Manifold

Steve and Jon discuss the global semiconductor industry with an emphasis on U.S.-China technology competition.

Show Notes

Jon Y produces Asianometry, which focuses on Asia technology, finance, and history: Podcast, YouTube channel, and Substack.

Steve and Jon discuss the global semiconductor industry with an emphasis on U.S.-China technology competition.
 
Topics discussed:
  • Jon's background and his move to Taipei.
  • Key components of the semiconductor ecosystem: fabs, lithography, chip design.
  • US-China tech war
  • TSMC, ASML, Huawei
  • Taiwan politics: Green and Blue parties, independence
  • PRC invasion / blockade of Taiwan?
Music used with permission from Blade Runner Blues Livestream improvisation by State Azure.

--
Steve Hsu is Professor of Theoretical Physics and of Computational Mathematics, Science, and Engineering at Michigan State University. Previously, he was Senior Vice President for Research and Innovation at MSU and Director of the Institute of Theoretical Science at the University of Oregon. Hsu is a startup founder (SafeWeb, Genomic Prediction, Othram) and advisor to venture capital and other investment firms. He was educated at Caltech and Berkeley, was a Harvard Junior Fellow, and has held faculty positions at Yale, the University of Oregon, and MSU.

Please send any questions or suggestions to manifold1podcast@gmail.com or Steve on Twitter @hsu_steve.

Creators & Guests

Host
Stephen Hsu
Steve Hsu is Professor of Theoretical Physics and of Computational Mathematics, Science, and Engineering at Michigan State University.

What is Manifold?

Steve Hsu is Professor of Theoretical Physics and Computational Mathematics, Science, and Engineering at Michigan State University. Join him for wide-ranging conversations with leading writers, scientists, technologists, academics, entrepreneurs, investors, and more.

Steve Hsu: Welcome to Manifold.

Today, we are going to discuss semiconductor technology and U.S.-China competition in high technology. My guest is Jon Y who lives and works in Taiwan. He has a Substack and a YouTube channel called Asianometry. I discovered Asianometry some time ago because I was trying to track the details of a U.S.-China tech competition. I was trying to track some of the details that I think we will get into in this discussion. And I found his work extremely useful in understanding things like the deep history of ASML, the competitive situation between TSMC and Samsung, and Intel. All kinds of questions that I think strategists, or even people who just want to understand how technology is going to evolve in this era of competition between the U.S. And China is interested in it.

So it's a real treat to have Jon on the podcast. Hello, Jon, how are you doing?

Jon Y: Doing well, doing well. How about you?

Steve Hsu: Awesome. So I always like to go into a little bit of background on each of my guests, just because I think the audience likes to kind of understand people's life history. And also my observation is that great minds don't just spring, fully formed onto the planet. They are somehow formed through their life experiences.

And it's always interesting to get to know what people have done in their lives. So maybe tell me a little bit about your background. I'm guessing you are of ethnic Chinese extraction, but maybe you grew up in the United States. Is that correct?

Jon Y: Yeah. great minds. Wow, that's... now I feel very pressured. I had a pretty, I had a pretty ordinary upbringing, down in southern California. After graduating from college in the Bay Area at Berkeley, I went to work in Silicon Valley. First in finance, I wrote research reports and valuation reports, and I did that for about three years. After that, I kinda got tired of working so long hours.

So I then kind of jumped around Silicon Valley for a while, tried to kind of live that tech, startup life, working in kind of, how it is with startups. You kind of do all sorts of different things. I wrote code, did data, did marketing, did a lot of things like, writing content, stuff like that.

And really burned out after that last start-up in which I made bras, of all things. And from there I decided to bum around Silicon Valley for about eight or nine months and decided I was going to go to Asia. I sent out applications to companies in four different cities, Singapore, Shanghai, Taipei, and Hong Kong.

I got offers from all of them. And the only one that gave the best offer was Taipei. So I literally got that offer, thought about it for two weeks, accepted, and packed all my stuff. Put the rest in storage and flew out with a bag. And didn't really expect I'd be here for so long. This was about five years ago. now I'm still here. It's kind of crazy.

Steve Hsu: Did you have family and friends in Taipei?

Jon Y: No. Half of my family is from Taiwan, but they never were in Taipei. They all moved to America just about when I went over to Taiwan. It was kind of like a funny reversal. My family on the other side is from Hong Kong. So when I touched down here I didn't really have any friends, knew kind of one person all right, but everyone else, it was just like a clean slate. It was really exciting.

Steve Hsu: Wow. I almost want to digress and just ask you what it was like to kind of break into the social circles in Taipei without any network to start with, that it's probably a story all by itself.

Jon Y: Yeah. No, like that's real fun. Like you, you find out that there are really tight circles here and, there's a lot of stuff going on underneath the surface. Yeah, I think the harder part is trying to break into the Taiwanese sector. Like circles, like the ABCs, ABTS out here are kind of the really close you'll actually know if you know, one, get plugged in to all the others.

Right. So you'll start knowing all the, one name and then hear all the other names and then like, find out this name is famous and that same, same as, and it's really funny, but, unfortunately, that's, that's not so easy with the Taiwanese people.

Steve Hsu: You know what I remind me, I got to hook you up with my cousin, Richard Chang. now you're too young to remember Richard Chang. I'm sorry. I have to digress. Richard Chang is six foot seven. He is, as far as I know, the only. ABC American-born Chinese to receive a division one college basketball scholarship.

Jeremy Lin didn't get one. Richard Chang was a power forward for Cal in the eighties. He played at Huntington, I think. Huntington Beach High School, down in SoCal. he was a big star on the Cal campus. When I came there as a graduate student, everyone asked me, they couldn't believe it when they heard like, what's your cousin because he was so famous, you just walk around campus and he just stood out, six-foot, seven Chinese guy.

So anyway, he is now the Asia manager. I think he's president of Costco Asia, and he lives in Taipei. So he's, he's older than you, but I should definitely introduce you guys at some points.

Jon Y: Oh yeah. I'd love to say hello

Yeah, so Asianometry started out as a YouTube channel for hiking videos. So it was kind of just like the blog of my life and Taiwan, kind of one of those things that you kind of see around nowadays. But I kind of moved towards kind of long-form content because I wanted to make videos that I wanted to watch.

Especially in content that are in areas that I felt that I didn't really see all that much in kind of, the wider YouTube space. And at that time it was kind of, you got a lot of like content about history in America or history in Europe and stuff like that, but nothing really in Asia. So that's changed a lot now. Like a lot of people have gotten more interested in certain events and stuff like that. But I worked on that for kind of three years and got no traction whatsoever.

But over time kind of, I started noticing a lot of it, it kind of started with TSMC cause TSMC is a company that has a lot of traction here.

Like it's very deeply embedded in the local culture. It's even though that, even back then where nobody really heard of TSMC, the company is very popular here. Extremely popular. You will hear it almost spoken. If you sit in the cafes across Taipei, all the Taiwan city, you'll hear people talk about this company.

And eventually got to the point where I started to kind of be religious. Like why do I keep hearing this term over and over again? And that's kinda why I wanted to share that with kind of the channel and with everyone who kind of watched my very small audience at the time. And it's been growing ever since, kind of crazy.

Steve Hsu: Yeah, so at least my gateway into your channel was mainly tech. But I did notice lots of, for example, videos on history, Asian history, things like this. So yeah, I wondered what the history of it was.

Now, as I recall, as you just said, you went to Taiwan originally because you had a job offer, but now it seems like, are you doing the channel full-time now? Or do you still have a day job?

Jon Y: Yeah, I still have a day job because I'm here on a visa. So basically, I work on the channel after work and before work. So I get up, like 5 or 6 AM, work until 8 or 9, and then head into the office, work until work a normal eight hour day, and then come home and until 10.

Jon Y: And then on weekends, right now, I have a little co-working space and I just work all day there.

Steve Hsu: Wow. Well, I appreciate you taking some time to chat with me. I hope I can steer some viewers or readers to your content.

Jon Y: No, no, I, I love to listen to your podcast and, I, I don't, it feels weird to be featured alongside like PhDs and professors and really, really smart people. And that, you know, it's just me. So, you probably have to edit all of my dumb history life. But you know, I'm excited.

Steve Hsu: No, I love it. I think that's the colorful stuff that I think helps listeners to understand you better.

I want to say, I want to compliment you because, the areas that you're exploring, like, okay, what's the deep history of ASML or what's the history of TSMC, are topics that are underexplored. Where else could I get similar content?

Let me, let me just say this. When I found your content, the other place where I would find kind of related content would be okay, maybe some tech journalism, but that wouldn't necessarily be particularly deep or insightful. It could be, but it isn't necessarily. The other thing would be investment analysts. But the investment analysts were always looking at very specific things like, okay, what's next quarter going to look like, et cetera, et cetera. Almost nobody would look at, well, the big break was when they finally were able to license this key patent from Hitachi or something like that. That's the kind of thing that you're interested in. You're interested in the actual history of these companies and, you know, in the modern world, these companies are, as you were just saying, in a sense of the Taiwan economy is TSMC, right?

And so it's fully deserving of a whole team of academic scholars, trying to understand what the hell happened with TSMC and Foxconn and all these companies, which helped, drive the history of Asia, broadly speaking. So I think it's deserving of serious attention. You're one of the few people who's actually giving it that attention.

