Fresh out of the studio, Chris Miller, associate professor of international history from Tufts University shared the key themes and takeaways from his new book "Chip War". While illustrating the origins of semiconductors that built Silicon Valley, Chris traced the story of semiconductors onwards from the 1970s to today that brought forth the rise of Asia in the past few decades and how Japan, Korea and Taiwan built their expertise within the chip supply chain along with other key players. He offered his perspectives on the present state of affairs and explored whether China can develop their semiconductor industry by decoupling from the rest of the world and risk a conflict with the United States over Taiwan. Last but not least, he provides a glimpse into the future of the semiconductor industry in the next decade.
"It's impossible today for any country to do it all on their own. And even if you looked at the United States, which is still the biggest player in the supply chain by far, it's still the case that the US can't do it all on its own. As you mentioned, it imports lithography equipment from the Netherlands. It imports chemicals and materials from Japan. And then the most advanced fabrication of processor chips is in Taiwan. So, no country can do it alone. And really no country is even close." - Chris Miller
Editor's note: Chris Miller has recently won the prestigious Financial Times book prize of the year in 2022, and our host and team congratulate him on his success. We are honoured to have him on Analyse Asia podcast before his win. We have obtained the approval to release the edited video and added the transcript for those who want to dive deep into the themes emerging from his book "Chip War". Do enjoy the transcript here!
Bernard Leong: Welcome to Analyse Asia, the premier podcast dedicated to dissecting the pulse of business, technology and media in Asia. I'm Bernard Leong and semiconductors brought the rise of Asia and are now the choke point technology that curtailed the rise of China. With me today, Chris Miller (@crmiller1 , Linkedin), associate professor in the Fletcher School of Tufts University and author of his new book "Chip War" (Amazon, Simon & Schuster) to help us to look at semiconductors from a historical lens and provide us with the perspectives to look at the future. Chris, welcome to the show.
Chris Miller: Thank you for having me.
Bernard Leong: Yes, and I have spent a weekend reading Chip War, which is a very good narrative read, and in fact, it seems to tell the history of the world from the lens of a semiconductor. So, in this podcast, we always start by getting to the origin story of the interviewee. So my first question to you is, how do you start your career?
Chris Miller: Well, I began as a historian, focused on the Russian economy and the history of the Soviet Union, and I previously wrote three books on different aspects of Russian history and political economy on the pathologies of the policymaking process in the late Soviet Union. Then my second book was on economic policymaking in Putin's Russia called Putin Omics. And then finally, a broader survey of the relationship between Russian economics and population trends and infrastructure with its foreign policy in Asia. And I came to this topic, semiconductors, not planning to write a book called Chip War, but planning to create a book analyzing the arms race and the missile.
During the Cold War, the question that I wanted to answer was, why was it? The Soviet Union, like the United States, could build nuclear weapons with, you know, extraordinary explosive power and in large quantities. Why was it the Soviet Union, like the US, could build rockets and send people into space and send the first satellite and the first person into space, but they couldn't?
Miniaturized computing power with any success at all and fell far behind in computing. And this struck me as an interesting puzzle because everyone realized that computing would be not only economically important but also militarily important. And so there was every incentive in the Soviet Union to invest heavily in computing, to try to find ways to apply computing power to military systems but didn't work. And so as a result, the Soviet Union, which kept up with the key technologies of the arms race in the 1950s, sixties and seventies, which were long-range missiles and nuclear weapons, failed to keep up with the key technological trends in the 1980s and nineties up to the present, which is the application of intelligence to military systems.
That was the puzzle that led me to this book. The more I learned about weapon systems and missile programs, the more I realize that the interesting part of contemporary weaponry is not the engines that power rockets, nor the metallurgy that's involved in making the case of the rockets, but the guidance computers inside of them, and that's true for basically all weapon systems today.
The interesting part and the part that makes them useful and pliable is the guidance system and the communications. Enable it in the broader infrastructure of acquiring, sending, and processing information that connects weapons to their targets. And that was fundamentally a question of computing power. The more I dug into that, the more I realized that semiconductors stand absolutely at the centre of the production of military power today. And not only military powers, but I'm also sure we'll discuss all of the key trends in recent history. I came to realize, not be understood properly without understanding the computer chips that enabled them and drove them.
Bernard Leong: So it's a pretty business discovery that actually from the military that leads you to the history of semiconductors and that spanned the history of the world in the last three or four decades, and still the story is ongoing. So I have one more question for you. What are the interesting career lessons that you can share?
Chris Miller: With my audience to your own experience, I guess, from being a historian academic now writing a book on semiconductors, I guess what drew me towards this topic originally was not any sort of plan, but just a desire to understand at a pretty granular level of the details of military technology. Then I decided to write this book, chip War, not solely because of the military aspect, but when I was able to understand some of the connections between that and several other contemporary themes today.
I was doing the initial research in the very early stages of when the US was beginning to dramatically restrict the supply of chip technology to certain companies in China, like ZTE and Huawei. I came to realize just from reading the headlines and newspapers in 2017 and 2018, the extent to which semiconductor production and design were centralized in a small number of countries, and a small number of companies. That helped me understand the lines that connected the dots from the early Cold War story that I was analyzing historically in the present day that was in the headlines.
