 0:01:53.600 --> 0:01:56.920 as well. Your first impressions and the most notable points 0:01:56.970 --> 0:01:57.670 of the technology. 0:01:59.140 --> 0:02:02.190 My first impression is that this device makes every foldable 0:02:02.220 --> 0:02:05.680 phone from Samsung, from Google, from Huawei, anyone else before it, 0:02:06.090 --> 0:02:08.230 look like a toy, look like a prototype, look like 0:02:08.270 --> 0:02:11.049 a beta test, basically, because the fit and finish we're 0:02:11.090 --> 0:02:15.010 seeing here is new to the foldable industry. The way 0:02:15.050 --> 0:02:18.769 that they are putting, they call it ceramic inserts into 0:02:18.810 --> 0:02:21.950 the cellular antenna cutouts, so when you go and try 0:02:21.990 --> 0:02:24.389 to bend the phone and break it, people who do that, 0:02:24.690 --> 0:02:27.290 it's not gonna bend in half. The way they put 0:02:27.430 --> 0:02:31.410 a special cover glass over the flexible display to hide 0:02:31.470 --> 0:02:33.769 the crease as much as possible and make it so 0:02:33.790 --> 0:02:36.790 your fingernails don't put cuts in the screen. The fact 0:02:36.810 --> 0:02:40.330 that they're using titanium around the edges to make the 0:02:40.350 --> 0:02:44.010 phone more durable, the fact that it's water and dust resistant, right? 0:02:44.070 --> 0:02:46.960 They've hit every checkbox here for it to feel like 0:02:47.000 --> 0:02:50.419 a normal iPhone, but to also be foldable. This really 0:02:50.600 --> 0:02:54.520 has very few compromises other than the lack of the 0:02:54.560 --> 0:02:58.140 telephoto camera on the back and battery life to some extent. 0:02:59.190 --> 0:03:01.550 It was amazing to get hands on with it yesterday. And, 0:03:01.690 --> 0:03:05.170 you know, everyone's sort of first reaction is to say 0:03:05.210 --> 0:03:08.440 there's no crease, there's no margin, there's no seam in 0:03:08.470 --> 0:03:10.920 the middle, not one that the naked eye can see. 0:03:10.940 --> 0:03:13.119 I do want to talk about the pricing, Mark. You 0:03:13.180 --> 0:03:20.400 said this would be $ 2, 000. It was in the end $ 1, 999. 0:03:18.840 --> 0:03:21.890 To what extent has Apple eaten some cost in the 0:03:21.950 --> 0:03:24.989 environment they're operating in? But also, you know, if you 0:03:25.010 --> 0:03:28.460 look at the higher tiers of storage, clearly there are 0:03:28.740 --> 0:03:30.130 some handsets out there up beyond $ 3, 000 too. 0:03:32.600 --> 0:03:36.280 So they're eating costs for sure because the memory shortage, 0:03:36.340 --> 0:03:39.300 you know, really is having an impact on so many companies, 0:03:39.340 --> 0:03:43.470 including Apple, delaying products, paying more for memory, needing to 0:03:43.510 --> 0:03:46.870 charge customers more. But they've offset it in two key ways. 0:03:47.210 --> 0:03:49.770 One is on the storage tiers. So you can get 0:03:49.810 --> 0:03:53.370 the 256 gigabytes for that $ 2, 000 price point. You can 0:03:53.390 --> 0:03:56.930 get 512 for the $ 2, 200 price point. But if you 0:03:56.970 --> 0:03:58.590 want to go to the higher tiers, you're talking 2,600 0:03:58.590 --> 0:03:59.490 for a terabyte and then 3,200 for. 0:04:01.610 --> 0:04:02.620 For two terabytes. 0:04:02.720 --> 0:04:06.680 So they're charging massive premiums on that storage. And their 0:04:06.720 --> 0:04:09.580 thinking is that most people probably aren't going to get 0:04:09.620 --> 0:04:12.440 the base level version. This is an early adopter product 0:04:12.460 --> 0:04:14.880 for all intents and purposes. And that group of people 0:04:14.900 --> 0:04:17.820 is going to go for higher storage tiers. So they 0:04:17.860 --> 0:04:20.080 are going to get their margin on that. The other 0:04:20.100 --> 0:04:23.000 thing they've done is they've really jacked up