Jon Y: Yeah. I came from an investment background, so I found myself reading, how hedge fund managers love to say, they'd like to read the 10Ks, annual reports. They'll read all day and stuff like that. The fact is, I found that I wasn't interested in that.

I wasn't interested in the valuation part. I wasn't interested in talking about multiples and all that. I was interested in when they do something called an initiation report. So when an investment bank starts coverage on a particular company, they'll write this long report about the history of the company. And I found that part really interesting. I wasn't interested in the valuation part. I wasn't. I didn't really care about the stock price because the stock price, none of them really hit the targets anyways, or like targets are way off.

So like, that part of the story was something I was really interested in. I thought to myself, how about a 20-minute video about that. And that's kind of how a lot of this stuff's got started. So, it's really interesting to talk about these critical points in Asian companies history, when you learn and dive deep, dive into all these different companies, you learn a lot about that. And you learn that some of these stories have a lot of familiar echoes and some, you know, don't. And, it's really interesting to find out what were these key points for certain companies. And then, in the case of TSMC, for example. Like I almost feel like that founding story is kind of old now. Like everyone talks about it, about how stronger and had these certain ideas and stuff like that. But you really want to dive deep into kind of what happened after that. There are certain ups and downs throughout that. And you learn a lot more, I think, I feel for academics, industrial professionals, and investment professionals as well, whether they have kind of been packaged up into this nice ribbon story. And that was really important to me to get those details and get them right. And mix that in with the technical part. Cause he can't really talk about technology without being technical, in my opinion.

Steve Hsu: Yeah. Sometimes when I'm listening to one of your segments, I'm thinking, how does he know that? Like where did you get that information? That's really interesting. I feel that someone like you, who has both on the investment advisory side of things or analysis side of things and tried to start a company as a founder or its startup guy, you really get the point that every company has its own unique history and its own unique inflection points and personalities, and it's every, every company is a kind of interesting story, potentially, if you look into it.

Jon Y: Yeah. Yeah. And, you know, none of them are the same. I have some that are more interesting than others. Like, there are a couple of Taiwanese companies I've done videos on that are billion-dollar companies nobody's heard of. And that's because they've had the most boring founding story ever.

But like, I'm sure it will be interesting to some guy who invests in them, but like, the learning that part is really special. Yeah.

Steve Hsu: Well, the companies that we're going to talk about have huge impacts on global supply chains and potentially even geopolitics, as well.

Yes.

Now I had to choose specifically the semiconductor industry because you actually cover lots of things. I chose the semiconductor industry because I think that's one of the main areas of U.S.-China competition.

Oh, I, let me say this. I have two goals for my listeners, from this podcast. One is to get some introduction to the complexity of the semiconductor industry, because most people, oh I bought a laptop. Oh, is it fast? Yeah, it seems fast. Your car has a lot of chips in it, I guess. I don't really see them. But I have an LCD panel on my dashboard now.

Steve Hsu: So people just to them, it's just part of the fabric of their lives. They don't really understand the incredible R&D and CapEx that went into building this highly developed IT world that we live in. So I would just like to give people a sense of how rich it is and how the different sub-sectors of the semiconductor ecosystem interact and things like this.

And then secondly, the thing I'd really like to get into is, to try to get your views on what is going to happen, in terms of China trying to catch up, trying to become maybe fully independent in its semiconductor capabilities, what will happen to TSMC and Samsung and other companies, ASML, and like that, as that evolves.

And maybe that last set of questions can't really fully be answered, but I'd love to just get your perspectives on that.

Jon Y: Sure.

Steve Hsu: So let me just start by introducing myself to my audience. So everybody knows what Moore's law is, right? It's the idea that every two years or so maybe there's a doubling of the density of transistors on a chip and Moore's law has, we could spend a whole episode talking about how Moore's law has kind of broken down recently that maybe it's not in effect anymore, but it was in effect for about 40 years.

And I believe Moore proposed it in 1965. So, it's incredible what physicists would call phenomenological prediction. There's no deep theoretical reason why it had to be true. And since it's an emergent property of a really complex industry, it's kind of amazing that this general rule applied. My own interaction with Moore's law comes from being in and around physics departments my whole life.

And even though I'm not in the area of physics that works on material science or semiconductor fabrication, none of that. I have been close to people, other professors, grad students, who worked in labs doing research on things related to semiconductors. So I've sat through a lot of talks on the core technology and even on sort of meta, talks about like, okay, what is the industry? Moore's law roadmap, and how the hell are we going to get past this. And this looks like a real bottleneck. And so I even sat through that kind of presentation, in physics departments before. So I have some insight into this.

And the thing I'll say to people that I think is the number one thing that they should remember is, anybody who says that we've had a tech stagnation or nothing's happened since the Apollo moon mission or some complete bullshit like that, does not realize that we got a factor of a million in computing power, mainly due to the efforts of huge armies of physicists, applied physicists, material scientists, and electrical engineers and chemists, which all kind of was under the radar. But the whole world was transformed by the work of these people and billions and billions of dollars of CapEx and fortunes made and fortunes lost.

So, it's kind of this underlying driver of all the change that's happened much of the change that's happened in a technological society in the last 40 years.

So having said that, Jon, let me say that, obviously, the ecosystem is super complicated, we can't explore all parts of it. But I kind of want to introduce the listeners to three different aspects of the ecosystem. Three different parts of it.

One is fabs. And an example of that is TSMC. Maybe their biggest competitor right now is Samsung. Previously Intel was one of the big competitors in the space, but it seems to be fading. So fabs is one.

Second is lithography, the actual etching of the circuit patterns onto the chip. We mentioned the company ASML, which is in a very unique market position right now and makes a uniquely complex machine.

Each of their leading-edge machines sells for like a hundred or $200 million. The PRC is desperately trying to catch up in that technology. So area number two is lithography.

And then area number three is chip design, which is a little more of a purely cognitive activity where chip designers use very specialized software to design chips, but then the chip is actually constructed at the fab.

So I'd like to just have a conversation with you about those three different parts of the ecosystem. And they're still, of course, much more. There are the people that make the wafers. There's quality testing, there are all kinds of other, still highly technical aspects of the industry. We have to draw the line somewhere.

So let's talk about fabs where the physical chips are actually made. And maybe you can just give us a few sentences introducing, what is the general lay of the land right now in the world.

Jon Y: I think like, when you talking about fabs, it's a fancy word for like factories, right? Right now you have different types of factories out there. And with the fabs you have, you wanted to talk about TSMC and Samsung. They are the two big, bleeding-edge foundries out there.

Intel kind of runs a different business model though, so they're trying to change that. Fab is basically a very large factory where on the inside you have this massive clean room in which you install the various parts of the equipment that will do the various parts of your wafer process. Right?

So when you create a wafer that's actually done in kind of something called mask layers, where you would reproduce a pattern onto a substrate, usually like the wafer itself. Each mask layer kind of requires like steps 15 to 20 steps. And some of the steps, I think you'll be very familiar with like photolithography, but there's also deposition, diffusion, oxidation, stuff like that.

So the fab is the place where that happens. And so when you have multiple mask layers, for example, and three, your three-nanometer process node has something like 80 mask layers. And with 15 to 20 steps per mask layer, that's over a thousand steps. You have to repeat these steps over the span until basically the wafer is done. And you have to get extremely good yield on each of those steps throughout all those processes. And not just like, oh, 80%, 90%, Because you have to deliver really good yields to the customer in order to maintain them as a customer. Because, if you are, if your fabulous customer receives a wafer and like, three of the dyes are good out of the whole wafer, that's not very good. And you're going to drop that fab, that foundry.

So fabs are not only a technical beast, they're an economic one, too. A lot of people kind of really focus on the magic of technology. UV and stuff like that. What I think is much harder to kind of bring to other people to understand it's like, this is an economic system. These are highly optimized systems that are at the intersection of technology.

This is a really expensive of paying for that and making sure that actually makes a profit for the of fabs and really key to that is scale to like, which is why a lot of foundries, other than TSMC, and maybe Samsung, they're not making a, like a substantial profit in the industry because they don't have enough scale. They're not learning enough. They're not putting enough wafers out there to make up for the immense capital costs that you need to put in to make these make sense.

Something I really enjoy about the semiconductor industry is that it's not only like leading-edge technology, but it's also a story of economics and a story of business as well.

I think that's that's, that's the special part about it. Yeah.

Steve Hsu: Let me jump in and elaborate on a few of those things. So first of all, TSMC, and correct me cause I'm sure I'm going to get some of these things wrong because I'm not an expert. TSMC is an order taker. So there's no actual TSMC chip. They might make a chip for Huawei where they used to be able to do that but they can't anymore.

They might make a chip for MediaTek or some other chip firm, but they are an order taker. They're an outsource manufacturer.

Jon Y: That is correct.

Steve Hsu: Whereas both Samsung and Intel have their own chips. Like if you buy an Intel Pentium processor, you buy a Samsung, I guess Exynos chip in your Samsung phone.

And I believe Samsung though also has some foundry activity where they'll take external orders and make chips for third party customers. Is that true?

Jon Y: That is correct. I think they make chips for Qualcomm and Nvidia as well. And, as well as some other customers, which we might not hear of, but, but yeah. They also used to fab Apple chips before Apple moved to TSMC.