The third data point that helped me triangulate the fact that there was something really important going on here was when I was studying some of the trade data in the late 2010s and came to realize that China has spent more money importing chips than it's spent importing oil, which is an extraordinary data point, and helped me realize that actually I didn't understand globalization in its full context or was missing something fundamental.
When you start digging into trade data across the Asia-Pacific region, what you find is that semiconductors are absolutely fundamental in Taiwan. Over 40% of exports are integrated circuits in the Philippines, Malaysia, 25%, Korea, over 15%, and. Any description of globalization that doesn't put the semiconductor trade absolutely at the centre is missing something important. I think piecing together those three different facets, the military story, the trade and globalization story and the US-China dynamics were helpful in helping me realize that this wasn't just a story of military technology, though it was important. It was a much broader story that brought together different strands and issues that I was interested in, but hadn't previously realized the connections between.
Bernard Leong: That comes to the main subject of the day on the book Chip War, and I completed it recently before this interview. From the reader's perspective, it did feel like the history of the world and even the rise of Asia were written from the semiconductor's point of view. You have talked about the inspiration of how you came to write this book by just tracing back from the Soviet Union's military history and weapons and looking at the trade history of what China is tracing. The first question I probably want to ask you is, what are the key themes and major takeaways from Chip War that you intended for your audience?
Chris Miller: There were a couple of things that stood out to me. First was the extent to which Moore's Law. Predicted exponential growth in the number of transistors that could be put on a chip, and therefore the amount of computing power that a chip can produce was a far more dramatic force than I'd previously realized. I'd heard of Moore's Law but hadn't thought through the consequences, and in fact, nothing in contemporary society can be understood without the fact that we're able to apply vastly more computing. To all sorts of problems because of Moore's law. So the first thing is just the overwhelming importance of exponential growth and computing power to all aspects of modern life.
The second is the extent to which military and strategic considerations have been a fundamental driver of the chip industry since the earliest days since the first chips were produced for use in missile guidance computers, and. Today over 90% or almost all chips go into civilian uses smartphones and PCs and things of that nature. But it's still the case that it's often that defence uses are driving cutting-edge research and development in semiconductors. It's increasingly the case today compared to the past couple of decades that concerns about.
The application of semiconductors to produce military power is at the core of what's driving government policy towards this industry. So the complicated balance between commercial considerations that drive companies and strategic considerations that drive countries has also been there since the founding of the industry and the. Facet is the extraordinary scale at which the semiconductor industry exists and the scale happens both at the macro level and the microscopic level.
They're related. The transistors on the chip in your smartphone, for example, are among the model devices humans have ever produced, and there are over 10 billion of them on your smartphone if you've got anywhere near a recent model. But it's only possible to have miniaturized transistors by the billions and billions because we've got extraordinary scale. The factories that produce these transistors on your ship are some of the most expensive and large-scale factories ever produced. And so there's a deep interrelationship between the massive scale of the biggest semiconductor companies. and the fact that they're able in a cost-efficient manner to produce the tiny scale of the transistors inside, and understanding that dynamic between macro and micro and the fact that you couldn't have the miniature scale without the large concentration by a couple of key producers was crucial in understanding the ways the industry has transformed over the past couple of decades and why we ended up in a situation with a tiny number of firms control the production of all the computing power that the rest of the world relies on, which is a bit ironic.
Bernard Leong: When I was growing up, a lot of people were saying that the semiconductors are commoditized, and then the investment bankers were froing all this thing about hardware being commoditized. Today this is the actual lifeblood that is controlling most of the trade between Asia. and the rest of the world. The best way to start talking about your book is to talk about the history of semiconductors, which came from Silicon Valley. In fact, a lot of people who don't study the lore of Silicon Valley are that actually whoever started Silicon Valley actually served in the military at the very start. What I want to get you to do is, can you summarize the story of the traitorous eight, and how part of this team, for example, Robert Noyce and Gordon Moore, who was famous for Moore's Law, eventually came up and ended up starting Intel?
Chris Miller: in. So the story starts with William Shockley, who was one of the three scientists who invented the transistor in 1947. Shockley was a brilliant theoretical physicist, but he was a horrible manager. And although he wanted to become a wealthy businessperson, he objectively failed in this task, and Shockley couldn't figure out how to commercialize the product. Because he was a horrible manager, no one else wanted to work with him. So, eight of his smartest young engineers abandoned his business in the 1950s and set out to found their own company, Fairchild Semiconductor. Although these eight individuals hadn't played a role in theorizing the transistor or inventing it, they played a fundamental role in turning it into a valuable product.
This story here is a great underlining of the huge difference between scientific invention and commercial commercialization, The semiconductor industry has relied on all sorts of scientific inventions and Nobel Prize-winning research, but it wouldn't exist in the way it does if it weren't for the commercialization step. That's why Bob Noyce, one of the founders of Fairchild was extraordinarily good at matching new science with commercially viable products. The pursuit of this goal played a major role in coveting the integrated circuit. So the first piece of silicon or germanium had several transistors on it and set the semiconductor industry and motion from that point. Noyce realized that in the early days that his primary customer would be the military and his first big sales were to the military. But he also realized from the very early stages that the military was going to be a limited customer in the long run for a couple of reasons. One is because the military wanted a lot of interference in the specific products that Fairchild was producing;
and two, because the military was willing to pay top dollar for small-volume production runs, but was never gonna buy millions and millions of semiconductors. He tried to capitalize on the fact that he had the military as a customer willing to pay high prices. But then he use that to springboard to a much broader civilian market. That was his vision really from day one. And so Fairchild grew quite rapidly, not primarily thanks to his military customers, although it was important because he was the first to realize that there was a vast civilian market out there and found a way to serve it.
