WEBVTT - How the Number One U.S. Semiconductor Company Stumbled 0:00:10.840 --> 0:00:14.720 Hello, and welcome to another episode of the Odd Lots Podcast. 0:00:14.800 --> 0:00:19.959 I'm Joe Wisenthal and I'm Tracy Allaway. So, Tracy, you know, 0:00:20.480 --> 0:00:23.080 we haven't talked about it in a while, but one 0:00:23.400 --> 0:00:27.720 recurring topic on the podcast over the years has been 0:00:28.480 --> 0:00:33.360 China's efforts to really sort of leap into the lead technologically. 0:00:33.520 --> 0:00:36.839 We've talked about it with a few different guests, invested 0:00:36.920 --> 0:00:43.479 aggressively in airlines, UH, semiconductors, UH and so forth, setting 0:00:43.520 --> 0:00:45.479 us at all the current noise. That's still like one 0:00:45.479 --> 0:00:48.880 of the bigger long term storylines that attempt to sort 0:00:48.880 --> 0:00:51.520 of supplant the US at the lead in the lead 0:00:51.560 --> 0:00:55.960 of a technological manufacturing and development. Yeah, that's right, and 0:00:56.000 --> 0:00:59.920 I think that's where a big portion of the trade war, 0:01:00.240 --> 0:01:03.200 or the Trump administration's trade war with China actually came from. 0:01:03.280 --> 0:01:05.959 And that's why we see so much of it centered 0:01:06.080 --> 0:01:10.960 around technology things like Bye Dance, TikTok and UH of 0:01:11.000 --> 0:01:15.080 course five G and Huawei and UH. Of course we're 0:01:15.120 --> 0:01:18.679 seeing Beijing rolling out it's it's new plans for I 0:01:18.680 --> 0:01:20.680 think it's for the next twenty years or something like that. 0:01:20.720 --> 0:01:24.240 And of course tech dominance features quite highly. Once again, 0:01:25.880 --> 0:01:28.920 absolutely but it also like it takes it takes two 0:01:28.959 --> 0:01:34.280 to tango because in addition to UH China's own endeavors 0:01:34.280 --> 0:01:37.520 to leap frog or surpassed the US in terms of 0:01:37.560 --> 0:01:41.560 technological development, there's also an another element, which is that 0:01:42.040 --> 0:01:44.440 a lot of the US companies that we sort of 0:01:44.480 --> 0:01:48.640 associate with, you know, the great heritage of US manufacturing, 0:01:49.240 --> 0:01:54.320 whether it's Boeing, g E or Intel with chips, all 0:01:54.360 --> 0:01:57.240 of them seemed to be not doing so great. So 0:01:57.320 --> 0:02:00.800 literally going in the other direction. Yeah, so this is 0:02:00.840 --> 0:02:03.840 something that I wasn't that aware of just because I've 0:02:03.840 --> 0:02:06.800 been outside of the US for a while. But Intel 0:02:07.280 --> 0:02:10.000 seems to be struggling. So I think in the summer 0:02:10.040 --> 0:02:13.160 they said they were something like twelve months behind schedule 0:02:13.200 --> 0:02:16.440 when it comes to developing like their new next generation 0:02:17.120 --> 0:02:20.480 of chips, and the share prices come down quite a 0:02:20.480 --> 0:02:23.680 bit since then. And it's sort of the polar opposite 0:02:23.720 --> 0:02:26.040 to what you see happening at some other chip companies 0:02:26.080 --> 0:02:29.560 like t S mc over in Taiwan for instance, they 0:02:29.600 --> 0:02:32.280 seem to be doing quite well during the pandemic, well 0:02:32.320 --> 0:02:35.560 relatively well yeah, I mean you haven't exactly right. There 0:02:35.760 --> 0:02:39.160 seems to be going in the exact opposite direction, falling 0:02:39.240 --> 0:02:42.560 further and further behind, and their stock has really been 0:02:42.840 --> 0:02:46.239 got hammered this summer. So, like what happened is sort 0:02:46.280 --> 0:02:48.840 of an interesting question because we could talk about China's 0:02:48.840 --> 0:02:54.760 research effort, but how this company that basically invented the industry, 0:02:54.840 --> 0:02:56.720 one of the you know, to put the silicon and 0:02:56.760 --> 0:03:01.680 Silicon valley, so to how it lost its lead and 0:03:01.760 --> 0:03:04.880 ability to be one of the world pre eminent manufacturers. 