Steve Hsu: That's right. So, so, so we've got a mix here where one player TSMC is always just working for somebody else, whose name actually goes on the chips. Samsung is a mix. And historically, at least as I understand Intel basically just made its own Intel products

Jon Y: Yeah. Which is why kind of Intel starting this Foundry 2.0 process, is sort of an interesting sort of world for them because they've tried it before, didn't quite work for various reasons. And prior to that, Intel and TSMC actually worked quite closely together. I think Intel was one of TSMC's major first customers. Actually, TSMC wanted Intel to be an investor. So sort of an interesting sort of situation for them.

Steve Hsu: But as far as the geopolitical story goes, Intel is no longer at the frontier, right? They fell out at, I don't know, it was a seven or five-nanometer process.

Jon Y: I would probably say that they're still there. They're still there. I mean, when you get to this frontier, it's not really like a line, it's not like… The word edge I think is a misnomer. You're more like, it's more like a cloud and everyone's sorta in that cloud. There are certain companies that have decided not to be at that cloud part, but Intel is still there.

I mean, like, if you're still committed to trying to get to that sort of boundary, I think you're still there. There are companies that explicitly decided they're not going to be.

Steve Hsu: Understood. This gets to your economics point that, you know, if TSMC has it really dialed in, they might be at that say three, five nanometers scale, and they're making money. Whereas Intel wants to be there and maybe they have the core technical competence that they think they can get there, but at the moment they're bleeding money and they're going to have to keep bleeding money till their efficiency gets better.

I mean, that, that could be the story, right? Something like that. So, when I was a professor at the University of Oregon, Intel had a, I think some, I forgot what they had. They had some fabs and designs in Oregon. And then also, this Korean company called Hynix also had some memory fabs in Oregon. So we had a master's program at the University of Oregon where you physics students could get a master's in semiconductor physics or semiconductor applied physics. And then they would immediately go to work for Intel and Hynix and maybe some other places.

And one of the things, again related to a comment you made, is that I noticed in designing the curriculum for these guys, it wasn't just that they had to learn some physics, some stuff about semiconductors. There was a lot of statistical education for them, because if you're running some process and you're trying to figure out, hey, this is not working. Our success rate is too low. I have these 10 different parameters I can tweak, it's an incredibly complex manufacturing process. They then have to be able to figure out like what, okay, what is the evidence telling me about the way I need to adjust the process in order to make it work effectively, the plant, the fab and Malaysia has this working. Why isn't it working here?

So it's a very, this sounds like a cliche, but it is a very high IQ sort of G-loaded business to be a fab because you need everybody up and down the line to be pretty smart, solve problems, figure things out, get things done efficiently. And so it's not an easy business to break into.

Like, if you're an investor you might ask, how can TSMC command such high valuations? What is it that they actually have that makes them so valuable as a company? And I think a big part of that is the experience in human capital just, running these things so that they're actually economically efficient and viable.

Jon Y: Yeah, that's part of the story. I think like, to have the human capital, to have the knowledge. Because human capital is very important because I think you should think of these fabs kind of like restaurants. I think I've made this kind of metaphor before. For example, a company like ASML provides ovens, capital equipment for cooking, but TSMC, Samsung, and, I guess Intel, they're cooks.

They're using that equipment. They're making recipes. They're learning their best recipes to create amazing food for people to see. And like a restaurant, like a fab depends on throughput. Depends on yield to some extent, right? How much dough you're using to make a finished piece of bread?

That sort of idea. And if you're not doing that right, your fab loses money and your fab closes down. It's very important.

Steve Hsu: Yeah, I think that's a great analogy. So the fab is the restaurant where the activity happens and they actually produce the physical product that the customer wants, the lithography, a machine builder, which we'll get to, they make some really super expensive piece of equipment that has to be used in the kitchen, like some super oven or something.

And, so they're a provider of the kind of infrastructure that the fab or the restaurant needs in the kitchen. And the chip designers are in a way, the ones who make the recipe up, right? They're like, this is, this is how you make a souffle and you need these ingredients and you have to combine them in this way.

Steve Hsu: So maybe that's, that is a nice way of explaining, these three components of the industry. I'd like to turn to lithography for a second. So there's this very famous company. If you're an investor, at least, or if you're a technologist it's called ASML. It's a Dutch company and they are the sole provider, correct me if I'm wrong, of extreme ultraviolet lithography machines, which, which you need to get to the, I believe seven-nanometer and below process. Is that roughly right?

Yeah. Yeah. I think seven nanometers is You can reach it without UV, but at this point, it's commercially viable. It's not economical. Right. You could do it with DUV, but it's not competitive.

Jon Y: Correct. competing against EUV itself.

Steve Hsu: Right. So now again, for the listeners, these very small, feature sizes, they're not actually the feature sizes, they're actually kind of fake, but the name for the process level say seven nanometers, five nanometers, even three nanometers. That's very bleeding edge stuff that might go into a kind of flagship cell phone or something like this.

Whereas the chip, the bulk of chips, they're used in your car, in your refrigerator, even internet of things devices. Those are made on much coarser processes, like maybe 28 nanometers. And they're the bulk of, in numerical terms, chips made. But the, but the, but the big profit margins are going to be at these extreme, extremely fine feature scales where you need EV.

So ASML, their machine, which I mentioned earlier, I think it could sell for something like a hundred or $200 million a shot and they sell maybe I think, 50 machines a year.Is that about right?

Jon Y: Yeah. Yeah.

Steve Hsu: These things, just to describe the extreme physics, okay. You're trying to make light in the extreme ultraviolet. So very blue, far bluer than the human eye can see short wavelengths.

Jon Y: They're basically X-rays. It's like low-power X-rays.

Steve Hsu: Yeah. Exactly. Low, soft x-rays. need that. Basically, the size of the wavelength of the light determines the smallest feature that you can reliably etch on the substrate, on the chip itself.

And the process, and this is amazing, the research that went into this. I happen to know a little bit about the history of this cause the U.S. Department of Energy, which is where I got a lot of research funding from DOE to do high energy physics. The Department of Energy was actually involved in a big consortium to develop this technology. So they take tin, the metal tin, they take 10 molecules and they superheat them with a laser until it's a plasma. And there are certain transitions among the outer shell electrons of tin, which are in the frequency range to generate EUV light. And so the process is unbelievable. Like you, you have a device where you're literally, you're, you're, you're not just melting 10, you're ionizing, a metal tin, a droplet of tin. You're capturing the UV light that comes from that plasma. And then you're using some crazy set of super precise Zeiss, that's a German company, mirrors to focus that or to point it at your mask. And that's the way that you make these cutting-edge chip designs. And you can just imagine it. It's one of the most complicated machines that exist on the planet.

Right? I mean, I mean, it's set of technologies that go into making one of these ASML machines work is, is just incredible. And it took, it was like a 20-year research project. You know, when it was first proposed, it was one of those moonshot projects on a Moore's law roadmap. And I could imagine like some wiseass theorists like me saying, you're never going to get that to work. How the hell can you possibly get to the work that. That's going to use as much energy as a small city, right. Just to, just to make that tin plasma, you know.

So all kinds of difficulties like that and who knows, maybe they started thinking about being able to do this, like in the nineties.

Jon Y: Hmm.

Steve Hsu: And only now it's become feasible. So, so I just back to my earlier comments, I think people just totally misunderstand and underestimate the difficulty it takes to keep a Moore's law going or to keep technological progress going because if you're not involved in it, you're not aware of it.

Jon Y: Yeah. And I think like the key point with like the Evie was that you wanted to continue for the first, I would think 40 years of kind of how, the roadmap of the was driving the concept that like you could X smaller feature sizes, and you first you use kind of as a very simple mercury lamp, and then you progress to eyeline, which is another type of light and then lasers, right?

Your krypton lasers your argon lasers. EUV was supposed to be the next sort of step towards that. Like, like you were supposed to go from like, argon fluoride to EUV. However, what happened was that they couldn't get enough power out. Right. So that's what you're talking about with the, with the amount of light. They weren't getting a good enough ratio for the light that they're putting in. The physics was always there. They always knew it could be used, but the problem was that they couldn't make it work in a way that was economical, like commercially viable. Again, there was that, that fat, the point about the commercial part, right? You could conceivably use this machine, but if you're not getting your money's worth, if you're not getting a strong enough light, then it's not commercially viable against the alternative, your DUV alternative.

Right.

Steve Hsu: Yup. That’s exactly right. I mean, in these old, these lectures that I went to in the past, where they would talk about the Moore's law roadmap, they would always have like three different technologies that, okay, one of these we're going to get working and that's how we're going to stay on this curve. And the other two, you know, after, many, many teams of physicists at universities and labs, national labs and companies would be trying all these things one of them might work and that's how the field kept going. Every time it was a miracle.

And I remember when I first heard them say, yeah, we're going to be using X-rays. What? You're going to be using X-rays to make microchips, you know. And if you just do some back-of-the-envelope calculation of efficiency, you're like, yeah this is going to cost as much energy as like a small city to run your fab. Right. But they're able to do it. They keep doing it.

Now, ASML has basically a kind of monopoly right now, right? For all the fabs. I mean, they have a limited number of customers because they have to be people that want to manufacture really bleeding edge chips. But they kind of have a stranglehold.

Just say a little bit about what the situation is about PRC companies, Chinese companies on the mainland, trying to get their hands on ASML machines.