Noyce built Fairchild along with Ford and others for about a decade. But eventually, they became frustrated with the fact that the company that had seated the funds refused to give him a large enough stake, and refused to give him stock options. He abandoned Fairchild, along with Gordon Moore founded Intel in the late 1960s with the vision of producing chips primarily for the mass civilian market.
Intel's initial product that had focused on was DRAM [dynamic random access memory] chips, which are the most widely produced ship since then in, in the history of semiconductors. From that point, he proved that actually, the big market was in chips for civilians. Even though margins were lower, even though the price per chip was lower, the volumes were exponentially higher. That set the industry on its civilian focus, which persists to this day.
Bernard Leong: Robert Noyce is the same person who eventually became given the nickname the mayor of Silicon Valley and then subsequently became the mentor of Steve Jobs and all the subsequent entrepreneurs in Silicon Valley and jumpstarted that growth. One interesting thing that I've really learned from the book, which I didn't appreciate was the reason why Russia didn't evolve the same way as the US in the military or even in developing other systems that are very related to semiconductors. And how did the semiconductor in the industry actually eventually end up with consumer electronics? Because the origins of this focused on the military and the space race that the Russians couldn't replicate that?
Chris Miller: Yeah, the Russians were from the very early stages aware that semiconductor chips would have military uses. They invested heavily and put a ton of capital expenditure into the chip industry and sent many of the best scientists to work in the chip industry and had leading physicists who won Nobel prizes and science related to semiconductors. In addition to that, they spent a lot of effort trying to acquire semiconductors and semiconductor production equipment from other countries, primarily from Europe and the United States. But the Soviets failed to develop a sustainable industry in the long run. They had a small industry focused on military production, but it was always far behind for a couple of reasons.
One was that they didn't have the civilian market at home with nearly as much to scale. They were always producing at far smaller production runs than in the US or Europe or eventually Japan, which could serve the rest of the world. The Soviets were producing for themselves and a small number of militaries in central and Eastern Europe. So that was one issue. The second issue equally significant was that the Soviet Union never adopted an outsourcing model for assembly and testing. And so whereas the Japanese and European firms very early on, turn to Taiwan, Hong Kong, and the Philippines as a place to find cheaper labour for assembly. The Soviet Union didn't do that. And so costs were always higher. The third factor was that the Soviet Union was focused on. Copying and catching up without any exposure to world markets would force companies to be efficient, and produce products that markets wanted. There was always a copying mentality in Soviet semiconductors that never produced any sort of durable, innovative structures or research and development processes that worked. These factors combined to produce an industry that existed for a long time. The Soviet Union was able to produce decent chips, but never to produce anything close to the cutting edge, and ultimately ended up failing the Soviet military because it remained so far behind with lower quality levels and lower production run than what the US military gets access via Silicon Valley.
Bernard Leong: This is interesting because if you think about it in parallel today for China, they have brought the chips from the US and they never developed the capability. One of the big major differences is that the Chinese have an open economy and at the same time, were actually able to have the economy to develop semiconductors, which they did and they also built on top of it. We'll come to the China story a little bit later because I want to trace the story as it tells the story of the rise of Asia from the semiconductors' point of view. So the interesting part is history doesn't repeat itself, but actually, it seems to rhyme. So can you talk about the rise of Japan and how it actually beat Intel in the process and then course, the very famous story that was told by any growth? Andy Grove, the CEO of Intel espouses that only the paranoid survived. This forced Intel to pivot to microprocessors in the 1980s.
Chris Miller: So Japan was a player in the chip industry from pretty early days, and the Japanese government tried hard to incubate successful semiconductor firms. As the chip industry began focusing on DRAM memory chips in the 1970s and into the 1980s, it opened up space for other countries to really compete in an active way. and the reason was that DRAM chips from the seventies up to the present have basically been singing along the same trajectory, and it was pretty clear what the trajectory would be. More advanced types of DRAM. Pretty regular production schedule. If you compare DRAM chips between different countries and companies, it mattered less which DRAM you were buying because they were all kind of the same. What mattered is who could get out the next generation the fastest, and being a couple of months in advance, made a difference, and who could produce them at the lowest cost of the highest quality? Those were parameters that were less. Thinking creatively about the markets you were serving or designing new products and more about effective execution plus the cost of capital investment and your willingness to invest capital.
Those were criteria by which Japanese firms were able to outcompete US firms in the late 1970s and throughout most of the 1980s in terms of quality in Japanese. Because they realized to a greater extent than US firms in this time period that semiconductor production, especially DRAM production, was a manufacturing problem rather than an innovation problem. They focused on manufacturing quality to agree that most American firms in the seventies and early eighties didn't. As a result, their quality rate was far higher. They had fewer defects than American firms. Secondly, because the DRAM business was a race to get to the next generation fastest, and if you didn't get to the next generation fastest. Your generation would produce a lot less money for you because your competitors would be able to compete with you for that generation.