pricing on 0:04:23.020 --> 0:04:28.090 the international customers. So in the US, right, the $ 2, 000 0:04:26.810 --> 0:04:28.750 price point globally is going to be the one everyone's 0:04:28.790 --> 0:04:32.270 talking about. But you go internationally, the prices in Canada, 0:04:32.310 --> 0:04:36.070 the price in Australia, the price in Asia, really sky high. 0:04:36.230 --> 0:04:38.529 And if you do some of the conversions, the prices 0:04:38.560 --> 0:04:41.159 in those other countries, it's not a one-to-one conversion for 0:04:41.200 --> 0:04:41.779 what the U.S. 0:04:41.839 --> 0:04:42.680 Price is. 0:04:42.760 --> 0:04:47.060 So, for instance, in Australia, there's an $ 800 Australian dollar 0:04:47.160 --> 0:04:50.810 premium on top of the converted rate. So that's how 0:04:50.830 --> 0:04:52.670 they're making up for that shortfall. 0:04:53.640 --> 0:04:56.400 Mark, to me, John Ternus tried to do two things 0:04:56.460 --> 0:05:00.289 with his first formal keynote as CEO. Talk about the 0:05:00.320 --> 0:05:03.750 smartphone being the device for the AI era, but also 0:05:03.830 --> 0:05:06.409 he put a lot of emphasis on your iPhone or 0:05:06.430 --> 0:05:09.230 whichever OS you're using knows you. It has your data. 0:05:09.250 --> 0:05:11.750 It has your context. What did you make of that 0:05:11.810 --> 0:05:16.130 messaging and Mr. Ternus's first full iPhone day event? 0:05:17.370 --> 0:05:19.790 Yeah, I mean, for the first time, Apple came out 0:05:19.930 --> 0:05:24.110 and showcased a real AI strategy. Whether it's an AI 0:05:24.150 --> 0:05:27.300 strategy that I agree with, that you agree with, anyone 0:05:27.330 --> 0:05:30.200 agrees with, it doesn't matter. At least they have a 0:05:30.260 --> 0:05:34.100 strategy now. And it's interesting they haven't really showcased a 0:05:34.140 --> 0:05:38.220 coherent strategy until now. He calls it the intelligent personal hub, 0:05:38.279 --> 0:05:42.279 the phone being the center of everything for artificial intelligence. 0:05:42.720 --> 0:05:46.720 It hearkens back to their previous intelligent hubs That all 0:05:46.760 --> 0:05:49.800 started with the Mac. Steve Jobs introduced the Mac as 0:05:49.839 --> 0:05:51.560 sort of the center of everything, and you have all 0:05:51.580 --> 0:05:53.330 those peripherals around it. 0:05:53.700 --> 0:05:57.050 The iPod, you name it. Then came iCloud. 0:05:57.510 --> 0:06:00.049 iCloud is the center of everything, and the Mac, the iPad, 0:06:00.089 --> 0:06:03.490 the phone, those are all peripherals around the cloud. Now, 0:06:03.610 --> 0:06:05.930 the center of everything is sort of that intelligent hub 0:06:06.330 --> 0:06:08.990 and the iPhone with everything built around that, whether it's 0:06:09.010 --> 0:06:11.890 your iPad, your Mac, your AirPods, your Apple Watch, and 0:06:11.910 --> 0:06:13.969 all the AI wearables are going to have in the future. 0:06:14.290 --> 0:06:16.710 It's all built around AI in the phone. So it's 0:06:16.730 --> 0:06:19.750 a pretty interesting strategy. I do think it's going to 0:06:19.830 --> 0:06:23.510 work to some extent, but obviously it's still early innings 0:06:23.550 --> 0:06:26.150 for their AI and there's more to come and there's 0:06:26.190 --> 0:06:26.570 more to do. 0:06:27.750 --> 0:06:30.110 Bloomberg's Mark Gurman, who leads our coverage of Apple, but 0:06:30.190 --> 0:06:34.440 also reported so deeply on iPhone Duo, including what its 0:06:34.480 --> 0:06:37.280 name would be prior to its announcement yesterday. Let's stick 0:06:37.300 --> 0:06:38.980 with Apple, but also we're going to talk more about 0:06:39.300 --> 0:06:43.720 technology markets. Nuveen's Chief Investment Officer, Sarah Malik, joins us now. Actually, 0:06:43.740 --> 0:06:46.490 the Apple move today is really interesting, up 2.5%. If 0:06:46.500 --> 0:06:48.150 you go back and look at sort of the average 0:06:48.170 --> 0:06:51.169 performance on the day itself of an iPhone event and 0:06:51.230 --> 0:06:54.290 next day, it doesn't often get the momentum we're seeing 0:06:54.330 --> 0:06:57.750 right now. I think you probably want to talk to 0:06:57.770 --> 0:06:59.810 what Mark was just talking about in terms of Apple 0:06:59.830 --> 0:07:02.779 having an AI strategy. That's been a bearing on the stock, Sarah. 