0:03:05.120 --> 0:03:10.000 Is itself pretty interesting storyline here. Yeah, absolutely, And I 0:03:10.080 --> 0:03:13.160 mean I remember during the early two thousand's Intel was 0:03:13.280 --> 0:03:17.040 this massive, massive company and it had the branding and 0:03:17.040 --> 0:03:20.160 the advertising around it and just a huge deal. So 0:03:20.240 --> 0:03:24.200 it's interesting to see it um sort of turning around 0:03:24.320 --> 0:03:27.920 this year. So we're going to talk more about Intel 0:03:27.960 --> 0:03:30.919 and what happened, and I'm very excited about our guest. 0:03:31.080 --> 0:03:35.480 We're going to be speaking with Stacy Raskin. He's managing 0:03:35.520 --> 0:03:42.480 director Senior analyst U s Semiconductors at Bernstein Research, and notably, 0:03:42.760 --> 0:03:46.040 at a recent uh Intel conference call after one of 0:03:46.040 --> 0:03:49.680 their quarters, the company didn't call on him, leading him 0:03:49.720 --> 0:03:53.320 to a subtweet a company on Twitter, which is always 0:03:53.360 --> 0:03:56.560 fun and doesn't happen that often with analysts. So Stacy, 0:03:56.640 --> 0:03:59.080 thank you very much for joining us. Oh my pleasure. 0:03:59.080 --> 0:04:02.040 I'm glad to be here. And by the way, um, uh, 0:04:02.080 --> 0:04:06.760 nobody knows what that tweet was about. Okay, it was 0:04:06.960 --> 0:04:10.920 a subtweet of anything. Okay, So just on the day 0:04:11.000 --> 0:04:14.240 that you didn't get called on by until during their 0:04:14.320 --> 0:04:18.680 quarterly conference call, you tweeted cowards dot dot dot. But 0:04:18.800 --> 0:04:21.440 we'll we'll all pretend that we have no idea what 0:04:21.480 --> 0:04:24.279 that was about. Maybe it was something completely unrelated to 0:04:24.320 --> 0:04:26.880 your professional career. This is one of the mysteries of 0:04:26.920 --> 0:04:29.520 the universe. We'll leave it at that. It'll be a 0:04:29.560 --> 0:04:34.240 mystery forever. We'll just leave that. We'll just leave that tweet. Um, 0:04:34.320 --> 0:04:36.320 but it is true that you didn't get called on 0:04:37.440 --> 0:04:41.400 that part. That is true. So let's forgetting the subtweet 0:04:41.400 --> 0:04:43.680 aside or whether it's a subtweet, what were you going 0:04:43.760 --> 0:04:46.560 to ask? Like, uh, you know, if here's your chance, 0:04:47.240 --> 0:04:51.760 Stacy Raskin at Bernstein Research Europe, what would you have asked? Oh? Boy, 0:04:51.760 --> 0:04:53.240 I don't even remember what I was going to ask 0:04:53.240 --> 0:04:54.360 at this point, and I have to go back and 0:04:54.400 --> 0:04:56.839 look at my question list. Um, I'm sure it was 0:04:56.920 --> 0:05:01.159 something pointed though, Um, it typically as I have a 0:05:01.200 --> 0:05:05.480 reputation I suppose for asking those kinds of questions. And 0:05:05.560 --> 0:05:07.920 it's not really my fault. I can't help it. My 0:05:07.920 --> 0:05:11.200 my BS detection threshold is set kind of low, and 0:05:11.279 --> 0:05:13.120 I tend to forget who I'm talking to once I 0:05:13.160 --> 0:05:15.080 get going, So I can't help it. But it doesn't 0:05:15.080 --> 0:05:19.040 always engender a lot of love. Um, but that's okay. 0:05:19.279 --> 0:05:21.320 I don't have to uh, I don't have to believe 0:05:21.440 --> 0:05:24.039 this is gonna be a good conversation. Yeah, this, this 0:05:24.080 --> 0:05:27.280 should be a great conversation. I'm looking forward to it. Well, well, 0:05:27.080 --> 0:05:31.000 we'll see if that threshold gets hit or not. I 0:05:31.000 --> 0:05:35.520 feel like cell side analysts with low BS detector is 0:05:35.600 --> 0:05:38.640 fairly unusual, or at least we don't get to talk 0:05:38.680 --> 0:05:41.320 to many of them. Um, it goes with the territory. 