The ability to invest at low-interest rates for a longer period of time was a huge advantage, and that's where economic differences between the US and Japan played a major role in part because. In the 1980s, US interest rates were high as the US was trying to fight inflation. Also, in part, because US firms had access to capital markets, which generally charged higher interest rates than Japanese firms, which were accessing bank loans from banks that were often closely tied to the companies themselves.
So Japanese firms faced lower interest rates, could invest more and won a lot of market share over time as a result. Now, they didn't actually make a lot of money over time as a result, because of Japanese firms. Probably overinvesting and competing with themselves at the end for market share. It wasn't a very profitable strategy, but it was a great strategy for winning market share. It challenged a lot of US firms at the time, which had to exit the DRAM business in most cases and focus on other markets. Many firms failed to make that transition somewhat bankrupt. But Intel did not fail. Andy Grove, the CEO of Intel at the time, abandoned the DRAM business. Intel was founded to create a new product called a microprocessor, which was at the time a very niche type of good, but it was growing rapidly and it was being put in a product called the personal computer, which at the time was small, but growing in importance. And because of that pivot from DRAM to microprocessors, Intel was the company that played the dominant role in producing the chips for PCs and defined the industry from the 1990s all the way through the end of the 2000s. The Japanese chipmaker strategy didn't actually serve them very well in general. They had a market share in the eighties, in the early 1990s, but profitability was low and they. Specialized in a type of chip that was more commoditized in its production and therefore more vulnerable to other competitors.
In the 1990s and 2000s Korean firms outcompeted them. Today, Japanese firms don't play much of a role at all in DRAM production. Korea is the biggest player in this sphere, and of course, with the history of Intel.
Bernard Leong: Because of the Japanese experience, Intel became vertically integrated. That is the problem that they're facing now in order that they need to transform as well because everybody has so outsourced that manufacturing to someone else to do, which they don't. That's where Intel has been having trouble in the last couple of years. One interesting part of the Japanese story, which I want to highlight, is what was done by the US government to constrain the Japanese so that they don't take over the US enterprise. Because I think that was a part of the story from your book that there was an alliance between Intel and the other chip makers to try to contain Japan's rice as well.
Chris Miller: Yeah. they were both industry efforts and government efforts. The government efforts I would say didn't really work. The US government threatened tariffs on Japanese ship exports to the US and eventually Japan in order to avoid tariffs imposed voluntary export quotas, which drove up DRAM prices but actually benefited Korean producers and not American producers. So that was, I think, a failed trade policy. it might, You can argue, you might have heard Japan on the margin, but it certainly didn't help the US. The US funded a research organization called SEMATECH, which was designed to boost research and development in the US and here the picture is mixed at best. I think the US debate in the 1980s was fixated on the collaborative research and development between the Japanese government and Japanese firms, was that explained Japan's success. It's probably not true, but that was the mainstream view. The US tried to replicate that via SEMATECH. SEMATECH had a mixed record at best in its efforts to save us.
Lithography firms totally failed in their efforts to help with methodology and inspection tools. The deposition space is where the companies were involved and I would say that SEMATECH played no role. So it's hard to argue that SEMATECH was really a major player in the turnaround. Maybe it helped on the margin, but I'm not sure. Then the US industry, not all, but many players in the US industry, as they exited DRAM, as many firms did, wanted to make sure that they had competitive research and development.
There was fear and probably not all that justified. There was fear in the US industry that Japanese firms might collude and drive up prices. So there was a desire to work with DRAM producers in other countries. This was just as in the late 1980s that a couple of Korean firms, Samsung most importantly, were beginning to move from the assembly and test work they'd done previously to the actual fabrication of chips. So there was a fair amount of technology transfer from the US to the South Korean firms, which might not otherwise have happened. The US firms were more comfortable with South Korean firms because DRAM was something that the Japanese were doing better at than they were. It seemed to be more of a commodity in which technology transfer was less problematic. They did have a tacit alliance between some big US firms and some rising Korean firms. To try to ensure that there was plenty of competition in the DRAM market, which for US firms by the late 1980s was seen as a good thing because so many of them had already exited the DRAM business. They wanted a more competitive DRAM market with both Japanese and Korean suppliers.
Bernard Leong: This is interesting because in the eighties, we talk about the rise of Japan and then the US did a comeback and then Japan start to fall. In the 1990s, the Koreans came up. How did South Korea manage to enter into the semiconductors race with Samsung? Today, Samsung's semiconductor business is still one of the best in the world, of course, with their solid-state drives, and they are OLED screens as well.
Chris Miller: Yeah, I think Samsung is one of the most extraordinary cases of going from a tiny role to playing a major role. It began in the chip industry already in the 1970s. It had some role in assembly, testing and a bit of fabrication, but it really wasn't until the late 1980s that the company really bet a lot on semiconductors versus Japan on building out its chip industry. Partly it was due to the fact that the company made a big gamble and put a lot of resources into it. Partly it was due to the fact that the South Korean government's educational system had spent a lot of resources learning about semiconductors, sending students abroad to study semiconductors, and attracting professors to South Korea to teach about semiconductors.