Jon Y: I think ASML has a monopoly on working UV machines. There is another company called Gigaphoton out in Japan. I think trying something new using magnetic fields to generate EUV light, which is kind of fun. And it kind of reflects Japan's past legacy of semiconductor suppliers.

So there has been kind of the big kerfuffle where SMIC is a company, they're China's leading foundry, and they tried to acquire UV so that they could kind of move from 10-nanometer processes to seven-nanometer processes economically. And basically, that was blocked, right?

I think there's been a lot of, kind of discussion about like, well, ASML is a Dutch company, the Netherlands, like how can America kind of have that sort of involvement, but like you said before, America does have a very strong stake in EUV technology. They're one of the first to kind of bring that and move that forward. They invested in it there. So they had a say in how that technology is being exported out there. I think there are competitors to lithography to try to kind of catch up in that space. There is, there is, SMEE is the kind of their leader, their commercial state bank leader.

Jon Y: I would probably say that so far as we understand, it's like these products they technically work. Science works, but it's not commercially viable in the market. So unless something horrendous happens where like they're not allowed to use ASML, DUV products or something, or even like prior older generation DUV products from ASML and, Canon and Nikon, cause Canon and Nikon are still present in the DUV space, that SMEE doesn't seem to be able to compete in that market.

But we, we don't really know. I mean, they make a big deal of the PR or the public relations of a certain achievement. They're really good at that. And like, I think something a lot of people need to think about when they look at a headline is that there has to be, you need to see the commercial part of that workout, before you're able to, yeah, go ahead.

Steve Hsu: Absolutely. Sorry to interrupt that. But yeah, the devil's in the details because I might be able to demonstrate in a lab. Oh yeah. I'm generating photons at this frequency, and everything's fine. But how reliable is that thing? And what's uptime, Properly integrated in the actual process, that's producing chips, you know, at the right price point that the customers will buy this stuff and there's a good margin for me.

So you're right. Like it's just one step of the puzzle. Like the mastery of basic physics is just the first step and there's so many additional steps. So I think one of the things people are keenly watching is how long will it take the Chinese to catch up across this whole gamut.

Now, the government there is pouring tremendous amounts of money into trying to catch up. Sometimes they have spectacular failures where some local enterprise eats up billions of dollars and it turns out to have basically done nothing and been a giant failure. On the other hand, you know, at the academic level and in terms of their government laboratories, they have a lot of really strong scientists, especially in these areas of chemistry and physics and laser physics, all of these areas they're quite strong. And they've been aggressively recruiting engineers from Korea, from Taiwan, from Japan, for some time. So it's a very dynamic situation. And what would You say is the best source of information about how all of this is developing?

Jon Y: You got to actually be, you got to talk to someone in there. That's what I found. Semiconductors in general are like a very big market and it's a market that needs to be really tightly integrated. Like people need to be working with one another. And you can make guesses based on certain understandings of how different companies work. You can make predictions but like to see the real status of anything that goes on in the mainland, this is somewhat difficult because I would presume that a lot of these companies are seen as state secrets.

And I think one thing you'll see big signs and big trends that I think kind of aligned with developmental trends in previous industries. For example, you'll see the adoption of native technologies and digitally developed standards, stuff like that. You'll see those as kind of main points. And those you'll hear very easily.

I think if you try to kind of follow it on a day-to-day basis, trying to see this thing happening or that thing happening, you're never going to, you're gonna get kinda lost in the details. Lost in the weeds. That's my opinion on that.

Steve Hsu: How about this angle? I hear all these stories about — now, this isn't necessarily directly lithography, but say engineers with lots of fab experience at TSMC or something. You hear all these stories about them getting recruited with huge packages to go to China.

Jon Y: Yeah. Yeah.

Steve Hsu: I'm imagining there's a lot of scuttlebutt in Taiwan about that kind of thing.

I don't see a situation where, unless someone feels like they're unhappy with the company itself, that someone's going to outbid TSMC on a salary basis. And I think you have to understand that TSMC employs 50,000 people, and some of these people are very important, right? I would say like your R&D engineers are very your R&D, your R&D your VPS of R&D and stuff like that will get paid very well. Like there are certain individuals, you know, can actually change the way a company's future fortunes are.

The current SMIC co-CEO, I think like he's an ex TSMC guy. He was one of their stars. And then he moved to Samsung and he helped Samsung catch up to TSMC. But for the most part, I would probably say that hiring hardware engineers is like hiring a Google engineer. It kind of, you're getting a name. And what are you actually really getting is has to be judged on an individual basis. In terms of technology, when you talk about trends of technology transfer from Taiwan to the mainland, it's, it's not so much of a big deal unless you see, Chinese companies start doing something and extending on that technology themselves.

Jon Y: Does that make sense? It's kind of like one of the key points was that when Taiwan TSMC was founded, they acquired a kind of an older node processing node to work with. And had they just stuck to that, that, that would have gotten them nowhere In the global marketplace, because always they're going to get an older version and like, they need to do something more with that.

they improved it on themselves with their own human capital, their own knowledge to, to, to eventually catch up to the leading edge and then they to expand beyond that, you need that sort of progress and you need that sort of, you need to see that stuff happen before you can take, technology, independence, and technology, certain industrial policies more seriously.

Not to say that they're not taking that the mainland is not taking it seriously, but I do sense that like it's when you have technology decisions driven by domestic politics, there is some concern there.

Steve Hsu: Yeah. It's, it's not clear how it's going to shake out. Now for these workhorse, feature sizes or process levels, like say 28 nanometers. Where is the mainland right now? At what level are they able to produce chips of that sort?

Jon Y: I think SMIC is doing a pretty good job. So the two leading foundries in China are SMIC Grace Semiconductor, or I think they've changed their name now, but like two are out in Shanghai and they were founded by people from Taiwan and, they're actually doing actually SMIC's founder was Richard Chang, a Richard Chang.

He, like those companies, are the leaders in the foundry space. I think, you know, it's an interesting, interesting world there.

Jon Y: TSMC China actually is the third leading so far as I know. So they have two fabs out in Nanjing and I don't remember the third, the second place, but like they were allowed to bring some high technology there and they're competing pretty hard over there.

So, traditionally, I would say that for the past eight, nine years, SMIC and Grace were basically getting the stuffing kicked out of them from TSMC moving over there because TSMC is so good at, at like basically wiping out the market's profit. And a lot of these companies, like SMIC, especially, they used western style. These were companies that were founded in a time when Chinese American and western relations were much friendlier. When you had transfer of technologies, adoption of these standards and adoption of exportability where like, companies could, western companies could commission SMIC to do chips and export them back to the United States.

Jon Y: And at that time, SMIC was basically a labor and tax play. Right. This was a company that didn't have didn't own the technology. Like if it was a restaurant, right. They were highly reliant. Yeah. Almost like a franchise, like, and not to say they were working really hard, but they're built in a way that it was kind of taking advantage of a certain location.

Which is fine when you start. But like, after a certain amount of time, certain things happened between TSMC and SMIC and all that. And then for the next couple of years, they kind of just stagnated. Now it's gotten to a time where they are becoming more important to the company or to the country's kind of technological future plans.

And 22 nanometers is fine for a lot of people and might be fine for a certain situation. But the problem is that it's like, it's not economical. Like, you can't stay there forever and make a profit because like I assume at some point you're barely making anything.

I don't know if that made sense. It's tough. It's an economic tough situation because like being on the trailing edge is not where you want to be.

Steve Hsu: Yeah, that's an interesting question though. But if for strategic reasons, China just wants to be independent in, whatever that means, like not reliant on the import of semiconductors from say from Taiwan or from Korea. And they're just happy to maybe the return on equity or return on capital is relatively low, but they are able to produce at say 22 nanometers domestically.

Is that a sufficient policy goal for them?

Jon Y: I'm sure that's fine. I mean, a lot of American military technologies are done with very trailing edge technologies of nodes as well. Like I know SkyWater, which is an American foundry, does like 180. Like that's their max, 180 nanometers.

Steve Hsu: Yeah, sorry to interrupt you. I didn't mean for the military because the military is like, totally like, very far from the cutting edge. Right? So, but I'm like, suppose these, these things are going to go into refrigerators and cars and internet of things and even low-end cell phones and things. Do they actually care that much strategically that they would actually, you know, almost subsidize that industry, even though it isn't even the cutting edge. But they just want it. They want to be able to produce this stuff there. I don't know. I don't know what the government there has actually said explicitly about this.

Jon Y: I think it's tough to make a call on that because it's a commercial decision too, right. I think if, think one of the reasons, like, are you, you're asking a Chinese consumer to say that you want a 22-nanometer chip. Because you want to support your own country. A lot of people I think will choose with their wallets and they'll choose a better product.

Steve Hsu: It's what you just said. It's where some people use this loose analogy between oil and semiconductors, right. Or oil and chips. Saying, oh, China imports more chips than it imports oil every year. In dollar terms. But of course, a lot of those chips it's, it's, it's embedding into a device which then it sells,

Jon Y: Yeah.

Steve Hsu: in Germany or, Australia or something. So that the whole analogy is not quite right. And then the thing is that, of course, one of the main strategic levers the U.S. has is they could cut off oil flow to China. But the analogy of them cutting off semiconductor flow is that even a viable analogy, like would that, is that even something that could happen strategically or that the Americans would ever actually want to do it?

