The Institute of Geoeconomics (Kazuto Suzuki, Director, Institute of Geoeconomics) is pleased to announce that recorded audio of the Geoeconomic Agenda, “Semiconductors & the Economy of the Future with Chris Miller,” and its transcript are now available. Geoeconomic Agenda is a new Pacific-centric, globally-minded podcast that investigates the connections between economics, geopolitics, business, and society. Hosted by Paul Nadeau, Geoeconomic Agenda will feature interviews with guests from across the Asia-Pacific region and beyond, from the policy world, academia, business, and more.

Transcript

June 30, 2023 (auto-generated by otter.ai)
Last updated: June 30, 2023

Paul Nadeau
Today I’m sitting down with Professor Chris Miller, the author of Chip War, and an associate professor of history at The Fletcher School of Law and Diplomacy, the best international affairs graduate school, which I say completely without bias as an alumni of that school. Professor Miller, thank you so much for joining us.

Chris Miller
Thank you for having me.

Paul Nadeau
So, I like to tell the story about how when I went to the library to look for your book, I typed the “Chip War” into the search bar. And sure enough, the Chip War was there. Except it was a 1989 book by Fred Warshofsky. And this book, the 1989 version, was about the chip war with Japan. Now I’ve read both books. So I have an idea myself of, you know, the difference? But what makes this chip war different from the 1989 version with Japan?

Chris Miller
Well, I think when the war, the trade disputes between the US and Japan over semiconductors, war was only in a metaphorical sense. The disputes were about market share, they were about allegations of dumping, it was a purely trade question. Whereas today, when you look at the semiconductor dispute between China and the US, Japan, and a couple other countries, the word “war” is not just a metaphorical sense because the security concerns are at the absolute center of why this is happening. They’re at the center of why China has for the last decade been trying to cut itself free from reliance on chips from the US, Japan, and Taiwan. And they’re also at the core of what is driving US, Japanese and policy-making other countries to try to cut off China’s access to advanced semiconductors because they believe that if China develops advanced chips, they will inevitably deploy them to defense and intelligence systems.

Paul Nadeau
So the threat this time is military. Primarily, this is not a question of economics or trade, there is a genuine military threat that is driving US concerns. Is that correct?

Chris Miller
Not just US concerns. I think too often this is described as the US-China dynamic. But we sit here in Tokyo, which is, I think, a good place to reflect on the fact that it’s not just a bilateral question. And it’s, I think, inaccurate to even see this as a US-driven question, because in a lot of ways, the US has been learning from others of its allies and partners about the importance of this issue and about the changes in China’s approach to this issue.

Paul Nadeau
So one of the big takeaways that you get from your book, you’re talking about the ecosystem of the semiconductor industry, is the really remarkable amount of concentration that’s in Taiwan. I think it’s 92% of advanced chip made semiconductor manufacturing occurs in Taiwan, that is more or less accurate. So that’s a really significant choke point. If you’re, you know, trying to game out the risks of a Taiwan invasion, a hypothetical Taiwan invasion, or at the very least a, you know, coercive reunification of the island. What would that do to the world’s semiconductor industry into the rest of the industries that depend on semiconductors?

Chris Miller
If there were a disruption to Taiwan’s chip exports, the result would be a disruption to global manufacturing of a scale we haven’t seen since the Great Depression. That big without really any doubt. In the last couple of years during the chip shortage, we saw hundreds of billions of dollars of disruption to the auto industry alone because car companies couldn’t sell cars they plan to sell because they couldn’t get all the chips they needed. But the quote unquote chip shortage of 2021, and 2022 occurred when the world produced more chips than prior years, there was a almost double digit rate of increase in chip production between 2021 and 2020. And there was a double digit rate of increase in chip production between 2021 and 2022. It was increasing to production, but demand grew faster, so we had a shortage that caused 100 million dollars in damage. Imagine what the impact would be if the world wants access to almost all of its most advanced processor chips, and a huge share of the less advanced processor chips which are critical and everything from dishwashers to automobiles.

Paul Nadeau
So this wouldn’t just be a bummer for, you know, anyone buying semiconductors, this would be felt throughout the entire manufacturing system. I remember, sorry, before you said there’s something like 3000 chips in a car nowadays. So this isn’t just, you know, computers or something, this is everything.