So there was a lot of knowledge transfer that benefited the entire country. But Samsung in particular also has a research and development facility in Silicon Valley as well which accelerated the transfer of knowledge. It was due to the fact that the Korean government had identified semiconductors as a priority and encouraged Korean banks to lend to semiconductor-focused businesses. Partly it was because Samsung had an extraordinary vision for how the business could be run and understood the DRAM market as well as anyone. Samsung was able to combine the fact that it had cheap access to capital with an understanding of the DRAM business that let it really edge out its Japanese competitors, which were main competitors in the late 1980s, and especially in the early 1990s. It benefited from the fact that Japan faced its financial crisis, which dramatically changed the landscape for Japanese firms. They had to start focusing on profitability the way they hadn't previously, and so that, that was a major help to Samsung. There was still a lot of competition that Samsung faced, coming from Taiwanese firms and from Micron in the US. Samsung outcompeted them all. So I think you have to look both at the structural factors, and the fact that Samsung executed very well and really has over most of the past 30 years.
Bernard Leong: This is where we head into the 1990s another story happened, which is Taiwan's turn to build their own semiconductors expertise. I think one of the underappreciated facts was about the story of Morris Chang. He was the number three most senior executive from Texas Instruments. He would have become the CEO, but then of course it was given to someone else and then he came to Taiwan. Can you talk about the story of why and how did he come to work with the Taiwanese government and essentially built TSMC [Taiwan Semiconductor Manufacturing Company]?
Chris Miller: So, I think that Morris Chang is one of the most interesting entrepreneurs in the last hundred years, and someone who is dramatically underappreciated. He's a name that most people outside of Taiwan don't know about. I think that they ought to. He was born in China before the cultural revolution and spent his wartime years fleeing the Japanese armies in Hong Kong. Then the communist took power and so he fled again. He enrolled at Harvard University as the only Chinese student in his class. He gave up on his study of Shakespeare at Harvard and decided to do something more useful. He transferred to MIT and became involved in the semiconductor industry in the very earliest days, and was hired by Texas Instruments [or TI in short], which is one of the hot startups of its time in the late 1950s. He played a really fundamental role in learning and teaching the rest of them in Texas Instruments how to produce chips effectively. At that time, manufacturing chips was an extraordinarily complicated business and quality was very low.
He played a major role there. Over the course of his work at TI, he helped TI to establish offshore assembly and packaging facilities. One of the places that he urged Texas Instruments to look was Taiwan. He has a couple of ex-classmates from Taiwan whom he met during his PhD at Stanford. They told him the conditions and they were good. He'd never been to Taiwan before but visited in 1968 as part of TI's effort to scout out the best facilities in the region. The other competing location was Singapore, where TI eventually also opened up the facility, but they chose Taiwan first. From that, Chang developed a relationship with several of the key leaders in Taiwan and these leaders played a really major role in Taiwan's overall economic development strategy. People like K. T. Li [former economy minister of Taiwan] got to know him quite well. They visited him whenever they were in the United States. He visited them occasionally when he was in Taiwan.
When he was passed over for the CEO job at TI and what must be one of the great errors of business history in the 20th century, he was looking for something new to do and he'd already been a really high-level executive, so there weren't that many other jobs that would've appealed to him. The Taiwanese government came to him and said, "Would you like essentially a blank check to build up a chip industry from scratch in Taiwan? He said that sounded like an interesting idea. So he moved to Taiwan where he'd only previously been on business trips in the mid-1980s. In 1987, he founded TSMC with a brand-new business model in mind. Rather than integrated designing chips, TSMC would only manufacture chips for any customers who design them. This business model has, as you alluded to earlier, revolutionized the chip.
Bernard Leong: I can add an anecdote that why Singapore wasn't chosen and the Singapore government in the 1980s didn't put a lot of emphasis on research and development in the semiconductors industry, and that was why TI chose Taiwan. It's not just about the infrastructure side but the research and development piece. So coming back to the Taiwan story, then, how did TSMC disrupt the vertically integrated model from Intel with the foundry model?
Chris Miller: When Morris Chang founded TSMC, there really weren't very many companies that only designed chips. There was a small number, but they were. They were really quite small, and at the time, if you wanted to design chips and have someone else manufacture them, you had to go to a company that had its own in-house ship designers and essentially compete with the in-house team to get space in their production line. Most companies would have preferred their in-house design team. So you'd never know when you could get capacity because it depended on who had spare capacity at different times. They were always worried about whatever company you went to, stealing your designs because they were designing their own chips too. You always had horrible customer service because you were a small share of their production lines. Their production processes were designed for their in-house team rather than for outsiders. It was a really tough business model to only design chips and have someone else manufacture them. But Morris Chang realized that the economics of chip fabrication was such that the more chips you produce, the more economies of scale you could reap. There was a lot of rationale for having a smaller number of bigger chip manufacturers than existed at the time. He betted very heavily on the idea that if you built a facility that only manufactured shifts, you would have more design firms emerging over time and they could be your customers. That didn't exist when he started the company. But after he started TSMC and proved that they could produce with effective levels of technology at a reasonable cost, and provide excellent customer service, more companies began to enter the design-only space and the fabless space.