Jon Y: It's a tough balance. I've actually been thinking about this, cause I wanted to do a video about cause Americans had this, actually, they had this debate with the Soviet kind of like a more thriving technology industry of their own.

And the Americans debated, they had the same sort of debates where they're trying to say, well, is it right to cut these off? Because then, maybe they'll develop indigenous capacity or let's just deny it and make a kind of, almost like, like a bet that the Soviet Union won't be able to catch up.

It's easy to look back at it and say, oh, they made a correct bet or something. But it was a very tough balance between giving them enough technology for them to get to being open enough to meet certain requirements in the markets, to try to open up markets and trying to kind of get the Soviet policymakers to think this is important. To make it important for them. But also cut off the most advanced stuff. And it's a very complicated story. I wouldn't say that there's one way or the other to do it, especially when it comes to China. I think that like, it's important to them to, for, I think right now the, the PRC acknowledges that they're not able to go beyond into the seven-nanometer, leading-edge world, and they're kind of building policy to deal with that, but does that mean that they're going to be able to develop their own capability, their own indigenous capacity?

And that means that Americans should continue denying exports or they should, actually be more open? I don't know. It's hard to say. Luckily I'm not the person who has to decide.

Steve Hsu: Yeah, let me, let me shift from, okay. So I was focused a little bit on the issue of not the bleeding edge, say seven, five nanometer processes, but maybe 28, where most of the chips actually are. And people who say like, well, it's not even really a realistic risk to the Chinese, that the U.S. would cut off, try to cut off, know, their access to that kind of chip because after all, like of those chips are basically turned into devices in China, through manufacturing processes, and then that's how the rest of the world gets their devices. So it doesn't even make sense that the U.S. would want to cut that off, because then you wouldn't be able to go to Walmart and get something. Right?

So maybe the whole thing is, is just symbolic that the Chinese want to be, that they somehow maybe make a false analogy between oil and chips and say, wouldn't it be great if we were quite self-sufficient in chips? That's one chain of argument.

But let me switch from the more commodity process scales to the bleeding edge. And let me just review how I understand the story of what happened to Huawei in the last few years. And you should correct me if I get any of this wrong. But a few years ago, I got interested in Huawei. I guess mainly because Trump had become president and Huawei was becoming like an issue and — I don't think we'll get into 5g in this conversation, maybe only at the end — but what I had noticed was Huawei was starting to make fully competitive world-class phones. Like I think the Mate 20, Mate 30. I don't know if you're into cell phones, but Huawei had, had made some phones, which were as good as anything Samsung was selling. And as long as you're not an iOS fanboy, it was in a sense as good as any iPhone. And I actually even bought a Huawei phone just to check it out. And it was using a … they have a design bureau within Huawei. It's called HiSilicon.

Jon Y: HiSilicon.

Steve Hsu: And the chip, the cutting-edge chip, which was made I think on like a seven-nanometer process was called a Kirin chip. And that was a big deal because they had designed that all in-house at Huawei. It was, you know, [leading] in the benchmarks and stuff, it was as good as Qualcomm and Exynos processors.

And so I thought, well, it was an interesting [phone], and this is a new milestone in climbing the value chain for a Chinese company. And effectively by not allowing TSMC to supply. So getting back to the design of the chip is Huawei's, but the fabbing had to be done at TSMC and the U.S. just by Fiat cutting off TSMC. Just said TSMC cannot supply any of these chips to Huawei. And now Huawei, around that time that I'm describing, had become close to the number one cell phone maker in the world.

Actually, if you look at their global sales, I think they were number one or number two, or close to number one briefly just for like one year. And then they got hammered by these sanctions. And so the U.S. was able to literally destroy a huge business of a major Chinese company, a flagship Chinese company, and it had nothing to do with 5g security. This technology had nothing to do with 5g base station. It seemed to me that Huawei got on the U.S. radar because of the 5g security threat. Like if Germany and the UK installed 5g networks all built by Huawei, the U.S., the NSA and U.S. security didn't like that.

Jon Y: Yeah.

Steve Hsu: U.S. intelligence. Somehow Huawei got on their roadmap, but then they destroyed the consumer cell phone businesses of Huawei, through their control of TSMC's fab. Do I have that correct?

Jon Y: I think there's some things that maybe we could yeah, I think there are, I think it's on the whole, correct. I think the thing that was key was that like HiSilicon and TSMC, when you're at the bleeding, at the leading edge, you work very closely together. So there's a lot of trade secrets being passed over from TSMC side to Huawei's side and Huawei side to TSMC side.

I don't really think the United States has a vested interest, it doesn't really care about Chinese cell phones in particular. Like they like OPPO, Vivo, all these other Chinese – Xiaomi – they're all kind of all around still. Right. Well, Xiaomi at the end nearly got smacked, but like I think the key was that like, what they wanted to do was remove that chip design part, that your chip design portion and to say what it is about what the Trump administration ended up deciding that decided to, they decided to pull the trigger on that.

It's a very fraught question. I got a lot of very angry comments about it.

Steve Hsu: Well, Yeah. I mean, we could talk more broadly about Huawei. I would, I would actually be very interested in that. But my view of this is okay. The U.S. can have a, especially U.S. intelligence services, signals, intelligence services, can have a legitimate concern that they don't want the whole rest of the world to switch to Huawei provided 5g.

Jon Y: Yeah.

Steve Hsu: Okay. can make that argument. I don't actually completely buy their argument about the security risks, but anyway, that's a separate discussion. But even if that is their issue, what they really did was it sort of like, know, like, I don't like what you're doing on the upper floor of the restaurant so I'm going to, I'm going to start a fire on the lower floor. It seemed to me very egregious. They basically, by fiat, destroyed a completely different part of the company, which that, what that part of the company was doing really didn't have any security implications, national security implications for the United States. So that's how I see it, but I'm, I'm curious what you think.

Jon Y: Hmmm. I reckon that what they saw was that they didn't want HiSilicon to become, to create chips that could be adopted by the rest of the world. Right. OPPO, Vivo. Xiaomi. They fought for all of that, even though they still make phones and they're very popular. They use Qualcomm chips. They use chips from Samsung or ..

Steve Hsu: Or MediaTek.

Jon Y: Yeah. MediaTek, MTK. Right. So like …

Steve Hsu: Sorry to interrupt you. But I, okay, I mean I was following this cause, I mean, I went so far as even to buy a Huawei phone just to check it out. But there was never any talk of other manufacturers using Kirin chips. Kirin chips, it was a little bit like Apple and M1. It was like, these are our chips. This is going to make us number one, cell phone, mobile phone company in the world, or one of the top phone companies. A competitor in a space, which, there are some strong U.S. companies like Apple and things like this.

But as far as I could tell it didn't have any connection to national security.

Jon Y: It wouldn't take a lot of knowledge to get a sense that it's because to get to, to feel like the next step would be to export the Kirin chips, right? The fact that HiSilicon has its own division. The fact that, that this company is trying to expand creating all these different lines, creating cell phones and all these other like different sorts of chips. I don't think it's like, very hard to understand that, like that's the next thing to do. And secondly, like I think, yeah.

Steve Hsu: I think the general, I mean, I think this is certainly true. The general intent of the U.S., the tech warriors, the people in the White House who wanted to hamstring Huawei. Yes. They wanted to hurt Huawei as a company. But I think for people who are somewhat impartial, like if you're some guy in Spain, you know, a lot of, there are a lot of people in Europe who really liked their Huawei phones and stuff, right. And you're just looking at this, what it really looks like is the U.S. is basically trying to destroy a top Chinese company. And some of the things that company does may affect U.S. national security, but the motivation to destroy their chip-making or cell phone-making business was the only motivation for that was primarily just economic. It wasn't actually based on national security.

Jon Y: I think what the Trump administration eventually saw, kind of the China situation, they saw it. Didn't see it as like a disparate set of different or autonomously moving parts. Right. They saw Huawei as like, if you find one part of Huawei, you're funding the other stuff that is competitive.

Steve Hsu: I totally agree with you. No, I totally agree with you that I think that's how they were thinking about it. They're like any way we can hurt Huawei we were going to hurt Huawei.

Jon Y: Yeah.

Steve Hsu: But if you take a step back, and you say, oh, you say to yourself three times, rules-based order, rules-based order, free markets. You know, what are you actually doing there? Right. I mean, you are really not operating in a very rules-based order way when you do stuff like that, right? Like why should TSMC not be allowed to fab chips for Huawei cell phones?

Just to close the loop on something. The base stations for 5g that Huawei makes, I believe they were using imported chips, not their own design and on like a fairly crude, like 20-nanometer, 28-nanometer process scale. So they were not cutting edge at all. As you would expect, it's a big blocky thing that you, attached to the side of a building. It's not a, it doesn't have to be bleeding edge. And they have also cut off I think their access to those chips that, that if you really wanted to cripple their ability to build 5g networks around the world, I can understand that like within that self-contained logic, okay. It's like, okay, let's cripple their 5g business. But I think if you look carefully, they crippled the cell phone business without anything like the same level of reason, national security reason, a justification as trying to attack the 5g business.

Jon Y: Yeah. I mean, you wouldn't take far for any, any American or non-American to kind of look, to see situations where, for instance, like the, the government would violate, its quote unquote principles to do a certain thing. That's not to say that I'm criticizing. I'm just, I'm saying that to like, for example, Japan and the states, they competed pretty hard in the '80s and '70s over semiconductors as well. And then the United States helped end that dominance too. What's the best way to say this? Like, I'm not surprised that the United States would do that. I would feel that.