0:05:42.360 --> 0:05:45.279 Can you set the scene for us, like, how bad 0:05:45.560 --> 0:05:48.680 are things at Intel at the moment? In your view, 0:05:48.760 --> 0:05:51.159 like Joe said, you're sort of noted critic of the 0:05:51.200 --> 0:05:54.400 company recently. What's going on there? How bad is it? 0:05:55.560 --> 0:05:58.160 I mean, it's it's problematic right now. And so you know, 0:05:58.200 --> 0:06:02.320 you started out the beginning of this call talking about 0:06:02.360 --> 0:06:05.960 some of the issues they're having with manufacturing. Um It's 0:06:05.960 --> 0:06:09.440 important to note these are not new issues. So the 0:06:09.839 --> 0:06:13.360 current issues are with seven nimeal seven animeter process, and 0:06:13.400 --> 0:06:17.120 you're right. In July they disclosed further delays there. They 0:06:17.120 --> 0:06:19.640 said the process was was twelve months delayed. They said 0:06:19.640 --> 0:06:22.680 the products, I think it versus the prior road net 0:06:22.680 --> 0:06:25.520 were six months delayed. That being said, this is not 0:06:25.600 --> 0:06:28.120 the first time they've had they've had a problem ten nnimeters, 0:06:28.160 --> 0:06:32.159 which was the prior generation manufacturing technology. And if you want, 0:06:32.200 --> 0:06:33.920 we can go into what the numbers and everything mean, 0:06:34.000 --> 0:06:36.880 but just in general, ten anners was the prior one. 0:06:37.320 --> 0:06:39.080 They had big problems with that one as well. That 0:06:39.080 --> 0:06:42.080 one was delayed for five or six years and forced 0:06:42.080 --> 0:06:44.680 them to sit on on the you know, the two 0:06:44.720 --> 0:06:47.440 generations back, which was called fourteen animeters. They were sitting 0:06:47.440 --> 0:06:49.480 on that one for many more years than they intended. 0:06:50.080 --> 0:06:53.680 But even fourteen nnimeters was delayed a little bit when 0:06:53.680 --> 0:06:55.640 they when they originally launched it, it it was it took 0:06:55.680 --> 0:06:58.920 about a year longer than they had originally anticipated to 0:06:59.080 --> 0:07:03.440 reach sort of like fully manufacturable yields and reach full volume. 0:07:03.520 --> 0:07:06.520 So these are not new problems that Intel. These have 0:07:06.560 --> 0:07:10.440 been building over probably half a decade plus and they 0:07:10.520 --> 0:07:13.240 finally hit a wall. So I mean that that's sort 0:07:13.240 --> 0:07:16.640 of the current state of affairs. So Tracy and I 0:07:16.720 --> 0:07:19.760 aren't technologists, but we see these terms or hear them 0:07:19.840 --> 0:07:23.440 ten men, ten nanimator, seven min nanimator. Why don't you 0:07:23.560 --> 0:07:26.600 take this moment to describe like what these means? I mean, 0:07:26.640 --> 0:07:29.560 I get it, okay, smaller and smaller chips and efficiency 0:07:29.640 --> 0:07:31.120 and all that, but talk to us a little bit 0:07:31.160 --> 0:07:33.720 about like what we're really talking about. Yeah, But so 0:07:33.840 --> 0:07:36.400 in theory, like it used to be, like some measure 0:07:36.480 --> 0:07:38.800 of the size of the transistor there was called the 0:07:38.840 --> 0:07:42.640 gate length that at one point um. In reality, these 0:07:42.720 --> 0:07:45.440 numbers are complete in total marketing. They don't actually mean 0:07:45.520 --> 0:07:49.360 anything in isolation um, and they're not comparable from one 0:07:49.440 --> 0:07:52.280 company to another. So, for example, an Intel tenometer part 0:07:52.400 --> 0:07:54.080 is not the same as as a t SMC ten 0:07:54.240 --> 0:07:57.280 nimet part. I'll give you an example of this, by 0:07:57.320 --> 0:07:59.360 the way, And when when t SMC went from what 0:07:59.480 --> 0:08:03.520 they purport really called twenty nimes to sixteen, they didn't 0:08:03.560 --> 0:08:05.480 shrink the transistors at all. What they did is they 0:08:05.560 --> 0:08:08.760 changed the structure of those transistors. Um. They went from 0:08:08.840 --> 0:08:11.760 what and again, I we can talk about what these 0:08:11.840 --> 0:08:13.880 terms mean. But they went from a transistor that was 0:08:13.920 --> 0:08:17.560 called a planar transistor structure to a three D or 0:08:17.600 --> 0:08:20.760 a fin fet transistor structure. But they didn't make them 0:08:20.760 --> 0:08:23.360 any smaller, They didn't squeeze them closer together, they didn't 0:08:23.400 --> 0:08:26.000 do anything. They just went from planer to fin fet 0:08:26.160 --> 0:08:29.400 and called it sixteen no shrink, know nothing. So that's 0:08:29.400 --> 0:08:32.000 so that's that's the thing