There is a trend later on where companies are divesting their manufacturing facilities and turning to design-only firms because they could count on TSMC to do the manufacturing for them and focus only on design. The margins were often higher as a result. It proved to be such an extraordinary business that today, TSMC is now the world's biggest chipmaker. It has reaped extraordinary economies of scale. From then on, they invest in what's now the most advanced process technology for logic chips.
Bernard Leong: Reaching economies of scale is one of the factors that made TSMC successful. But there are other factors, for example, their partnership with ASML and TSMC's internal talent bench. It has to execute chip production for major customers such as Nvidia.
Chris Miller: I think you're right that the ASML relationship has been crucial and it's existed now in a really deep partnership for three decades. One of the initial steps in the founding of TSMC was that TSMC licensed production technology from Phillips, the Dutch semiconductor producer, from which ASML was spun out. So there was a natural fit between ASML and TSMC from the earliest days. The relationship between TSMC's customers and TSMC is also really important because although TSMC serves a lot of customers, it's got a couple of key customers that have been with it for some time. Apple, their largest customer, is very important in terms of providing TSMC with guaranteed funding every year to know that it can invest very heavily.
Other major customers are Qualcomm, Nvidia and AMD, which have really given TSMC this scale. That's been necessary to expand the way it has over the past decade and a half. You couldn't have TSMC without the small number of pretty massive customers and so there is a deep relationship between TSMC and Silicon Valley as a result.
Bernard Leong: I still remember that famous video [which is TSMC's 30th-anniversary video featuring a panel of key movers and shakers in the semiconductors industry]. A lot of people don't know about it, and I'm going to repeat this for the third time. Here is Morris Chang who is about to retire and there were Jeff Williams (the COO of Apple), Steve Mollenkopf (the CEO of Qualcomm), Simon Sagars (CEO of ARM), and other movers and shakers on stage for two hours. I literally watched that entire video for two hours and I learned everything about the future of semiconductors in that two hours. That's the best YouTube video on the subject, and I really do not understand why nobody goes there and watches it because I think you can learn a lot about semiconductors
I want to come back to the discussion. If you look at the semiconductor supply chain today, it's extremely specialized. Taiwan produced the chips. The Netherlands built the ASML machines for the EUV [extreme ultraviolet] lithography that's required, and the United Kingdom (UK) has the ARM architecture, which provides the low-power chips that are inside most automobiles and phones from Android and Apple iPhones. Then you have Japan and Korea with their own expertise. Samsung is most well-known for its own solid-state drives and OLED screens. It does make it very difficult for any country to be able to do it all. You will know which country I'm alluding to. I'm talking about China here.
Chris Miller: It's impossible today for any country to do it all on their own. Even if you looked at the United States, which is still the biggest player in the supply chain, by far, it's still the case that the US can't do it all on its own. As you mentioned, the imports such as lithography equipment are from the Netherlands, chemicals and materials are from Japan, and the most advanced fabrication of processor chips is in Taiwan. So, so no country can do it alone and really no country is even close. The US is closest, but it's got those gaps that I just mentioned. Every other country is far less close. Taiwan, for example, is the most advanced producer of logic chips but uses machinery and software and chip designs and chemicals that are all imported. So Taiwan is far from self-sufficient and quite the opposite.
If you look at China, the reality is that China plays even less of a role in the supply chain than Taiwan or Korea or Japan, which are themselves very reliant on imported tools and designs and chemicals. Today, China is investing very heavily, it's trying very hard to play a bigger role over time. It's trying hard to produce, if not a completely domesticated supply chain. At least a de-Americanized supply chain is a more plausible goal. But the reality is that goal is still a very long way away. And so for now, China remain. Like the rest of the world, dependent on imports for key parts of its chip supply chain.
For China, that's of course a problem because as tensions with the US have escalated over the past decade, the US has imposed more restrictions on the types of equipment and software that China is able to import.
Bernard Leong: So we have come to the major part of this conversation: China. I think the early origins of semiconductors are meant for the military. The reason why the Russians couldn't replicate it was because of the supply chain and the fact that the US has turned semiconductors for civilian use and it spark a lot of technological innovation as such. Why is China so severely limited by this chokepoint technology?
Chris Miller: Well, the reality is that if you wanna make an advanced ship, you need equipment and software and chemicals from countries that are for China geopolitical competitors, if not rivals, whether it's Japan, whether it's the US, whether it's the Netherlands, which is defined China as a strategic competitor because of China's fraught relations with so many of its neighbours and so many technology leaders in the world. It's so bad that there is a whole set of restrictions on what China can import. The Chinese government realizes this limitation.
There's no easy way around it because of the immense specialization in the supply chain, and the immense complexity evolved in producing the tools and designs and software means that you can't just catch up in a couple of years. It means capital investment, which is the thing that China's really good at. On its own panacea, you need to invest a lot. The Chinese government has been investing a lot, but there's a lot more that you need to do. You need to develop firms that are deeply integrated into supply chains. They understand what the market needs, and learn from the best in the business. The incentives inside of China's system are no longer allowed to integrate with supply chains, but actually the opposite, to pursue more domestic development and more self-sufficiency. And that's exactly the opposite of what Samsung did when it was catching up on what TSMC has done.