0:07:02.820 --> 0:07:05.080 Your thoughts, please. Well, definitely. 0:07:05.100 --> 0:07:07.780 There's a lot of consumer interest and excitement over the 0:07:07.839 --> 0:07:11.060 new foldable phone. But there's two issues that investors are 0:07:11.160 --> 0:07:14.140 thinking about around Apple and this new phone. So, first 0:07:14.180 --> 0:07:18.030 of all, with the phone, the $ 2, 000 price point or 0:07:18.030 --> 0:07:20.830 $ 19. 99 may lead to a more modest addressable market from 0:07:20.850 --> 0:07:24.420 the start. And we'll grow from there. But secondarily, is 0:07:24.460 --> 0:07:27.240 this question around AI? The new CEO has a hardware 0:07:27.260 --> 0:07:30.440 background and Apple's very strong on hardware. But what investors 0:07:30.480 --> 0:07:33.460 want to know is, can Apple become more competitive in AI? 0:07:33.740 --> 0:07:36.360 They have somewhat lagged their peers in terms of their 0:07:36.400 --> 0:07:38.920 push into AI. So really, that is the key question 0:07:39.200 --> 0:07:42.300 for investors and Apple going forward. Is this AI strategy 0:07:42.340 --> 0:07:45.710 going to turn into something that has monetization for the company? 0:07:45.750 --> 0:07:47.960 Of course, this is a wait and see. So happy 0:07:47.990 --> 0:07:50.010 to see the stock up based on the excitement over 0:07:50.030 --> 0:07:51.990 a new phone. But I think the AI question is 0:07:52.030 --> 0:07:54.370 the medium term question for whether the stock. 0:07:54.230 --> 0:07:55.030 Outperforms or not. 0:07:56.010 --> 0:07:58.450 There is lots happening in financial markets right now. You 0:07:58.470 --> 0:08:01.710 look at oil, Brent, $ 105 a barrel. Yields are higher. 0:08:01.810 --> 0:08:05.230 Technology stocks are lower. Chip stocks are underperforming. And there 0:08:05.250 --> 0:08:08.460 are lots of pieces of economic data and macro events 0:08:08.500 --> 0:08:11.560 that are driving this market. For you this morning, what's 0:08:11.600 --> 0:08:12.619 going on? What's happening? 0:08:13.520 --> 0:08:16.200 It is all about higher rates and higher rates are 0:08:16.280 --> 0:08:20.260 tougher for long duration technology stocks. And really, even though 0:08:20.280 --> 0:08:22.760 we had a very strong second quarter earnings season driven 0:08:22.840 --> 0:08:27.610 by the tech sector, it's higher oil price prices, potentially 0:08:27.630 --> 0:08:31.250 higher rates at the FOMC meeting next week, a rate 0:08:31.270 --> 0:08:33.470 hike out of the ECB. All of that is causing 0:08:33.510 --> 0:08:37.050 this turmoil for the markets. The biggest data point we're 0:08:37.080 --> 0:08:39.960 going to see before the FOMC meets next week is CPI, 0:08:40.000 --> 0:08:43.260 which comes out tomorrow. If that number comes out hot, 0:08:43.280 --> 0:08:47.120 and by that I mean above 0.3% on the plus 0:08:47.140 --> 0:08:49.380 side month over month, I think you're going to lock 0:08:49.420 --> 0:08:51.340 in a hike for next week. If it comes in soft, 0:08:51.920 --> 0:08:55.210 under 0.2% month over month, that might cause the Fed 0:08:55.470 --> 0:08:58.870 to pause. But either way, we're seeing long-term treasury yields 0:08:59.150 --> 0:09:02.450 at up over 5%, oil prices over 105. That is 0:09:02.470 --> 0:09:04.730 very concerning for the markets. And there