that the members themselves actually 0:08:32.040 --> 0:08:36.080 stopped really having actual meaning or more than more than 0:08:36.120 --> 0:08:38.800 a decade ago. It's been marketing ever ever since. What 0:08:38.960 --> 0:08:40.960 you can think about them though, in terms of meaning, 0:08:41.000 --> 0:08:46.000 it is typically some measure of improved transistor density. So 0:08:46.120 --> 0:08:49.400 squeezing more and more transistors onto a given area of silicon, 0:08:49.480 --> 0:08:52.640 it is some measure of that. And with every further note, 0:08:52.800 --> 0:08:55.640 in every further transition in theory, you should be getting 0:08:55.679 --> 0:08:59.720 better performance out of these chips and and better power efficiency. 0:09:00.160 --> 0:09:01.839 This is really what Moore's law is. By the way, 0:09:01.840 --> 0:09:03.520 it's sort of these three legs in the stools. It's 0:09:03.559 --> 0:09:06.640 better performance, better power, and lower cost. And by the way, 0:09:06.640 --> 0:09:08.959 when people talk about Moore's law dying, which is sort 0:09:08.960 --> 0:09:11.319 of the situation we're in right now. It does not 0:09:11.520 --> 0:09:13.800 mean that it's impossible to to to shrink these things 0:09:13.920 --> 0:09:16.680 to make the transistors any smaller. What it actually means 0:09:16.720 --> 0:09:18.600 is that the cost leg of that three legged stool 0:09:18.640 --> 0:09:21.040 is going out the window. So we can still do this, 0:09:21.200 --> 0:09:23.599 like engineers are smart. If there's a if there's a 0:09:23.720 --> 0:09:25.319 business case to shrink, they'll do it. But now we 0:09:25.400 --> 0:09:27.440 have to pay for it, whereas before it we used 0:09:27.480 --> 0:09:30.160 to get it every two years for free. It was fantastic, right, 0:09:30.160 --> 0:09:32.120 But that's kind of the situation of what's going on now. 0:09:34.080 --> 0:09:38.360 So we keep comparing in this discussion Intel to uh 0:09:38.960 --> 0:09:43.120 T SMC. But my understanding is they're they're not exactly 0:09:43.280 --> 0:09:46.360 the same company, right, the business model is slightly different. 0:09:46.400 --> 0:09:50.120 Could you maybe explain how they differ? You bet? So. 0:09:50.200 --> 0:09:54.400 There are two primary um uh semiconductor broad semic connector 0:09:54.440 --> 0:09:57.800 business models in the industry. The first is called they 0:09:57.840 --> 0:10:00.520 called an i DM or integrated device man facturer, and 0:10:00.600 --> 0:10:03.400 these are companies of both design and manufacture their own 0:10:03.440 --> 0:10:07.719 ships like Intel. Right. The issue is, you know, a 0:10:07.880 --> 0:10:11.839 leading edge semiconductor manufacturing facility can cost these days, you know, 0:10:12.000 --> 0:10:15.520 ten billion dollars, Like it's very expensive. And if you're 0:10:15.520 --> 0:10:17.319 going to build a factory that costs that much, you 0:10:17.440 --> 0:10:19.360 better have a lot of revenue in order to put 0:10:19.400 --> 0:10:21.160 through in order to cover those costs, and most most 0:10:21.240 --> 0:10:24.400 semi conduct companies do not. And so decades ago this 0:10:24.559 --> 0:10:29.000 became clear and the supply chain atomized, it split apart, 0:10:29.080 --> 0:10:31.280 and he gave rise to what's known as the fabulous 0:10:31.440 --> 0:10:35.320 foundry model, where you have companies, um say, like an 0:10:35.400 --> 0:10:37.320 Nvidio or a qual Colm. These are referred to as 0:10:37.400 --> 0:10:39.880 fabulous companies in the sense that they do not have 0:10:40.080 --> 0:10:43.760 fabs fab as as a semiconductor manufacturing facility. They only 0:10:43.840 --> 0:10:47.040 design chips and then they outsource the manufacturing to a 0:10:47.120 --> 0:10:49.679 company known as a foundry. This is what t SMC does, 0:10:50.240 --> 0:10:52.800 and they're able because of this model. They can agglomerate 0:10:52.960 --> 0:10:56.760 demand together from many semiconductor companies and build up the 0:10:56.840 --> 0:10:59.840 revenue scale that is required to support them the high 0:11:00.040 --> 0:11:04.120 costs