Waging the High-Stakes “Chip War”: My Long-Read Q&A with Chris Miller

By James Pethokoukis and Chris Miller (Miller is an Associate Professor of International History at The Fletcher School, and author of Chip War: The Fight for the World’s Most Critical Technology. He is also a non-resident fellow at the American Enterprise Institute.)

Computer chips are the driving force behind everything from smartphones and cars to military defense systems and artificial intelligence. Not only are they the essential element of modern digital infrastructure, they are a critical element in the global balance of power.

Taiwan is home to the most advanced and productive chip plants in the world, precariously placing the technology between Communist China and the democratic West. In today’s geopolitical landscape, control over semiconductor supply chains is more than just an economic issue; it’s a matter of national security. Today on Political Economy, I’m talking with Chris Miller, author of Chip War: The Fight for the World’s Most Critical Technology.

Miller is a nonresident senior fellow here at AEI, where his research focuses on Russian foreign policy, politics, economics, as well as Eurasian geopolitics and the geopolitics of technology. He is an assistant professor of international history and co-director of the Russia and Eurasia program at the Fletcher School of Law and Diplomacy at Tufts University. He is also the director of the Eurasia program at the Foreign Policy Research Institute.

What follows is a lightly edited transcript of our conversation. You can download the episode here, and don’t forget to subscribe to my podcast on iTunes or Stitcher. Tell your friends, leave a review.

Pethokoukis: Chris, welcome to the podcast.

Miller: Thank you for having me.

If China were on the path that many of us hoped it would be 20 years ago, maybe moving toward being a South Korea, maybe a Japan, being a slowly more democratic member of the community of nations, would we care where our chips were manufactured? Would we care that so many chips were manufactured off the coast of China?

Well, I think we’d care a lot less. The reason why we’re worried right now is that today, when it comes to advanced chips, around 90 percent of them are made in one island, Taiwan, which is an extraordinary concentration, both from the perspective of earthquake risk, but, more importantly, from geopolitical risk, and that’s why there’s been so much public focus the last couple of years on this question of: Can you diversify the manufacturing base of where chips are made?

What happens if China retakes Taiwan in 2027, all of a sudden there are Chinese troops on the factory floors? What is the economic and geopolitical impact then? So all our good efforts fail, and these high- end chips are still being made in Taiwan, but now Taiwan is under the control of China. What does that world look like?

I think it’s hard to imagine Taiwan falling into Chinese hands without some sort of conflict, and the reality is that, to make chips, which are manufactured in Taiwan, you need inputs from Japan, from the United States, from Europe, from a number of other Western countries, without which chip factories don’t operate. And so the moment there’s any sort of conflict around Taiwan, chip supply chains would get gummed up, and the production of chips will collapse.

So I think that the two scenarios we have to consider are the status quo, in which Taiwan’s still able to export chips, and the alternative, where China tries something, supply chains freeze up, and the entire world faces vast deficits of chips. Because the scenario in which China is able to take the facilities without actually destroying the supply chains that keep them functioning, that seems to me a highly unlikely scenario, and I think the United States would have a lot of ways in which it could make that basically impossible to continue operation.

Wouldn’t that be a temporary state of affairs? The world needs chips, they’re made there, wouldn’t at some point we just acquiesce to the new status quo and all those inputs start rolling again into Taiwan?

You’d need the acquiescence of the US, of Japan, of European countries, but also, most importantly, of Taiwan, of the workers in those facilities, many of whom don’t want to be part of China, some of whom have been educated in the United States, some of whom have family in the United States. So the theory that the workers, who are the absolute essential component of these factories, would just stick around under a Chinese occupation regime seems to me highly, highly implausible. And you don’t just need most of them to stick around, you need all of them, because every segment of the chip-making process is so complex that it’s not just something where you can take people off the street and throw them in the factory. They need years of training using the most complex machinery humans have ever made. So any sort of Chinese operation would throw extraordinary disruptions into the industry and cause many, if not most, of the chip-making capabilities of Taiwan to freeze up.

I have one more “what if” question: Would there be sabotage? Are there Taiwanese plans like, “If we can’t have them, nobody has them”?

I think you’d have to assume that’s a real possibility, both on the hand of the Taiwanese government, on the hand of the US government, you had multiple senior US government officials—out of office, but former officials—say that the US should destroy the facilities if they were falling into Chinese hands, but also the workers themselves, because if you had a Chinese manager parachuted in, they wouldn’t even know if a production process was being tweaked in a way that caused chips to malfunction, and if they found that there was a malfunction at the end of the process, they couldn’t identify where in the process it was happening. These supply chains are so delicate, the manufacturing processes are so precise, that the idea you’re going to have political overloads arrive in the factory and run them without any disruptions is just highly implausible.