is no end 0:09:04.790 --> 0:09:07.290 in sight at this point for the geopolitical issues and 0:09:07.320 --> 0:09:07.689 the war. 0:09:07.580 --> 0:09:08.199 In the Middle East. 0:09:08.220 --> 0:09:10.540 So I think this turmoil for the markets is going 0:09:10.580 --> 0:09:13.840 to continue because it's hard to see the path for 0:09:13.900 --> 0:09:15.960 yields to start to decline from here, especially with the 0:09:15.980 --> 0:09:18.160 Fed either pausing or hiking rates next week. 0:09:19.200 --> 0:09:21.040 For me, Sarah, I'm going to go back to Apple, actually, 0:09:21.080 --> 0:09:23.380 because if you think about what the market, at least 0:09:23.420 --> 0:09:26.849 on the equity side, is doing this morning, Apple's massively 0:09:26.910 --> 0:09:30.510 outperforming from a points perspective at the index level. Would 0:09:30.530 --> 0:09:33.880 that indicate to you that investors are genuinely cheering what 0:09:33.960 --> 0:09:36.020 was outlined by the company yesterday? 0:09:37.160 --> 0:09:39.740 The positive here for Apple and out of the Goldman 0:09:39.760 --> 0:09:42.420 Sachs Technology Conference that's happening this week is we are 0:09:42.480 --> 0:09:46.920 hearing positive data points on monetization of artificial intelligence, which 0:09:46.980 --> 0:09:49.570 is what everybody has been waiting for. Today, what you're 0:09:49.590 --> 0:09:53.530 seeing for Apple is great. It's fundamentals outperforming the macro issue. 0:09:53.550 --> 0:09:56.850 So overall for the markets, the macro overhang of higher 0:09:56.990 --> 0:10:00.520 oil prices and potential spark of inflation because of that 0:10:00.920 --> 0:10:03.380 is definitely a cloud over the markets. But Apple's new 0:10:03.420 --> 0:10:07.180 product launch, it's exciting. It's a new style of phone. 0:10:07.480 --> 0:10:09.020 It is at a higher price point, but that new 0:10:09.059 --> 0:10:12.080 style of phone can start a new product cycle for Apple. 0:10:12.100 --> 0:10:14.640 And that's what people are excited about. But as I mentioned, 0:10:14.660 --> 0:10:16.760 there is a key question for Apple, which is how 0:10:16.840 --> 0:10:19.860 are they going to gain more traction in in AI 0:10:19.960 --> 0:10:22.140 with their phones. And it's great to see a strategy. 0:10:22.520 --> 0:10:24.109 And now we'd like to see the traction and the 0:10:24.150 --> 0:10:25.209 monetization from that. 0:10:26.390 --> 0:10:29.209 We are kind of late in a technology earnings cycle 0:10:29.230 --> 0:10:33.190 where they keep coming. Oracle and Adobe's after the bell. 0:10:33.990 --> 0:10:37.480 One software name, one infrastructure name with a software element. 0:10:37.600 --> 0:10:41.060 How much of importance do investors continue to assign the 0:10:41.100 --> 0:10:43.340 data points coming out of earnings in terms of where 0:10:43.380 --> 0:10:44.500 we are in the cycle right now? 0:10:45.580 --> 0:10:49.170 Well, every technology data point counts because technology earnings are 0:10:49.230 --> 0:10:52.330 what are leading overall earnings growth, especially for the S & 0:10:52.050 --> 0:10:54.980 P 500 and in the U.S. So Oracle and Adobe 0:10:55.240 --> 0:10:59.600 will be important. We may get some commentary around hyperscalers 0:10:59.740 --> 0:11:03.750 and CapEx and whether we're seeing any monetization. from AI investments, 0:11:03.830 --> 0:11:05.630 all of that's going to be important. But I think 0:11:05.650 --> 0:11:08.290 the issue is with an FOMC meeting next week and 0:11:08.330 --> 0:11:11.410 CPI data coming out tomorrow, people are more focused right 0:11:11.470 --> 0:11:14.790 now on what do these higher oil prices mean for inflation? 