Ironically, the more China tries to have this catch-up strategy play, the dominant feature in its semiconductor industry, the more it's actually diverging from what TSMC or Samsung or other firms did to catch up, and that's a paradox that I think driven as much by China's domestic politics as it is by the chip industry. I suspect that most of the players in China's private sector, specifically the tech industry understand this is not the best strategy for technological purposes, but that's the strategy they have given the political context.
Bernard Leong: When I talk to my friends and then when they talk about why the US is losing against China, I will always draw them back to the lessons of Japan in the eighties, and then Korea in the nineties. The Chinese situation is different from Japan and Korea. First, it's not a democracy, and second, China starting to decouple from the supply chain with its own version of the GPS navigation system: beidou. They have a 1.4 billion population, which is the largest market in the world, which means they have their own domestic market as compared to say, Japan and Korea, which both do not have a very big population. The question now would become, even with all the blocking that's coming from the U, will it still be able to replicate the semiconductor chain? I think there's YMTC and there's SMIC from China, but I think they're still playing the catching-up game. From just looking at today's world, do you think they can actually end up replicating the supply chain anytime soon?
Chris Miller: I think the answer is not anytime soon but maybe in a decade's time they will catch up on key parameters, but it's going to be a long way away, I think for a couple of reasons. First, if you compare China today to Japan and Korea in the eighties and nineties respectively, there have been some similarities for a long time. If you look at specific firms, you can find a lot of comparison points. But if you look at the political context, it's really quite different. The US right now has made it clear that it doesn't want to see China catch up technologically. Whereas the US was never concerned about Korea's technological advances, and it was only concerned really in very specific instances for a short period of time about Japanese technological advances. Even then, there was extraordinary cooperation scientifically between the US and Japan that is still ongoing. Whereas now there is a sort of selective decoupling underway between the US and Chinese tech sectors. It's not just the US partly due to US regulations, but also partly due to other countries' perception of their interests. You know, Japan passed this series of new regulations and legislation that is going to reduce the science and technology cooperation with China. So there is a lot less cooperation that we're going to see going forward between China and other key tech powers than we saw between the US and Japan or the US and Korea. So that's a big difference.
The second big difference is that, on the one hand, it's true that China's got this vast domestic market, which Japan and especially Korea didn't. On the other hand, I don't know if that's a good thing because the Chinese domestic market is still a lot smaller than the world market. If the Chinese government and firms think that they're going to be able to sell primarily or solely to the domestic market and thereby give up the world market, they're gonna end up with a market that is 20% of the size of the world market. If you've gotten one supply chain selling to a market that's four times as large as the Chinese supply chain, well, the economy's scale that we've talked about, benefits the larger supply chain. I wouldn't want to be in a position of focusing solely on China's domestic market. Big though it is, I'd wanna focus on the world market, and so I think that's a risk to China as well, that it might make it seem to Chinese policymakers or some corporate leaders, that they can only focus domestically. But the reality is that I don't think that they have a great strategy given just the importance of these massive economies of scale that we've discussed.
Bernard Leong: What about the global south, a market that China can tap into? The global south has to contend with two systems now. Living in Southeast Asia, I don't have to deal with just the US technology companies. I also have to deal with Chinese technology companies on the same end and do business with either of them. We are becoming like a router of two systems because of the decoupling.
Chris Miller: No, I think that's true, but I think it's also true that so long as the economy's a scale that we've discussed persists, my sense is that TSMC is going to stay far ahead and they'll be able to provide better products at lower prices than competitors in China. There are other markets where that's not the case. Chinese firms and Taiwanese firms will produce comparable goods or Chinese firms and other countries' firms will produce comparable goods. But in semiconductors, I don't think that's going to be the case. I think it'll be very hard for China to compete in the markets of Southeast Asia, for example, if the chips it's producing have orders of magnitude fewer transistors on them. If you're three generations behind semiconductor production because of Moore's law, you're miles behind because you're exponentially behind each generation. So long as that dynamic persists, so long as economies of scale produce better technology, and so long as better technology is exponentially better because of Moore's law, I think China faces real difficulties if it focuses largely on its domestic market just because the dollars are much more focused in terms of sales and in the rest of the global economy. So it's gonna be difficult to win primarily focusing domestically unless it leaps forward [beyond the present technology].
Bernard Leong: Like China taking on something like a quantum computer if it developed it correctly. Of course that in today's world, it is not possible from a realistic point of view.
Chris Miller: Correct.
Bernard Leong: I want to talk about this question and I am gonna take history back to the 1940s, just around the World War II timing. Given that you're a historian, if we start from the precedent down into Graham Ellison's "Destined For War" on the Thucydides trap, how Japan ended up attacking the United States on Pearl Harbor was because of the tough sanctions of oil as the final straw. If I pick different words, I replace China with Japan and oil with semiconductors. What's the livelihood of China attacking Taiwan with the recent conclusion of the 20th party congress?