Chris Miller
Today basically everything with an on off switch, except for the simplest of light bulbs has at least one, and often dozens or hundreds of semiconductors inside. And the complexity of supply chains today is such that, if you take a device with a dozen chips inside, it will often have a dozen chips produced by multiple different companies from multiple different countries. And you need all of those chips to make that one device work. So even if just one of those semiconductors is from Taiwan, if you lose access to it, you can’t produce that device until you find an alternative source of supply. But in a crisis where we lost access to chips made in Taiwan because of Chinese aggression, there would be no alternative source of supply, there’d be a global deficit.

Paul Nadeau
So even beyond, you know, the risk of invasion or coercion, there’s just a simple fact, like you pointed out, if there’s any kind of disruption to anything going out of the streets of, excuse me anything out of the island of Taiwan, that’s going to be, you know, felt throughout the system. So how much of that concentration would you say is driving these initiatives towards supply chain relocation or the subsidies like we’ve seen in the United States through the Chip Act and the IRA, how much of this is motivating all of that?

Chris Miller
I think there are two factors that are driving these new policies or industry. One is, the concentration in Taiwan and the growing Chinese threat to Taiwan. And the second is the fact that China has, over the last decade, been pouring tens of billions of dollars a year into its own chip industry trying to catch up to the cutting edge. And it’s made progress in certain spheres, less progress in those spheres. But China has in aggregate made some strides. And until the restrictive measures began to be put in place, its rate of progress was, I think, increasing. And so a key driver of the new restrictions and new policies has been to stop China from catching up.

Paul Nadeau
So one of the points tangential to that, one of the points that you’ve made consistently is the questions that you have about whether China can sustain these kinds of investments over a long period of time. Can you elaborate on that a little bit for our listeners?

Chris Miller
If you look at the scale of investment that China is putting into the chip industry, it’s really unprecedented in the history of China’s industrial policy. We’re used to a Chinese state that is very active in funding certain businesses and directing investment. But this scale of support for the chip industry goes far beyond the scale of support for any other segment of the economy that’s operating at the technological frontier, because building something up is not like building apartment blocks or building textile factories. Every country can build apartment blocks. And so the question is not can you do it, it’s just will you spend the money to do it. With advanced facilities, there’s huge uncertainties where China can actually undertake the activities and spending forward. So there’s a lot more risk associated with the dollar spend. And so kind of spending a lot of money. It’s unclear what if any of this money will lead to productive results. And so there is a lot of financial risk that has already built up and is increasingly building up every single year in the fact that China is spending these large sums on this chip industry.

Paul Nadeau
So I just found out that your day job, or at least one of your past day jobs is as an economic historian of the Soviet Union. I don’t know if that’s the exact title, but you’ve at least written on the economic history of the Soviet Union in the 1980s. So when you talk about, you know, dumping money, few into a futile project. To me, that sounds a lot like the goal of Reagan’s defense spending in the 1980s, is try to encourage the Soviets to outspend out, excuse me, encourage the Soviets to overspend, overspend, overspend and eventually. crack under the weight of that, that excessive spending. So whether or not it worked for the Soviet Union, quote, unquote, is a discussion for another day. But, you know, given that template, if you want to call it that, why not just let China dump money into a losing effort and let them spend themselves into oblivion until they can’t anymore. Why do we have to take that extra step of export controls and subsidies and all the expense and the friction that comes with that?

Chris Miller
Well, I would say two things. I think first, the analogy with Reagan strategy in the Soviet Union is an interesting one. And it’s certainly true that under Xi, the Chinese policy making apparatus has been increasingly making more of the type of mistakes that the Soviet Union made moving away from a market economy towards a state-driven economy. But I think we shouldn’t underestimate the amount of spending China has ahead of it. And to predict that China only has a couple of years of capacity. I think people underestimate their physical capabilities, there’s uncertainty. But yeah, I would rather overestimate than underestimate their physical capability. And the second aspect is that the Chinese ship industry was struggling to advance on its own. But it was succeeding in advancing where it had access to Western technology. And what we saw from the early 2010s, up until recent years, was a very accurate campaign of buying foreign companies, partnering taxes foreign technology of succeeding in buying many of the most advanced machine tools of a series of somewhat dubious joint ventures with foreign firms designed to extract technology. And it’s only because all of these activities have been cut down to two export controls due to investment restrictions, that China’s faced a lot more difficulty in catching up.