0:11:15.130 --> 0:11:17.830 What does it mean for the US economy? As an example, 0:11:18.130 --> 0:11:21.709 we did get very strong payrolls last week, but if 0:11:22.090 --> 0:11:24.830 inflation stays high and we become worried that oil prices 0:11:24.850 --> 0:11:27.550 will cause the economy to slow, then we could start 0:11:27.590 --> 0:11:30.130 talking about stagflation. I think all of that for now 0:11:30.620 --> 0:11:32.600 is going to be a bigger issue for the markets 0:11:32.970 --> 0:11:35.660 than a couple of more incremental technology earnings data points. 0:11:36.179 --> 0:11:38.940 We've covered an iPhone day. We covered technology earnings. We 0:11:38.960 --> 0:11:42.350 covered macros. We covered higher rates. We covered oil. We 0:11:42.370 --> 0:11:44.390 covered inflation and then the Fed meeting next week. That's 0:11:44.450 --> 0:11:47.610 pretty good going. Nuveen Chief Investments Officer Sarah Malik back 0:11:47.650 --> 0:11:51.250 on Bloomberg Tech. Thank you very much. Coming up, D-Matrix 0:11:51.410 --> 0:11:55.620 is plugging into NVIDIA's AI ecosystem. CEO Sid Sheff joins 0:11:55.700 --> 0:11:58.160 us on the new partnership and why the next big 0:11:58.260 --> 0:12:01.700 battle in AI chips is inference. This is Bloomberg Tech. 0:12:09.030 --> 0:12:12.770 TSMC reported better than expected sales in August, but its 0:12:12.860 --> 0:12:16.179 US-listed shares are under pressure. As investors and analysts start 0:12:16.200 --> 0:12:18.800 to worry the company doesn't have enough supply to meet 0:12:18.840 --> 0:12:21.780 robust demand, let's discuss with Bloomberg's Peter Elstrom. I mean, 0:12:21.800 --> 0:12:24.360 this is the dynamic of industry right now. In the 0:12:24.460 --> 0:12:27.400 AI context, demand for lead-edge chips is running ahead of 0:12:27.440 --> 0:12:30.740 the industry's ability to supply. What did TSMC say? 0:12:32.260 --> 0:12:34.780 Yeah, the numbers are pretty staggering. I think we're familiar 0:12:34.830 --> 0:12:37.310 with the trend, but the numbers are quite staggering here. 0:12:37.350 --> 0:12:41.190 So First of all, TSMC reported a 53% rise in 0:12:41.250 --> 0:12:44.770 revenue for August. Taiwan does this thing where they have 0:12:44.830 --> 0:12:47.150 companies report every month. So we get a sense of 0:12:47.200 --> 0:12:49.400 how the quarter is going for them. That's above the 0:12:50.210 --> 0:12:53.189 expectations that analysts have for the company to boost revenues 0:12:53.250 --> 0:12:56.290 about 47% for the quarter. We had an interview with 0:12:56.330 --> 0:12:58.630 the COO though, Cliff Ho, and he talked to us 0:12:58.690 --> 0:13:01.770 a bit about the constraints that they've got on supply 0:13:01.830 --> 0:13:04.270 right now. They say that typically in a typical year, 0:13:04.590 --> 0:13:07.790 they'd be building four or five fabs someplace. Right now 0:13:07.840 --> 0:13:10.510 they're trying to build 20 fabs. They're so far behind 0:13:10.600 --> 0:13:12.840 on demand here that they just can't keep up. So 0:13:12.860 --> 0:13:14.820 they're trying to build fabs as quickly as they can. 0:13:14.880 --> 0:13:18.160 Of course, fabs takes years to build. They're very complicated. 0:13:18.200 --> 0:13:20.910 They're very expensive. TSMC is the best in the world 0:13:21.010 --> 0:13:23.680 at building them, but they're having a hard time getting 0:13:23.740 --> 0:13:26.640 them built in as quick of a manner as they 0:13:26.679 --> 0:13:28.480 would like right now. And partly it's what you were 0:13:28.520 --> 0:13:32.220 talking about before. Apple certainly wants more advanced chips, especially 0:13:32.280 --> 0:13:35.490 Nvidia in the AI ecosystem wants to be able to 0:13:35.520 --> 0:13:38.330 get more capacity online as soon as they possibly. 