Steve Hsu: I think I agree with you. I think you're saying that the U.S. has behaved in a strongly mercantilist way, using governmental power to help its companies compete against foreign companies many times in the past. And so you're not surprised that they did it to Huawei. And I agree with you, except that I'm sometimes in policy circles or think tank circles where people don't seem to understand this.

They actually, they actually think that this whole Huawei story was Huawei evil. Huawei, of course there's this little inconsistency cause the NSA is even more evil in a sense like, the NSA was caught monitoring Merkel's cell phone and Macron's and Huawei never was able to do anything like that.

But anyway, Huawei bad, so we destroyed Huawei and that's the whole story.

Jon Y: There's questions about that company, and I'm not saying this to be anti-American or that like I like red, white and blue Boba as much as anyone else, but I think like, they saw Huawei as like a front for the military, for the PLA. And they couldn't trace a provision for the ownership of this company. I don't know if they did, maybe they did see it. Maybe they looked at it and they said like, this is a commercial entity for the People's Liberation Army. And in the end, that's like, if Huawei was public and publicly owned and it was like it wasn't and kind of distanced itself from the government, I feel that that would kind of help the situation.

If you had a company that in the United States acted as a commercial front for the United States military forces, I think that company would get heat too, abroad as well.

Steve Hsu: Yeah, I mean, look, I'm as pro-American as anybody, but I think that a foreigner would look at Boeing and say, okay, well, they get these huge government military contracts. So isn't Boeing effectively subsidized by the U.S. military industrial complex? And should I therefore boycott their 737s?

It's a kind of similar situation. There might be a matter of degree here. You might say I was secretly Huawei's getting so much subsidy from the Chinese government that there's a mercantilist element already in its own existence. Like we can't compete, our companies can't compete against them because they're so heavily subsidized by the Chinese government.

That's kind of a separate argument, which, which could be correct. It's also true that many U.S. IT companies have been caught cooperating with the NSA to put backdoors in so that the U.S. intelligence services can spy on foreign heads of state and business leaders and things like this.

So that's a real thing. And sometimes in foreign policy, in geopolitics, it's all projection. It's like we've been doing this for 50 years and we think you're going to do it too. So that's why we will just point the finger at you and assume that Huawei's doing this stuff because U.S. companies have been doing this.

Jon Y: Or better yet, don't put them in a position so they are able to do it in the first place. I mean, yeah, it's a chess move. It's a chess move for sure. It's a pretty nasty one too.

Steve Hsu: Yeah. So coming back to semiconductors, like in the next year or two, is there a milestone that you're on the lookout for that will tell you either a, hey, these Chinese guys are trying so hard to catch up with one of the most, if not the most competitive global industries and you just can't government, can't just snap his fingers and make it happen.

So they're failing or these guys have their shit together and it looks like they are moving fast. Anything you're on the lookout for as a signal?

Jon Y: I think what will be very interesting to look at is what happens when the Silicon shortage ends, right. When the chip shortage ends and the tide goes out and suddenly all these companies are building like SMIC's building like a $3 billion fab, right? If that $3 billion fab goes online and suddenly no one wants to buy chips anymore because we're in a big recession globally, then you see kind of like how committed companies and governments are to certain policy directions, because that's when you start seeing kind of the wheat separated from the chaff. Because right now, when it's easy to make money selling any sort of chip, I mean, then that's great. Everyone can talk about all these technology things and all that. But at some point, that's going to turn. I don't see any reason why it wouldn't.

And at that point, like, there's going to be companies that's going to get hurt. It's a lot easier to think about throwing $30, $50 billion at your semiconductor industry when everyone's screaming about chips, but suddenly when everyone could get a 28 nanometer chip, well, does it make sense to put $50 billion if you're the Chinese government, as opposed to, like pensions or healthcare or something like that?

Steve Hsu: Yeah, I think, I think almost every semiconductor analyst now is pointing out that it's a cyclical industry. And so despite the shortages that we're having now, the massive investments now are definitely going to lead to overcapacity at some point. And so we'll have a bust.

It's actually kind of hard to convince people that, because I, I mean, the comments are, are very like, no, you're not you're this is wrong. We're in the new world. We're in this news like everything's different now. AI, cloud, gaming, crypto like that. It's hard.

Steve Hsu: That's a good sign that you're near the peak of a bubble is that people can't even entertain the other hypothesis.

There are sort of two aspects of this that I'm interested in. But, you know, you keep reminding me that they're interrelated because unit economics is so important in this industry.

One is the technical capability. Are the Chinese catching up in the really bleeding edge technical capability? Or even not bleeding edge, but just how do you run your fab so it's efficient and things like this? But, separate from that is, are they making a profit? Right. But you're sort of saying those are the same thing.

Like if you're not making a profit, it doesn't matter whether you can do X, right. No, one's going to pay you to do X if it costs too much, right?

Jon Y: Yeah. It's basically a show pony. Like a science fair project, right? It's not interesting. Like it has no future. It's a dead-end technologically.

Steve Hsu: Right. Well, it could have a future if the... Well, yeah. I mean, obviously that having those engineers around and having that capability and the machines prototypes built, it can be useful if you just, if you just have to catch up and get to three nanometers, like, it's just a crash program, like the Manhattan project. It's better to have those people around and not have those people around.

Jon Y: I mean, you'll have it. Then what can you really do with it? You've made a $152,000 chip. I mean, unless it can succeed in the commercial world, I really wonder if like, then you're going to have a situation where they put out a press release on it. You have the propaganda kind of whip up a whole bunch of fake YouTube videos about it, and then like, you're done with it.

It's just, I really want to press on that. Cause it's like, you don't have an industry if there's no one to buy your stuff.

Steve Hsu: Right. But, okay. But I just want to probe that. I agree. I agree with that point of view. I, but I want to probe it a little bit. So, if you say, this kind of stuff, it's only, you only really have the capability if you can do it in a commercially viable way, commercially competitive way.

How is that related to the idea that one country can actually cut off another country from complete access to some of these products? Right. So, given the U.S. has shown that pre-election right, they've almost killed one of the flagship tech companies in China, by just cutting off its access to TSMC. Does that mean that even if you can't do it at completely market competitive rates you still need to have control of that technology and be able to do it yourself, even if the government has to subsidize these companies to keep them afloat?

Jon Y: I guess. Yeah. It's important to kind of know that. And I think a lot of the technology concepts are out there. So. if you were a country that's opposed to the United States and or in their bad graces. And I would for sure say that, like it's important to have latent knowledge and latent understanding of how to build a conceptually, what would be a very advanced, leading-edge chip, whatever that might be.

Yeah. There's no doubt that that strategy is important. Right. And like, there are some applications, I think, to other areas of the world. I think it's true. I think that's very true.

Steve Hsu: Do you think that the current Chinese plan vis-a-vis semiconductors is rational or do you think they're just overreacting? Maybe they should wait and buy all this stuff when it's cheaper. Like when the bust comes or something?

Jon Y: Yeah. I think import substitution policies are very commonly practiced by a bunch of countries in Asia, right? It's a key first step towards building competency and expertise in a certain industry. Right. But it doesn't always succeed. And I think import substitution is very, I think China's quite good at it. They're very good at it. But it doesn't, the first two or three years is not really going to tell you if they succeed. You need a lot more time.

Steve Hsu: Yeah. That's so that's an answer to my question. Maybe your answer is there's no, we're not going to know in two, even two years from now. We're not quite going to know how this is all going to play out. It's still gonna be uncertain. Yeah.

I, I'm of two minds when, when you talk about the Chinese succeeding in climbing the technology chain for semiconductors, I can sort of give arguments on both sides for why they'll succeed or why they won't succeed.

On their side. They have a very deep pool of human capital and they have a very deep pool of capital, right. And very strong organizational skills, when they want to get things, to produce things, efficiently.

On the downside, this is, really in my mind, one of the most globally competitive industries. And so, that makes it tougher to just flip a switch and catch up. And you have to have stamina to really keep at it for a long time, probably to catch up. And a lot of the stuff like that, had to be done to get EUV working and stuff was just crazy and took 20 years. So, from that perspective, it's like, okay, maybe they're nowhere close to fully catching up. So I don't know.

Jon Y: I think it reflects, like, I think people's thoughts on it reflects what they think. It's a Rorschach blot, kind of showing what you really think works. In my case, I feel like the industry is going to be a tough uphill climb for China.

They have a lot of stuff they need to do anyways. I think the domestic market isn't big enough for them to kind of really succeed there. And at the same time, it is important enough that they really, that they really care about it. And they're going to keep trying to hammer at it. And they're really good at hammering at stuff until it finally works.

Steve Hsu: But is it really just their domestic market? Because if I want a device, which uses chips and I'm in Texas, or I'm in Brazil. If that chip doesn't go to China, then to have the plastic built around it, and then it's shipped to Brazil, if that doesn't happen, the rest of the world has to then rebuild that entire supply chain to actually manufacture the device.

Okay. Yes. The chip came from Taiwan or the chip came from Texas or something, but there's this whole other part of the supply chain that has to be rebuilt if you try to excise China from it. So can the world really cut China off from semiconductors? Maybe not.