Paul Nadeau
So relatedly, is we’ve talked before, it’s easy to spend money, it’s easy to throw money at a problem, China’s doing it the US, Japan, everyone else is doing it to a degree. And that can get you a certain point down the road. But you can’t spend your way into innovation. It’ll help, it can help at least. But it’s not usually the key that unlocks the door. So can you talk a little bit about what the United States or China for that matter, anyone in this space needs to do beyond just simply spending lots of money and providing lots of subsidies and tax credits, and so on? What’s the ecosystem around all this money?

Chris Miller
I think what you find historically is that the countries, the companies that have done the best in this competition have integrated themselves into the international supply chain, learning from their component suppliers, learning from the toolmakers, learning from their customers. Because in the chip industry, the supply chain is broken into many different pieces. And no matter where you sit in a supply chain, you’ve got to understand the expertise of all the other parts to make the right products at the most effective matter. And so if you look at a company like TSMC, we can discuss extraordinary, success story in our industry. And it succeeded because it was so deeply integrated, learning from the toolmakers, learning from its customers, and that was a key driver of TSMC success. I think for Chinese firms, this is a real dilemma going forward because they’re being systematically cut out of the leading edge portions of supply chains, and so they can’t learn from other companies in the supply chain anymore.

Paul Nadeau
So what kind of specific innovations does that hold them back from or what sort of advantages, innovation wise, would that give economies like the United States or Japan, and so on, so forth.

Chris Miller
I think that the challenge of making is not to make one innovation, but to make a whole series of innovations constantly, because you’ve got to be keeping up with Moore’s law, which says there’ll be twice as many transistors per chip every two years. And that is a rate of growth that is unparalleled in any other segment of the economy. And, and to make that possible, is not just about changing one thing every two years, it’s about making everything vastly better. So your chemicals have to become better, your transistor design has to become more refined, your lithography tools to be used more accurately. And to have this broad array of improvements happening on a constant basis is extraordinarily difficult. You only do it through this type of collaboration and collective learning that I’ve described. And one of the reasons why there are very few firms that emerged out of the blue in the semiconductor industry is because it’s very difficult to emerge out of the blue when you already have an understanding of the rest of the supply chain. Because if you have a tool that is not perfectly aligned with the materials that tool requires, your tool is not valuable. So you’ve got to have a relationship with the materials suppliers and with the fabrication companies so that you can understand how your specific expertise fits into whatever else we’re doing.

Paul Nadeau
You know, maybe to go off of that, if you could, you know, drill down into a specific example, one of my favorite parts of the book was your discussion of ASML’s, lithography machines, and that absolutely incredible amount of I don’t even know how to describe it. It’s like detail a, we’re talking about nanometers. The complexity of these machines are absolutely incredible. Can you just describe like one of ASML’s machines for our listeners to give them a sense of, you know, obviously, I encourage them to read the book, needless to say, but until they do, can you just give them a quick preview of what these kinds of machines are like? What goes into them?

Chris Miller
So these tools used for extreme ultraviolet, they are the most complex tools humans have ever made, they have hundreds of 1000s of components in them, including the flattest mirrors ever made, one of the most powerful lasers ever deployed, and commercial device and an explosion happening constantly at many times hotter than the surface of the sun. It’s mine, it’s unbelievable. It’s unbelievable that they work once. But what’s even more unbelievable is that they work every single day in high volume manufacturing. And if you think what does it take to get a machine with hundreds of, thousands of components to work almost all the time, each one of those components has to basically never break? Because if the average lifetime of your component is a year, your machine never works.

Paul Nadeau
Yeah, exactly. When you make it an investment like that into a machine – that’s that complex – you’d expect to have a return that makes it pay off. And A, that’s an illustration of how much, you know, money and expertise has to go into something like this. But B, that also shows you that the returns are there, the economics are there that justify these massive capital expenditures? Is that fair to say?

Chris Miller
Well, these machines sell for $150 million a piece. Never heard of something like that, before. the most expensive machine tools ever mass produced, the next generation of these tools will sell for around 350 million. Can’t even imagine that. Which is why building a new shipping facility cost $20 billion. Because you’re not just one but multiple of these tools, and many other types of tools as well.