0:13:37.910 --> 0:13:40.510 Can get it. That's on the manufacturing side. On the 0:13:40.590 --> 0:13:43.589 fabless chip maker side, MediaTek, a company we've been covering 0:13:43.610 --> 0:13:46.870 very closely on the show recently, out with very strong numbers, 0:13:46.929 --> 0:13:48.270 a lot of momentum for their chips. 0:13:49.530 --> 0:13:52.840 Yeah, I mean, you alluded to this earlier that we've 0:13:52.860 --> 0:13:55.260 had MediaTek CEO on the show talking about this new 0:13:55.300 --> 0:13:59.610 joint venture that they've got with NVIDIA in particular. MediaTek 0:14:00.050 --> 0:14:03.130 is also growing very quickly. They reported strong revenue growth 0:14:03.170 --> 0:14:06.630 of 44%. They historically have done very well on the 0:14:06.650 --> 0:14:10.000 smartphone market. Now they're moving into these adjacent markets, especially 0:14:10.050 --> 0:14:13.480 in AI services. They've struck this deal with NVIDIA where 0:14:13.500 --> 0:14:17.060 they want to be able to incorporate their technology in 0:14:17.100 --> 0:14:19.840 with NVIDIA's technology. They want to also move into being 0:14:19.880 --> 0:14:22.740 able to make chips that will work in laptops and 0:14:22.780 --> 0:14:26.250 computers and bring more AI capabilities there too. So the 0:14:26.290 --> 0:14:28.750 company is growing quite strong at this point. They also 0:14:28.830 --> 0:14:29.970 are trying to keep up with demand. 0:14:31.390 --> 0:14:34.710 Bloomberg's Peter Elstrom, who leads our coverage of Asian technology companies. 0:14:34.730 --> 0:14:37.460 Thank you very much indeed. NVIDIA is confronting a world 0:14:37.800 --> 0:14:40.880 of more custom chips for more players and growth in 0:14:40.900 --> 0:14:44.490 specialized inference systems. Its response has been embrace it. 0:14:45.310 --> 0:14:47.930 There's no question NVIDIA's footprint is the largest. There's no 0:14:47.970 --> 0:14:52.680 question we're growing share in closed models and open models. 0:14:53.100 --> 0:14:58.260 We're growing share across the entire AI opportunity. And so 0:14:58.500 --> 0:15:04.140 we're not intimidated by this. We're, in fact, welcoming it 0:15:04.640 --> 0:15:07.460 and opening our platform so that they could connect XPUs 0:15:07.500 --> 0:15:07.800 into it. 0:15:08.880 --> 0:15:11.060 So that was NVIDIA CEO Jensen Wang speaking to me 0:15:11.180 --> 0:15:14.480 after the deal with MediaTek on XPUs last month. Today, 0:15:14.980 --> 0:15:18.150 startup D-Matrix is the latest to partner with NVIDIA to 0:15:18.170 --> 0:15:22.690 plug its specialized inference XPUs into NVIDIA's platforms, targeting ultra-low 0:15:22.730 --> 0:15:27.620 latency AI workloads, where its tech works alongside NVIDIA GPUs 0:15:27.840 --> 0:15:31.590 using NVLink fusions. Dematrix CEO Sid Shash joins us now. 0:15:32.360 --> 0:15:36.380 Very interesting. Why is it a benefit to you to 0:15:36.400 --> 0:15:40.640 be able to plug into a largely NVIDIA-based architecture? Yes. 0:15:40.800 --> 0:15:42.560 First of all, Ed, I'm a fan of the show, 0:15:42.580 --> 0:15:45.200 so thanks for having me on it. This is a 0:15:45.240 --> 0:15:51.550 landmark announcement for Dematrix, obviously, and the biggest reason we 0:15:51.670 --> 0:15:54.110 embarked on this partnership with NVIDIA, and we've been working 0:15:54.170 --> 0:15:58.180 on this for six-plus months with them, is exactly what 0:15:58.220 --> 0:16:03.040 Jensen said in his comments earlier. This is probably, most certainly, 0:16:03.600 --> 0:16:07.660 the largest deployed infrastructure base for AI in the world. 0:16:08.660 --> 0:16:09.460 It is in every cloud. 