Jon Y: I think so. I think it really depends on where the bulk of the value comes from, right? The bulk of the value right now comes from fabrication, packaging, testing, all these other services come from different parts of the world. That can be moved. Right?

For example, Malaysia, Indonesia, Philippines, Intel put packaging factories there. And then it's kind of moving them around to certain areas like Israel and stuff like that.

And then when you talk about actually putting the chips into the actual electronics product, that part is simply a function of how much consumers are willing to pay for it. Like it used to be that China had a substantial portion of shoe-making, right? The shoes were mostly assembled in China. That's starting to change. Now you have factories in Vietnam, Myanmar, Indonesia.

Steve Hsu: Let's take the shipping. Let's take the container that holds the most bleeding-edge chips, which is the mobile phone. Right. Isn't every single close to a flagship mobile phone basically made… Well, okay, I guess I'm leaving out South Korea, but almost every, leading-edge mobile phone is made in China, right?

Jon Y: I think to kind of take what I said before about what China's like. If a government is determined enough and is persistent enough in the plant, I don't think anything's impossible. These companies, like Foxconn, are Taiwanese companies. I think if there is a policy that turns out to be in that sort of way, I don't doubt that anything can be done.

I don't think it will be either, I think it will be an economically viable situation for both parties. I mean, nobody really wants that sort of thing to happen.

Steve Hsu: I think I was complimenting you earlier about how your, your segments are, like many histories of these companies and stuff like this. If I were to write a mini-history of what happened to Huawei in the last five years, it's kind of a strange story because they got so good at 5g, they got way ahead of everybody in 5g, which is a totally different set of technologies. It's like radio millimeter wave, radio technologies, and stuff like this. Right. They got so far out ahead of that, that the U.S. got afraid that they were going to build the whole next generation of, you know, information networks in friendly countries like in Western Europe and stuff like this. And as a totally parallel effort, they also became world-class in making cell phones. And their main use of really leading-edge chips was on the cell phone side, not on the 5g side. And yet they just got smacked down as a company because they were involved in 5g.

But if they had been two separate companies, I don't think anybody would have cared about Kirin processors or Mate 30 pro phones and things like that. I think it's a weird conjunction, something that the spooks really cared about a lot. The 5g networks around the world and, and, and mobile phones, which I don't think they actually care that much about. Cause I mean, they haven't knocked off Xiaomi or OPPO or anything like that.

Jon Y: Yeah, think it's, it's a, it's a weird situation and Huawei is a weird company. And I think it's really, difficult to talk about them because it's so flavored with the politics of the day. It's a tough situation. I don't think I'll ever do a video in the short term about this, about the Huawei saga, because I don't really want to get screamed at. But like, I think it's a demonstration of kinda like how maybe a state-owned company or a state-backed company can really kind of grow in a western and competitive market. And it's also a demonstration of kind of like, the old, certain overreaches of power by a government as well.

Steve Hsu: Yes. I think if somebody didn't already have a somewhat cynical or dark view of U.S. mercantilism. The fact the U.S. clearly practices mercantilism, just like every other country, Looking closely at the Huawei story, if you take off your, red, white, and blue blinders, at it that says you're a German looking at or something you can't help, but conclude that they whacked this company, really deliberately whacked this company, just to whack it basically.

Sorry. I keep repeating that. I'm being repetitive.

By the way, I have a little bit of insight into Huawei because one of my post-docs, who was trained in theoretical physics, but you know, a lot of bleeding off of people from theoretical physics into other fields. His first job after he left working for me in my lab was at Huawei's AI research center. They have a research lab for AI in Silicon Valley. So he went there and worked. He's originally from Hong Kong. And so he went and worked for Huawei for a couple of years. And now he's at, I won't say exactly where, but he's at, he's pretty senior at one of the other cell phone [companies]. You know, the Oppo, Vivo, you know, those companies. And he's an AI director there. But he had a pretty good inside view of what was happening in Huawei at least for some period of time. And it's nowhere, it's not an invincible giant or anything, but it's also not like an insidious arm of the PLA, it's not an insidious arm of the PLA either. It's a company that's just trying to make money.

Jon Y: Definitely. Yes.

Steve Hsu: Anyway, the truth is much more complex, I think, than people suspect. Yeah, so I've taken up, I think, probably too much of your time. I think we've covered most of the things that I thought we'd set out to cover. Are there any final remarks you want to make maybe more broadly about U.S.-China tech competition or whether China is going to invade Taiwan? And anything you want to say as, as we close out?

Jon Y: First, I hope China doesn't invade Taiwan. I live here. That'd be very unpleasant.

It's a tough situation. I hope it really doesn't go to that. My ideal situation is that, you know, China, Taiwan continue to co-exist with one another and continue to move towards peace in some way from both sides. So I think it would be really the ideal situation because I don't think anyone would want war.

No one would want war. No, there's no, I don't think there's any contingent of people who would want to, to, have a war of that sort.

Steve Hsu: It would have really global consequences for sure.

Jon Y: It would have global and personal consequences.

Steve Hsu: yeah, actually, if you don't mind staying on five more minutes, I would like to just explore this topic a little bit more and I'm sure my listeners would be interested too.

So one of the issues is, cause I'm much older than you are. So I don't feel like I'm in touch even though I have extended family in Taiwan and I've spent time there. I don't really quite understand how serious the shift toward the greens is in Taiwan and towards sort of a more nativist, native, native, Taiwanese, perspective. So just for my listeners, let me explain that, like my, the part of my family that passed through Taiwan, they were people who fled the civil war.

So when the communists won the KMT nationalists government relocated to Taiwan, and there are a bunch of mainlanders who didn't have any connection to Taiwan, except that they flooded there at the end of the civil war in 1949. And they're actually called 49ers, that whole generation of people who went to Taiwan. They took over Taiwan. They suppressed the Chinese that were already there. And the very small population of Aboriginal people that even predate Han Chinese presence in Taiwan. And, they're the ones who still effectively claim to be the true government of greater China. And that, there's still a kind of one China policy, both officially, from the Taiwanese side, and from the PRC side.

Now you have these younger people who grew up, in a democracy, they may be at home, don't speak, Mandarin. They don't speak Putonghua, they speak the Taiwanese dialect, which is a derivative of the Fujianese, I think. And they don't feel very Chinese. They don't feel like they're part of the mainland.

And so this pro-independence movement, which is led by the Green Party, which is currently in power in Taiwan, has become a stronger and stronger force in Taiwan. So the question is, is it now basically hopeless for reunification because a strong enough majority of young Taiwanese just don't want it?

Jon Y: I think the basic understanding of how kind of the Green Party is, the Green Party was like a pro-independence party. And it was in their charter and it was kind of something they strove for. They got beaten really hard by the Blue Kuomintang Party because of this particular issue, right?

Presidential elections are voted based on the China issue. Domestic primaries are based on domestic issues. And I think the Green Party realized that they can't be pro-independence explicitly. Like it's still in their charter and all that, but that's not something they can no longer sort of, kind of, call for. I think over time their views are kind of, kind of modulated to the point. You don't hear it called out to this particular point. To this particular sort of extent.

Steve Hsu: They're less visibly pro-independence. I guess there are some real firebrand radicals in the party that still are, the president for example.

Jon Y: Those people don't really understand. Yeah. And they wouldn't understand that those people should probably never get into power because they wouldn't understand. Cause I think China's pretty serious about this sort of particular situation. And Americans wouldn't want it either cause it's one of those situations that would trigger a war.

But I think what the Green Party has done is sort of this very sort of slicing, creating a sort of like a distinct Taiwanese culture. And that's part natural and that's part encouraged by government policy.

The Green Party, they're not like some sort of insidious sort of beast and they're not sort of like, like, I feel like they're being portrayed outwardly as kind of like this, this power or something like that. They're just like another party trying to kind of shift policies and their own sort of way. And while supporting all these other things and doing all these other things and winning elections at the same time.

I think a lot of what you see here, a lot of what I see on the ground in Taiwan is that like they've accepted that there is some sort of identity separate from China, mostly amongst the young, younger people, the older people kind of see it as like, well, we see ourselves as Chinese, but right now what's going on in the mainline is kind of weird. So we don't really get involved with that. So it's kind of like a tacit sort of alliance between those two if that makes sense. So that's why.

Steve Hsu: No, that was great. But when you say the stuff, that's weird, what's going on? Is that because of Hong Kong primarily or because of Xi Jinping?

Jon Y: Yeah, there was a period of time when it really seemed like the two sides would be able to sort of find a negotiated peace. And I think it was very close and I think partially because of Hong Kong, but partially because like the CPC did change like did make a sort of a hard turn away and who knows who started it, but it's like, the CPC will have their own perspective to say that like, the Taiwanese have been trying, taking more than they can give. Right. Like them, they give more rope, take more rope. Right?

So it's like, it's sort of a he says, she says thing. But there was a point in time where like, you know, it really seemed like the two could have been really tight together. There was increasing integration And there's really close, sort of perhaps there was a satisfaction between the two things that have changed a lot more. And I think it's partially because of, you know, Xi Jinping COVID, Hong Kong, all that. And I think it will take some time to overcome it, but I think like it's harder to make that argument now because like, it feels like the government has made a sort of explicit turn away.