Chris Miller: First off, we should recognize that when Chinese leaders think about Taiwan, they don't primarily think about semiconductors. The CCCP leadership has wanted to assert control over Taiwan since 1949 well before the semiconductor was first invented. If you listen to the language that they used in the party Congress, for example, it's not somebody that could be focused, nor is it different than what we heard a decade ago. So we shouldn't say that because Taiwan is important because of the chip that produces. Therefore, China wants control over Taiwan because of the chips that produce, because China has always wanted control over Taiwan since 1949. So for people like you and me who are interested in semiconductors, I think it's easy to overestimate the importance of semiconductors.
But I think you're right to say that there is some danger that the Chinese leadership believes that the new controls will be so damaging to its position in the long run, that they're better off acting in the short run. Now, I think the comparison to Japan in 1941 has some similarities and some differences. The differences are that oil was a lot more important to Japan's economy writ large than cutting-edge chips and chip-making equipment are today. You know, the US hasn't banned the sale of all chips to China. Quite the contrary, almost all chips that could be sold to China last month can be sold to China this month.
It's just a small segment of the ultra cutting-edge GPU chips for data centres. So that's 1% of the chips that are going into China are being cut off, whereas the oil embargo on Japan before World War II was far larger. So that's different. The second thing that's different is that it's, it was clear to Japanese leaders right after the oil embargo that this was going to be devastating to them both in the short run and the long run. So they had to act. It's not immediately clear to me that Chinese leaders believe that this will be so problematic to them because they may believe that they can get around it or they can domestically produce their way out of it. And while I think that's probably the wrong judgment, There's uncertainty around that. If you are sitting in Chinese leadership meetings, you could easily tell yourself that this will be damaging in a couple of years but we'll find our way around it. That introduces enough uncertainty where for China, the optimal strategy is to wait and to try to see if you can get your way around it.
Maybe you can, maybe you can't. But that compared to an attack in Taiwan could produce a disastrous war for China. Is there uncertainty about that? Yes. Am I more worried about a Chinese attack on Taiwan than I was several years ago? No doubt. As the military balance continues to swing in China's favour as it has over the past several decades, does my worry increase? I think it is inevitable. Do I think that the move to export control certain chips and chip-making technology is the same as an oil embargo in Japan? No, it is fundamentally different because we're not saying "no" to chips into China. We're saying a tiny share of the most advanced chips in chip-making tools can't go into China. That's different from saying an economy can't have any oil.
Bernard Leong: The chips that we are talking about limit their military capability development. I appreciate getting a historian's point of view on this. So I ask this last question, and I think it's a perspective question. If we fast forward to the next decade, what will be the story of semiconductors going ahead?
Chris Miller: We've got a pretty clear pathway through 2030 of advances in chips that will keep Moore's law going at least for the end of the decade. We're gonna see semiconductors applied to even more types of devices and the two big growth markets I think, will be in autos, which have always had chips inside of them for a long time. But we'll have more and more advanced chips inside of them. The second growth market will be for high-performance computing and artificial intelligence (AI) in data centres primarily. That is a market that has grown a lot and will grow dramatically over the next couple of years. As we see more demands for AI across different parts of the economy, that will create massive demand for more chips.
In terms of the industry's structure, one of the big trends that are emerging right now and will continue is that more and more big companies are designing their own chips. Today it's Google and Amazon, but I think we're gonna see more companies deciding to design their own chips because they can get more performance out of chips that are optimized to the specific processes they want to run on them. This will be an ongoing trend that will benefit foundries like TSMC and others. Companies that are trying to set up foundries because of Amazon and Google, I don't think ever going to run their manufacturing facilities, but they will keep designing their own chips. They'll need to outsource the fabrication of these chips to someone else. This is gonna be another industry trend that's already started but will continue through the decade.
Bernard Leong: What about the CHIPS Act with now the 52 billion that is given to the US industry?
Chris Miller: Will Intel get a comeback as a result? Well, this is a big question. I think whether or not Intel comes back won't depend on the CHIPS Act because Intel's challenge has not been a lack of money. Their challenge has been getting the technology and business model to work. There's a turnaround underway right now at Intel. Whether they can succeed in turning it around remains unclear on the comparatively small amount of money that the CHIPS Act will give to Intel, as the funding will be divided between a number of different companies. That alone won't make the difference in Intel. I think that the much more important [questions] are: Can they get their business model to work? Can they get their technology to catch up or at least get closer to where TSMC is?
Bernard Leong: Thanks for coming on the show and sharing the story of Asia with the rise of semiconductors through your book, which I highly recommend my audience to go and read it. In fact, the technology audience has been very curious about your book, and that's why I thought I should do this interview with you. In closing, I've read two very quick questions. My number one question. Any recommendations that have inspired you recently?
Chris Miller: Well, the best book that I recently read is a book by a historian at American University called Joseph Torigian's "Prestige, Manipulation, and Coercion: Elite Power Struggles in the Soviet Union and China after Stalin and Mao". He's got an extraordinarily deep analysis of the post-Mao era succession in China based on extraordinary archival works. And so it's a minute-by-minute account of the succession struggle after Mao Tse Dong's tenure, which I highly recommend.
Bernard: Thank you for the recommendation. I'm going to check it out. How did my audience find you?
Bernard Leong: You can find us on any podcast platform and definitely tweet to us if you have any feedback or give us a five-star rating on Apple Podcast. Chris, many thanks for coming on the show and I look forward to speaking to you again sometime in the future.
Chris Miller: Great. Thanks for having me.