of of of manufacturing um that ordinarily all these 0:11:04.120 --> 0:11:05.840 companies would not be able to support on their own. 0:11:06.040 --> 0:11:08.360 And so that's what t SMC does and it's it's 0:11:08.360 --> 0:11:11.439 been kind of amazing, I mean in the sense that 0:11:12.160 --> 0:11:14.280 you know, for the longest time, Intel talked about being 0:11:14.320 --> 0:11:16.560 an i DM as being an advantage because you could 0:11:17.160 --> 0:11:20.160 verily tightly couple both the process and the design together 0:11:20.320 --> 0:11:23.040 and get really good products. And I guess as long 0:11:23.120 --> 0:11:26.000 as as both of those things were on the right trajectory, 0:11:26.080 --> 0:11:28.880 that was true. The problem is now because they're their 0:11:28.960 --> 0:11:31.040 their manufacturing side has followed by the way side, it's 0:11:31.040 --> 0:11:33.679 actually really impacting them. Not only can they not tightly 0:11:33.760 --> 0:11:37.559 couple those two things together anymore, but it interferes with 0:11:37.679 --> 0:11:40.400 the design because you designed for a specific manufacturing process. 0:11:40.520 --> 0:11:42.520 If the process isn't ready, you gotta throw that stuff out. 0:11:43.080 --> 0:11:46.400 And so the fabuless founding model, the founding model, it 0:11:46.400 --> 0:11:50.280 stuff actually seems to be growing. It's ascending now. Um 0:11:50.360 --> 0:11:52.320 it seems to have the advantage of especially because they 0:11:52.320 --> 0:11:54.280 are able to at least at this moment, to stay 0:11:54.320 --> 0:11:58.559 on their manufacturing road map. So we can kind of 0:11:58.760 --> 0:12:03.640 think of Intel in theory as you know, designing chips 0:12:03.880 --> 0:12:08.320 all in video and also having a fab t SMC. 0:12:08.960 --> 0:12:11.920 What you're saying is it sounds like you can't really 0:12:13.000 --> 0:12:15.840 like sort of neatly divide the two and if they're 0:12:15.880 --> 0:12:19.400 having trouble on the fab side, on the foundry side, 0:12:19.679 --> 0:12:23.280 then that also bleeds through to the design side. Yeah, 0:12:23.320 --> 0:12:25.959 because you don't design in isolation, right, you have to 0:12:26.040 --> 0:12:29.559 design for a specific set There's a set of design 0:12:29.679 --> 0:12:33.400 rules that go with a specific manufacturing process. So, for example, 0:12:33.600 --> 0:12:35.839 if I'm Intel, you know, and we'll probably get to this, 0:12:35.920 --> 0:12:39.120 but until it's talking about using outsourcing potentially in a 0:12:39.160 --> 0:12:42.520 bigger way, until cannot just take their current designs and 0:12:42.600 --> 0:12:44.600 just throw them over the wall to t SMC, they 0:12:44.640 --> 0:12:47.959 have to completely redesign them um to to correspond to 0:12:48.040 --> 0:12:51.160 t SMCS manufacturing process, and and and and and t 0:12:51.280 --> 0:12:54.079 SMCS design. So these two things are are coupled. And 0:12:54.160 --> 0:12:55.959 if you have problems with one that causes problems with 0:12:56.000 --> 0:13:00.440 the other. Sure, Wait, could could we do of into 0:13:00.480 --> 0:13:03.319 the manufacturing problems Intel a little bit more than like 0:13:03.600 --> 0:13:07.520 what exactly is happening there and what's gone wrong? Because 0:13:07.520 --> 0:13:09.719 of course, as we mentioned in the intro, you know, 0:13:09.960 --> 0:13:13.360 Intel was supposed to be the state of the art 0:13:13.480 --> 0:13:17.000 global standard for chip making for many many years, and 0:13:17.320 --> 0:13:20.840 now it seems like it's not. Yes, So so first 0:13:20.880 --> 0:13:23.800 I want to step back and and say, um, this 0:13:23.960 --> 0:13:26.920 stuff is very very difficult to do. So the fact 0:13:26.960 --> 0:13:29.280 that people are having problems is not in and of 0:13:29.360 --> 0:13:34.840 itself a shock. These are the most technologically advanced products 0:13:34.880 --> 0:13:37.679 that humanity has ever devised. And I don't know if 0:13:37.920 --> 0:13:39.160 either of you have ever been in in a in 0:13:39.240 --> 0:13:42.679 a in a semiconductor factory, but