0:16:10.120 --> 0:16:13.440 NVIDIA hardware is in every cloud, every country, pretty much 0:16:13.520 --> 0:16:16.910 every data center in the world. At D-Matrix, we have 0:16:16.930 --> 0:16:20.510 a very unique memory-centric inference computing technology. We have been 0:16:21.010 --> 0:16:24.350 working on this for seven plus years, really targeted for 0:16:24.450 --> 0:16:29.880 ultra-low latency inference. And ultra low latency inference computing just 0:16:30.010 --> 0:16:33.060 took off in a massive way in the last 12 months, 0:16:33.160 --> 0:16:36.620 post GPT Codecs and Cloud Code and with the arrival 0:16:36.660 --> 0:16:40.340 of agentic coding, everyone is looking for how they can 0:16:40.400 --> 0:16:43.790 quickly get access to low latency compute so that they 0:16:43.810 --> 0:16:46.610 can really interact with these agentic coding tools. 0:16:46.770 --> 0:16:48.220 And that's where D-Matrix comes in. 0:16:49.210 --> 0:16:52.410 And this partnership will really allow us to take our 0:16:52.430 --> 0:16:55.770 products to market very rapidly with the NVIDIA partnership. 0:16:57.120 --> 0:17:02.260 Sid, in Mediatek's case, NVIDIA made a very significant investment 0:17:02.620 --> 0:17:06.840 through the purchase of convertible notes. Under your arrangement with NVIDIA, 0:17:06.880 --> 0:17:09.350 are they making an equity or any other kind of 0:17:09.420 --> 0:17:12.550 investment into D-Matrix to help you get moving? 0:17:13.570 --> 0:17:15.770 At this point, Ted, I couldn't comment on the investment 0:17:15.810 --> 0:17:20.159 piece specifically, but look, this is a partnership that is 0:17:20.300 --> 0:17:24.320 very strategic to NVIDIA. And they are putting a lot 0:17:24.340 --> 0:17:30.330 of resources, engineering resources. There is obviously a lot of 0:17:30.530 --> 0:17:33.170 work the two teams have been doing for six plus months. 0:17:34.210 --> 0:17:37.390 And our first goal is to really make sure that 0:17:37.430 --> 0:17:40.130 the Raptor XPU that is going to be integrated into 0:17:40.170 --> 0:17:43.020 the NVIDIA ecosystem is going to be ready to come 0:17:43.060 --> 0:17:47.040 to market in about 12 months. So the biggest opportunity 0:17:47.100 --> 0:17:50.460 here is really about getting that product in customers' hands 0:17:50.520 --> 0:17:53.050 because we've got a lot of customer pull for that 0:17:53.170 --> 0:17:55.050 specific product in the NVIDIA ecosystem. 0:17:55.650 --> 0:17:57.950 So on your newest chip or your newest platform, I 0:17:57.990 --> 0:18:00.950 should say, Is there any data point you can give 0:18:00.990 --> 0:18:04.109 me on how the partnership with NVIDIA tangibly allows you 0:18:04.130 --> 0:18:06.750 to move faster? In other words, will this deploy into 0:18:06.790 --> 0:18:09.550 the real world quicker than it otherwise would had you 0:18:09.609 --> 0:18:12.070 not struck a deal with NVIDIA for NVLink Fusion? 0:18:12.790 --> 0:18:18.010 Yeah, 100%. Look, the Raptor XPU, which is the specific 0:18:18.109 --> 0:18:22.189 XPU that's in partnership with NVIDIA and will be launched 0:18:22.290 --> 0:18:24.570 as part of this partnership first, is going to be 0:18:24.590 --> 0:18:30.760 the world's first 3D-stacked DRAM XPU, right? And just let 0:18:30.780 --> 0:18:33.419 me touch upon why that is so important. It is 0:18:33.460 --> 0:18:37.800 important because memory is really the biggest bottleneck for AI 0:18:37.840 --> 0:18:38.139 right now. 0:18:38.240 --> 0:18:39.060 And we all know that. 0:18:39.460 --> 0:18:41.850 And what we have done is instead of using HBM, 0:18:41.869 --> 0:18:45.010 we don't use any HBM at D-Matrix. What we have 0:18:45.050 --> 0:18:48.629 done is we have taken DRAM and found a way 0:18:49.190 --> 0:18:51.870 to package it directly. 0:18:51.510 --> 0:18:52.370 With the compute. 0:18:53.270 --> 0:18:56.230 And it is 3D stacked exactly like you just saw 0:18:56.270 --> 0:19:00.870 in the picture there. And that allows for the architecture 0:19:00.910 --> 0:19:03.350 to punch through the memory wall, which is really the 0:19:03.490 --> 0:19:07.310 biggest bottleneck in AI today in terms of addressing latency, 0:19:07.410 --> 0:19:11.920 addressing energy efficiency, addressing cost. And we have found a recipe. 