Paul Nadeau
So they go back to the ecosystem, then, you know. It’s sort of a joke that I have that I am not technologically adept, I do not imagine that I could build one of these myself, that you would need, you know, people that have extensive training and extensive expertise. And these people do not come out of thin air. Needless to say, these are not, you know, I’m not going to speak to your engineering skills. But I doubt you could probably do that as well, even though you have very specific knowledge of these things. So where did these people come from? How do you cultivate this kind of talent and draw upon it, you can create machines that will sell for $350 million?

Chris Miller
You know, I think that the easy part of the talent creation process is having good universities, and certainly in universities that will train people in optics and physics and neuroscience. But actually, when you get to manufacturing, the real expertise is not in the physics or the material science, the expertise is in the know-how, and how to actually manufacture these machines. Because there’s, there’s no physics textbook that will tell you how these tools are made. There’s no textbook at all. It’s just in the minds of the couple hundred people who have spent the last three decades actually making them. And that’s why copying and replicating these types of tools is so difficult because there’s no one file you can copy and paste. There’s just lots of really unique know-how that’s in the minds of the engineers. And it’s proven extraordinarily hard to replicate.

Paul Nadeau
So with those expenditures in mind, you know, the the amount of expertise and the amount of just simple know-how, I can start to see how a company like TSMC might end up accruing 92% of the market share because once you start down this road, once you make the expenditures, then it compounds. And you know, 20, 30 years later, you know going back to Fred Warshofsky’s book when he’s talking about TSMC. You can see how it ends up with 92% of the market share. How do you catch up? How do you diversify away so that, you know, not only are you insulated from something like, you know, the worst case scenario of a violent invasion, but also just any kind of disruption?

Chris Miller
I think if you look historically at the industry, what you find is that companies with dominant positions have been disrupted when technology shifts, when technology shifts in a way that the established company isn’t prepared for, and which provides benefits to new business models. So for example, throughout the first three decades, chip industry’s existence, almost all chips were both designed and manufactured by the same companies. And then in 1987, TSMC was founded around the idea that it would be better only to manufacture chips designed by multiple customers and manufactured by TSMC. And that became more viable because chip-making got more and more and more complex, more and more expensive, as a TSMC could specialize and make life much easier for all of its customers who are worried about manufacturing. And so I think your question to ask right now is, well, are there inflections on the horizon that could create a big shift, in the business model, whether it’s new architectures for semiconductors, we’ve seen Nvidia, for example, become a very successful, most successful company by pioneering this new architecture that’s perfectly honed for AI. But also shifting the manufacturing process could be underway. There’s a lot more focus today, for example, on how do you package chips together once you’ve manufactured them, and that’s an increasingly important part of the value add that companies provide.

Paul Nadeau
So you mentioned inflection points, why don’t we wrap up by you know, discussing the future? To the extent that you’re able, I know you’re a historian, so you’re not usually look, you’re more backward looking and forward looking a lot of the time, but what inflection points, what developments, what changes, evolutions do you see happening in the next five to ten years and beyond?

Chris Miller
I think one, one trend we’ve seen very clearly already is that more and more companies are designing specialized chips for themselves. Because it’s become clear over the last couple of years that if you’re someone who uses a lot of computing power, like Google or Amazon, you can have chips that are much more efficient if they’re specifically tuned to the type of workloads you’re undertaking. And so all the big tech companies have been designing their own chips, manufacturing is outsource they design themselves. To get that efficiency from specialized chips. That’s one factor. The second has been, focusing on new types of chip architectures for artificial intelligence. We’ve discussed in a video which is producing chips used for training AI systems, or in the very early stages, I think of seeing new chips that are focused on inference on the edge of networks. So in autos, or in consumer devices, this is a huge growth market for both establishing new selector firms. The third facet is packaging is as I mentioned, traditionally, most of the dollar value spent, most of the complex technology was in the actual manufacturer of chips, that’s still really hard and like blocks, but there’s now just as much focused on how do you put different types of chips together, which is important for the data transfer speed, the interconnect speed for heat dissipation, for packaging, alternative capabilities, like photonics, so optical connections onto a existing chip. There’s lots of new things you can do with packaging and companies are spending just as much time focusing on packaging and manufacturing itself.

Paul Nadeau
This has been fascinating. We could go on for hours probably talking about this, but I want to thank you very much for your time professor Chris Miller, this has been great. Come again anytime. Thank you.

(Disclaimer)
The views expressed in the IOG Geoeconomic Insights do not necessarily reflect those of IOG or any other organizations to which the speakers belong.