you can think about this. 0:13:42.800 --> 0:13:44.640 I mean, if if if I just take these nodes 0:13:44.679 --> 0:13:47.319 sizes as as gospel, which I know they're not, but 0:13:47.360 --> 0:13:49.640 I mean, let's just just take them. You know, at 0:13:50.000 --> 0:13:52.960 at Intel's right now is delivering ten anometer products right 0:13:53.400 --> 0:13:56.400 when when they're actually doing the manufacturing and they're imprinting 0:13:56.640 --> 0:13:59.640 these these features onto the onto the wafers. I'll try 0:13:59.679 --> 0:14:01.800 to be plaistic here, but they use use a laser 0:14:02.160 --> 0:14:04.760 light to to do this. The laser right now has 0:14:04.800 --> 0:14:07.559 a wavelength of a hundred and nine three nimes. So 0:14:07.600 --> 0:14:12.079 they're printing purportedly ten nnimeter features using a wavelength of 0:14:12.200 --> 0:14:14.800 light that's almost twenty times the size of the feature 0:14:14.840 --> 0:14:16.559 that they're trying to print. Like, I'm amazed that any 0:14:16.600 --> 0:14:18.880 of this stuff works at all. So just just to 0:14:18.960 --> 0:14:21.040 get that out of the way, like it's it's it's 0:14:21.040 --> 0:14:23.280 astonishing by the This is why one reason that I 0:14:23.440 --> 0:14:26.480 love this space, because it's it's like I'm just continually 0:14:26.560 --> 0:14:29.640 in awe of the things that that that that humanity 0:14:29.720 --> 0:14:32.280 collectively has has been able to pull together here. But 0:14:32.360 --> 0:14:34.880 in terms of specifically what's going on it Intel. So 0:14:35.960 --> 0:14:38.960 I'll step back at fourteen nimes. Like I said, that 0:14:39.080 --> 0:14:42.120 was a minor delay. Um it was. Like I said, 0:14:42.120 --> 0:14:45.120 it took them about an extra year to yield the process. 0:14:45.240 --> 0:14:47.720 Whether when I when I say yields, that's a simple 0:14:47.800 --> 0:14:50.120 concept that is, you know what, I'm making a silicon 0:14:50.160 --> 0:14:52.880 wafer full of chips. How many of those chips are good? 0:14:53.720 --> 0:14:56.280 That percentages your yield. The more the higher yield, the 0:14:56.320 --> 0:14:58.640 lower your costs. So that's a good thing, um. And 0:14:58.720 --> 0:15:01.080 it's the yield that determine how quickly you ran for 0:15:01.200 --> 0:15:04.360 process into production. Fourteen anometers took a little bit longer. 0:15:04.440 --> 0:15:06.840 We still don't know why, but it didn't hurt them 0:15:06.880 --> 0:15:09.160 because back then they had a genuine five plus year 0:15:09.200 --> 0:15:11.080 process leadership. So they burned a little bit of it. 0:15:11.120 --> 0:15:14.000 But it was fine. With ten n we kind of 0:15:14.080 --> 0:15:17.760 know qualitatively what was what was wrong, um until had 0:15:17.960 --> 0:15:20.520 was using multiple patterning on some of their layers for 0:15:20.560 --> 0:15:23.040 the first time. Again, we can talk about what that 0:15:23.320 --> 0:15:25.680 is if if we need to, but they were using 0:15:25.720 --> 0:15:29.800 a specific type of of of advanced process for the 0:15:29.880 --> 0:15:33.960 first time, and it caused problems. They were using um 0:15:34.480 --> 0:15:37.360 new materials at the time, like cobalt and other things, 0:15:37.400 --> 0:15:40.440 and they were using other techniques to get a much 0:15:40.680 --> 0:15:44.880 greater density improvement that was normal with a normal node transition. 0:15:44.960 --> 0:15:48.800 They their their transistor density transitions permillimeter squared of of 0:15:49.040 --> 0:15:51.920 silicon area typically went up by maybe two point two 0:15:51.960 --> 0:15:54.800 to two point four times. With ten anometers, it was 0:15:54.840 --> 0:15:57.360 a two point seven times improvement. So the way the 0:15:57.440 --> 0:15:59.480 company has discussed this is it was just too big 0:15:59.520 --> 0:16:02.440 of a lead and they bit off more than they 0:16:02.440 --> 0:16:05.240 could chew and it caused problems. UM, I have no 0:16:05.400 --> 0:16:07.720 idea what's going on at seven animeters. All I know 0:16:07.920 --> 0:16:11.360 is that it is it