0:19:11.980 --> 0:19:13.550 We have cracked the code on how to do it. 0:19:13.640 --> 0:19:15.800 And we'll be the first to bring that solution to market. 0:19:16.220 --> 0:19:19.160 So there is tremendous customer pull for that product right 0:19:19.200 --> 0:19:22.400 now because no other company is going to be within 0:19:22.440 --> 0:19:25.520 a two-year window of getting access to that technology. And 0:19:26.040 --> 0:19:28.570 this NVIDIA partnership will truly allow us to take that 0:19:28.590 --> 0:19:31.250 to market the fastest because there is already a deployed 0:19:31.290 --> 0:19:33.270 base of all that infrastructure and all the data centers. 0:19:33.310 --> 0:19:35.170 And we don't want to reinvent that part of the wheel. 0:19:35.650 --> 0:19:37.889 We'd much rather just ride on the NVIDIA ecosystem and 0:19:37.930 --> 0:19:40.129 get the product to our customers as soon as possible. 0:19:40.790 --> 0:19:42.650 So, Sid, we're short of time here. Who are those 0:19:42.690 --> 0:19:45.010 new customers? If I look at your cap table, there's 0:19:45.050 --> 0:19:47.270 lots of sovereign wealth funds on there who have lots 0:19:47.310 --> 0:19:49.930 of national AI projects. We just have 30 seconds. 0:19:50.720 --> 0:19:51.160 Yes, yes. 0:19:51.880 --> 0:19:54.520 It's a combination, Ed, and we'll be announcing these customers. 0:19:54.560 --> 0:19:58.160 Stay tuned. It is a combination of hyperscalers, frontier labs, sovereigns, 0:19:58.220 --> 0:20:00.660 inference clouds, high-frequency traders. 0:20:01.740 --> 0:20:02.160 Stay tuned. 0:20:02.180 --> 0:20:04.290 We're going to be announcing these customers in pretty short order. 0:20:05.400 --> 0:20:08.060 Sid Sheff of D-Matrix, thank you so much for your time. 0:20:08.440 --> 0:20:11.510 There's another important story. As NVIDIA expands its ties with 0:20:11.550 --> 0:20:15.450 specialized AI chip makers, one of those deals is drawing scrutiny. 0:20:15.609 --> 0:20:20.390 The Justice Department is investigating NVIDIA's $ 20 billion tie-up with 0:20:20.470 --> 0:20:24.229 AI inference startup GROK and whether it was structured to 0:20:24.310 --> 0:20:29.110 avoid an antitrust review, according to sources. Grok remained independent, 0:20:29.230 --> 0:20:32.590 but its CEO and COO moved to NVIDIA as a 0:20:32.609 --> 0:20:35.480 part of the deal. US lawmakers have called the arrangement 0:20:35.540 --> 0:20:40.480 a takeover that stifled competition. NVIDIA representatives and a DOJ 0:20:40.540 --> 0:20:43.859 spokesperson declined to comment. A representative for Grok did not 0:20:43.920 --> 0:20:45.940 respond to requests for comment. 0:20:47.540 --> 0:20:49.750 It's time now for Talking Tech. I'm Yahaira Ana. And 0:20:49.830 --> 0:20:53.190 first up, OpenAI is ending its deal for government agencies 0:20:53.250 --> 0:20:57.119 to use its models for $ 1 per year. That's according 0:20:57.150 --> 0:21:01.160 to the General Services Administration. Starting October 1st, agencies will 0:21:01.200 --> 0:21:03.660 have to pay based on how much they use them 0:21:03.820 --> 0:21:08.419 with a 50% discount off standard pricing. Plus, Anthropic is 0:21:08.460 --> 0:21:12.600 giving the EU's cybersecurity agency access to its most powerful 0:21:12.940 --> 0:21:16.390 mythos AI model after months of back and forth over 0:21:16.470 --> 0:21:18.930 who should be allowed to use it. Of course, Anthropic 0:21:18.970 --> 0:21:22.260 has tightly restricted access because it says the model is 0:21:22.410 --> 0:21:26.460 so powerful it can uncover cybersecurity flaws on its own. 0:21:27.660 --> 0:21:31.020 And DeepSeek unveiled its latest AI model that charges as 0:21:31.180 --> 0:21:35.140