Steve Hsu: I think to me, the best-case scenario is probably that we have status quo for another decade and maybe China takes a more reformist turn again, right after Xi Jinping, whoever succeeds Xi Jinping. And then maybe at that point, China will be perhaps very developed and there can be a kind of two-state, one state, two system kind of agreement between the two.

And hopefully, nobody wants to go to war and that's to me the most optimistic thing that can happen.

Jon Y: Yeah.

Steve Hsu: I Was at Academia Sinica, which is you know, pretty, I guess, elite science research Institute in Taipei. I was on sabbatical there. This was over 10 years ago. I remember conversations with the other physicists who are very rational, smart people, but a mix of greens and blues. So greens are the pro-independence party, or at least somewhat pro-independence party. And the blues of the definitely not independent KMT party. And I used, we used to have these pretty difficult conversations, but you know, like I think they would not argue with each other cause they, they, they knew each other personally and just had to work with each other every day.

But I was a strange American guy so if I brought it up, they would humor me. And I used to ask the greens. I say, what's your scenario that you guys are going to escape, being reunified because I'm pretty sure they really want, talking about the mainland, to reunify. And assuming they continue to develop economically, eventually, they're going to be able to do it against your will.

So I only see two possibilities. Either you do reunify at some point. You don't have a choice. The other option is the Chinese under the Chinese civilization undergo some collapse, again, like the mainland, you know, self-destructs. That's the only way you're going to escape. If the mainland doesn't self-destruct, you're not going to escape, you're going to become part of China again.

And these green physicists after I pose it that way would say, well, there's no hole in your logic. I can't disagree with you, but I just hope emotionally that we somehow managed to be independent someday. But I said like, would you be willing to be independent if the cost was a billion people in the mainland were thrown back into savagery because of the collapse of their system? Because that's the only way I see that you get out?

Do you agree with me? Or do you think that's crazy, what I just said?

Jon Y: Hmm, I don't really think it would need to be that extreme. I think for the most part, people are pretty happy with the status quo. I think those people who are explicitly calling for pro-independence right now are probably the younger types who feel like they would be able to, you know, this is the things that should be, but over time they will gain perspective too. Like the party itself.

I think there's still a lot of cultural ties between both sides and, you know, it's for the most part, the individual people seem to be quite cordial to one another. So like, I think in the heart of it, it's just, the window's not open right now.

The window can open again.

Steve Hsu: Yeah, no, I agree with that. I think everybody right now just wants to kick the can down the road and not make waves. And I think that's totally the right perspective, but I guess what I was trying to emphasize is that it doesn't seem likely that the way out is that the mainlanders are going to give up on this.

And if they continue to develop economically, then you're not going to have a choice.

Jon Y: Conducting an invasion across 100 miles of sea onto an island that is largely difficult to, that's been quite militarized or fortified over 50 years is going to be difficult. This is the most I'll ever say about this because I'm not an expert at all, but I think it's going to be really difficult.

It's really difficult. And I don't think any military would want to try that unless they have to.

Steve Hsu: If you're interested in this topic, I could have a whole other episode to discuss it with you. But the one thing I will say is that the ease with which the mainlanders could blockade, a hundred percent blockade Taiwan, using missiles or subs, anything because the technologies are so advanced now. People forget how war is different today than what they're used to. They could interdict all shipping and air travel in and out of Taiwan. Like it was nothing. And when you get 50%,

Jon Y: Right. But then you have to consider the counter steps beyond that, right. Like, obviously they would have done it a long time ago. So it's like saying to them, what did they think was going to happen?

Steve Hsu: This isn't a situation where they're the largest economy. In the world. Right? So this is like, what's going to happen to you if China continues to develop and their GDP reaches a half U.S. per capita level. So their total GDP is twice the U.S. level. And they just say, hey tomorrow, no ships come in and out of Kaohsiung. Cause we have the missiles. We have the satellite coverage, we have the subs. Just no energy. There's no energy. And there's no food coming into Taiwan. I'll just wait until you guys surrender.

Jon Y: I think like, I see them possibly doing a blockade, but I don't see it coming simply because they're richer.

Steve Hsu: Like they're going to be at a point where they can basically say to the Americans and these other guys, like, I don't really think you want to intervene in this one and we're not going to invade them, but we're just going to cut them off and then they're going to surrender. And I don't see a counter strategy for that.

In the scenario where we're total Chinese GDP is twice U.S. GDP and they have continued to develop technologically the way that they're developing.

Jon Y: We'll see if that happens. I think things are a lot more complicated than that. I think there's always communication between the two sides and like the United States has always can get involved as well. And there also should be considerations on what it will do to whatever wealth that's owned on both sides of this, of the strait. So I think it's a very sensitive topic.

Steve Hsu: No, I wasn't saying that that super simplistic scenario was actually going to happen. I'm just saying just as a strategic exercise, I think they could do that if they wanted to do it. But just in terms of actual capabilities in this future scenario.

Jon Y: It would represent the end of lifestyles for all of us on this island, including myself. And I think it's, difficult for myself to think about or kind of consider that.

Yeah, it's hard to say.

Steve Hsu: But when they threatened that blockade, that the piece of paper that they're going to dangle for you is like, no, it's two systems. You guys do whatever you want. I think it can be something like that where like, we grant you the autonomy that we originally granted Hong Kong or something. Right?

Jon Y: I don't think that option is so much on the table anymore.

Right now.

Steve Hsu: Yeah, I think, think right now, but in 10 or 20 years, people will maybe forget what happened in Hong Kong.

Well, we'll see. I mean, it depends on what happens in the next 10 or 20 years in Hong Kong.

Jon Y: Very true.

Steve Hsu: Yeah, but you know, just one final comment, you know, like when, when east and west Germany were reunified, I think the cultural gap between the east and west Germans was much larger than what the cultural gap is really between-- I could be wrong on this, but this is just my impression-- between Taiwanese and mainlanders.

They mix pretty much effortlessly and there's pretty good overlap in things like the media that they consume and the things they think about and what they know about. And, you know, it doesn't seem like there's a huge gap.

Now, maybe that's because mostly I'm seeing these mainlanders and Taiwanese mix in academic settings and stuff. And there just really isn't much difference. But it just seems like it wouldn't be that difficult. Um, just purely cultural grounds to reunify.

Jon Y: There are differences in terms of kind of like, you know, snobbishness and stuff like that. But I wouldn't say that. Yeah, I would say the differences are quite subtle. Dialect, values. I think it's maybe nice to say that Taiwan is like China's Canada. They're quite nice.

Steve Hsu: Yup.

Jon Y: Meanwhile, you know, the mainlanders are a bit more aggressive, a little more, little more...

Steve Hsu: Blunt.

Jon Y: Yeah. A little more blunt. It's different, but, but yeah, they're cordial with each other and like they speak the same language. So it's like, they're always going to be. I think, personally, like they're very kind to each other. But you know, politics is always complicated, with the way things are.

And on both sides, I think both sides don't really like their own politics to some extent.

Steve Hsu: Yeah, no, I mean, another solution would be the mainlanders do so well that they just forget about Taiwan. They're like, what do we care? Let these guys be independent. Like, why is this an existential thing for us? Let them be. We're already so successful on our own.

That would be a great outcome too. Right? I would love that. I would love it if that just happened.

Jon Y: That's unlikely, but yeah. That, that would be very nice as well.

But I would say it's also just like, you know, just, they continue to hold the claim, and they never, they could also just like, they got other stuff, they got bigger fish to fry right now. Right? Like they have other things they need to do at some point. Like you can still hold the claim, but do you have other things you need to deal with? And especially considering that Taiwan is literally smaller than Beijing. I think China's got other things they really need to focus on.

Steve Hsu: Yeah, I totally agree. I sometimes wonder why. I think it must be some internal domestic politics thing where hardliners just won't let it happen. But if you are a little bit more of a reformist leader in the PRC government, I can just see you wanting to just throw all the propaganda in reverse. And just say, oh, Taiwan who needs it anymore. They're such lovely people. We should just let them have their own experiment in their own governance. You know, like it just throws the propaganda in reverse, and then maybe then it won't be a big deal. And it's, everyone will just forget about Taiwan. But I think the hardliners wouldn't let you do it.

Jon Y: Probably. Yeah.

And it's also, I do think that that does misunderstand how deeply inculcated this sort of stuff is.It drives back to the hundred years of humiliation, drives back to all that other stuff, Imperial interlude. So it's like Imperial intervention.

So I think it's going to be really hard to tear that out, but I do think it might be simpler to simply just let it stand.

Steve Hsu: No, for now, for sure, yes. That's the best solution. The reason I said to these greens, these Korean physicists that I just don't think the mainlanders are going to forget about it. And if they continue to get stronger, then you're not going to have a choice.

That was my, you know, my primary assumption was that it is so embedded in the hundred years of humiliation and it is still an unrecovered part of the planet. Right.

So that's my baseline, I think.

Jon Y: All right. Well, yeah.

Steve Hsu: Great. Hey, I've taken way too much of your time. I really appreciate it. If we ever meet in person, I'm going to buy you multiple drinks.

Jon Y: Are you going to edit two hours of the podcast? That's crazy.

Steve Hsu: Well, maybe I'm going the Joe Rogan route. We'll see. Well, maybe we'll edit it down if it's too boring, but, uh, I thought it was great.

Jon Y: Sounds good. Fantastic. Thank you.

Steve Hsu: Take care, Jon.

Jon Y: Take care.