is something that's completely different 0:16:11.400 --> 0:16:13.760 from anything they hit them at ten and seven was 0:16:13.800 --> 0:16:15.920 supposed to fix those problems attend it. It was a 0:16:16.040 --> 0:16:19.520 smaller density improvement, it was only one point seven. They're 0:16:19.600 --> 0:16:22.760 using a new lithography technique that's called EUV or Extreme 0:16:22.800 --> 0:16:25.800 multra violet lithography, that was going to replace that multi 0:16:25.960 --> 0:16:28.680 multi patterning and fix those problems. And so they were 0:16:29.200 --> 0:16:32.560 taking steps in order to to to to learn lessons 0:16:32.600 --> 0:16:35.280 from the ten nnimeter debaucle and and and to repair. 0:16:35.320 --> 0:16:38.480 And that that's why the seven nimeter announcment came as 0:16:38.520 --> 0:16:41.400 such a shock, because they've been telling everybody, you know, 0:16:41.480 --> 0:16:43.760 for a year or seven animeters is on track, we've 0:16:43.840 --> 0:16:45.480 learned the lessons, it's gonna be good, and then they 0:16:45.520 --> 0:16:48.280 came out and just dropped the bomb on us. And 0:16:48.600 --> 0:16:51.880 this is why going forward it's so problematic, because their 0:16:51.880 --> 0:16:55.560 credibility on this stuff is now zero. Right, so when 0:16:55.560 --> 0:16:57.480 they're telling us where we think we've got to handle 0:16:57.560 --> 0:16:59.040 on it, we think we know what's wrong. I mean, 0:16:59.080 --> 0:17:17.719 like nobody knows what to believe anymore. So many good questions. 0:17:17.760 --> 0:17:20.040 I think we're gonna have to extend this episode to 0:17:20.080 --> 0:17:23.840 about two or three hours. But in all seriousness, we saw, 0:17:24.040 --> 0:17:27.000 you know, Tracy mentioned it when they dropped this bomb 0:17:27.119 --> 0:17:30.600 this summer. Uh, the stock took quite a reaction. I mean, 0:17:30.640 --> 0:17:32.320 it was a major plunge. I don't think it's like 0:17:32.400 --> 0:17:36.760 a really recovered very much since then. Was that because 0:17:36.960 --> 0:17:40.639 this was understood to be fundamentally different, as you describe it, 0:17:40.840 --> 0:17:45.080 than those other delays, Because there's this nobody really understand, 0:17:45.119 --> 0:17:47.520 like explain the sort of financial market reaction to this 0:17:47.680 --> 0:17:51.560 one versus the previous setbacks. They've bet, yeah, you you bet, 0:17:51.800 --> 0:17:53.080 and I need to full d A m D. I 0:17:53.119 --> 0:17:55.600 think and do this discussion a little bit. Um. So 0:17:55.720 --> 0:17:58.359 if you go look at A m D stock versus intels, 0:17:58.359 --> 0:18:02.320 their polar opposites, right, and and and part of the 0:18:02.359 --> 0:18:04.359 reason there was a A m D is now on 0:18:04.480 --> 0:18:06.680 a trajectory and and they've come a long way am D. 0:18:06.760 --> 0:18:09.280 You know, four five years ago, the controversy on that 0:18:09.320 --> 0:18:11.639 stock was are they going bankrupt or not? And the 0:18:11.720 --> 0:18:14.520 stock bottomed out, you know, in below two dollars. That's 0:18:14.560 --> 0:18:16.119 that's kind of where it was. The stock now was 0:18:16.160 --> 0:18:17.679 you know, close to eight and it's been it's been 0:18:17.720 --> 0:18:20.360 even higher. The reason is a A m D. Actually 0:18:20.520 --> 0:18:23.280 um uh, they had already gone fabulous back in the 0:18:23.320 --> 0:18:27.280 financial crisis. Um they had been using Global Foundries, which 0:18:27.320 --> 0:18:30.159 is as another foundry and Global Founderies back in the 0:18:32.440 --> 0:18:35.240 time from shot am B in the foot um Global 0:18:35.280 --> 0:18:38.480 Founders had problems with their process technology as well. And 0:18:38.560 --> 0:18:40.119 then A m D. I mean the business was in 0:18:40.320 --> 0:18:43.240 was in total free fall. I mean they're there, their 0:18:43.280 --> 0:18:45.040 revenues a back cut in half. They had four or 0:18:45.080 --> 0:18:47.679 five different rounds of restructuring was pretty awful. They had 0:18:47.720 --> 0:18:51.919 a last Hail Mary effort to get a new architecture, 0:18:52.040 --> 0:18:54.560 which they're delivering now. They're on their third generation UM 0:18:55.080 --> 0:18:58.600 and that architecture was extremely successful, and they made the 0:18:58.640 --> 0:19:01.439 shift from Global Founderies to t SMC. And so now 0:19:01.560 --> 0:19:03.960 you're in a situation where where am D actually has 0:19:04.480 --> 0:19:07.520 a competitive and in some cases even superior architecture, and 0:19:07.600 --> 0:19:09.919 they have a superior process technology. And so they've been 0:19:09.960 --> 0:19:13.399 taking share from Intel both on the client side pecs 0:19:13.400 --> 0:19:16.720 as well as in servers. Right now, the real bulls 0:19:16.760 --> 0:19:19.399 on am D. They have this vision right that A 0:19:19.520 --> 0:19:21.600 and D is gonna go And I'll make up the numbers, 0:19:21.600 --> 0:19:25.200 but I mean twenty two there were twenty three. They'll 0:19:25.200 --> 0:19:27.560 be doing two fifty or three bucks and earnings, and 0:19:27.560 --> 0:19:30.399 by five they'll be doing five or six bucks and earnings. 0:19:30.840 --> 0:19:33.040 And and this is a duopoli market. It's them an 0:19:33.080 --> 0:19:35.719 Intel primarily, so all that would be coming at Intel's expense. 0:19:36.359 --> 0:19:39.480 But am D has been kind of a dream. And 0:19:39.560 --> 0:19:42.399 so before Intel announced these seven animeter issues, it was 0:19:42.520 --> 0:19:45.399 kind of like, well, okay, maybe am you'll get to 0:19:45.480 --> 0:19:47.240 this two or three dollar number. In a few years. 0:19:47.280 --> 0:19:49.600 But by then a Intel will be going harder and 0:19:49.720 --> 0:19:52.200 seven they'll have fix the issues. At ten they gave 0:19:52.280 --> 0:19:54.880 am D an advantage. And at that point, like it's 0:19:54.880 --> 0:19:56.920 all over right, and and Intel is going to be 0:19:57.000 --> 0:20:00.880 back on on on the ascendancy now because of these 0:20:00.920 --> 0:20:03.040 issues with seven animeters, like with A and DNA, you 0:20:03.200 --> 0:20:05.320 you could believe whatever you want, like there's there's no 0:20:05.440 --> 0:20:08.440 pushback because you have no idea what Intel is gonna 0:20:08.440 --> 0:20:10.240 be doing at that point um, and it makes the 0:20:10.359 --> 0:20:14.160 idea of the the x C E six CPU market 0:20:14.200 --> 0:20:16.000 and PCs and service, which today is sort of like 0:20:16.080 --> 0:20:18.160 a nine ten or an eight twenty kind of market. 0:20:19.000 --> 0:20:21.040 You know, if you wanted to believe that it could 0:20:21.080 --> 0:20:24.160 be a true duopoly, a fifty fifth or sixty market 0:20:24.200 --> 0:20:25.840 in fears, if you want to believe that, you can 0:20:26.760 --> 0:20:29.520 and that all comes if it happens at Intel's expense, 0:20:29.680 --> 0:20:33.359 and the driving force behind it is a continued loss 0:20:33.400 --> 0:20:36.320 potentially of competitiveness on Intel's part. And with the seven 0:20:36.400 --> 0:20:39.000 animeter delay, which like I said, was was really a 0:20:39.040 --> 0:20:41.760 bomb that was dropped on people, it becomes very easy 0:20:41.840 --> 0:20:43.399 if you want to believe that sort of thing, to 0:20:43.480 --> 0:20:46.960 believe it, So this was potentially a thesis changing announcement 0:20:47.359 --> 0:20:49.280 for Intel. That's where the stock reacted like it did. 0:20:50.480 --> 0:20:52.960 So is this just I don't know? Is all of 0:20:53.040 --> 0:20:56.960 this just down to Intel doing basically a chip version 0:20:57.240 --> 0:21:02.840 of vertical integration for far too long when other companies 0:21:03.480 --> 0:21:07.720 decided to specialize in various ways. Yeah, so again, I 0:21:07.760 --> 0:21:09.800 don't know why Intel's having this, but it's kind of 0:21:09.840 --> 0:21:11.399 amazing that they are, because they looked. Intel is a 0:21:11.520 --> 0:21:15.639 hundred and thirteen thousand employees. They spend over thirteen billion 0:21:15.720 --> 0:21:18.280 dollars a year in revenue, which which is like almost 0:21:18.320 --> 0:21:21.399 double amb's revenue. And in R and D they Intel 0:21:21.400 --> 0:21:24.239