Why is it important that the most advanced chips are made in the United States, rather than Japan, or South Korea, or, I don’t know, the Netherlands, or something? The domestic aspect, why is that so important? I understand why we wouldn’t want them all made in an earthquake-prone island right next to a country that may invade it, but why not these other places? Why is the domestic part so important?

I think if advanced ships were made in Canada, I probably wouldn’t have that many concerns, but if you zoom out and look around the world and ask, “Where are the places where it’s plausibly efficient to produce high-end chips?” the answer is, there aren’t that many. And the reason is that manufacturing chips is not like manufacturing many other types of goods. Labor costs matter a bit, but not that much. What matters is having the specialized workers and the dense networks of suppliers of chemicals and materials that are needed in manufacturing. And so it’s not as though low-cost locations for most manufacturing are actually low-cost in the chip industry. If you go to a typical country in Southeast Asia, you can get low wages, but if you don’t have the suppliers or the trained workers, your costs are going to be high, not low. The US is actually one of the relatively more attractive places to produce advanced chips, alongside Korea and Taiwan, alongside a couple countries in Europe, and perhaps Japan, where it’s actually plausible to make high-end chip-making facilities. And then I think we can debate, are we more or less comfortable in Germany, or Japan, or Korea, or the United States, but there’s a short list of countries where you could plausibly set up a $20 or $40 billion chip-making plant and have it run efficiently, and it’s not primarily based on labor costs.

Chip War was published in October 2022. The next month, OpenAI rolled out ChatGPT, and whatever the importance of high-end chips were in October 2022, they seem a lot more important, from my perspective, now that this chip competition is now extended into cloud computing, and these data centers. Does this change your thesis at all? Does it advance your thesis? It seems, from a layman’s perspective, that chips are wildly more important now than they were a year and a half ago.

Yeah, I think that’s right, and that’s why we’ve seen all of the biggest tech companies, all of the leading AI startups, pour not just billions, but actually tens of billions of dollars into buying high-end AI chips from companies like Nvidia, which is the most important supplier of them. And they’re doing this because, what AI researchers have learned over the past couple of years, is that if you want a better AI system, you need an AI system trained on more, and more, and more data. And if you want to train systems on larger quantities of data, you need more processing power to do it, which means the more advanced your chip, the more advanced your AI system. It’s a pretty simple equation, it’s held up over the past several decades, and the leading figures in the tech world, from Sam Altman at OpenAI, to CEOs of all the big tech companies, they’re all making essentially the same bet, that if they want better AI systems, they need bigger AI systems, and that means lots and lots of high-powered chips.

Given this recognition by the private sector that these are important—tons of investment—does it make the government component less important than maybe what you thought it would be in October 2022?

I think the challenge in the chip industry is that governments in different countries have always been putting their thumbs on the scales in terms of where chips are being built, and so the industry, left to its own devices, would just pursue the most efficient route possible, but the reality is that the map of chip production today is a map that has been influenced by governments’ decisions in the past. And so today, because of industrial policy in Taiwan over several decades, Taiwan is the best place in the world to make chips. They’ve got the densest network of suppliers, they’ve got the deepest pool of labor, and so, left to own devices, the chip industry and the AI industry would buy chips from Taiwan far, far into the future. The challenge is that, of course, if you look at the geopolitical map, that’s a problem for the reasons that we’ve discussed.

And so I think governments are simultaneously putting their fingers in the scales and shaping where investments take place, but they’re also having to interact with companies that are investing at a scale that is, at times, even larger than what governments can afford. And to me, I think the CHIPS Act in the US is a case in point. The US government is going to invest $39 billion in subsidizing chip making in the US. That sounds like a lot of money until you look at the tens of billions of dollars that Big Tech firms in the US are going to spend buying AI processors just this year. The capital expenditure is so large in the private sector that even significant public-sector programs look small by comparison.

Given the geological and geopolitical precariousness of Taiwan, why wouldn’t other companies say, “Boy, this is crazy that we’re so dependent on this one place.” Why wasn’t there already pressure for more geographic diversity with chip making? I understand why, on a pure market analysis: They can make them, they can make them super efficient, they can do it like nobody else, but given those other factors, why wouldn’t, at some point, we see that diversification without government trying to encourage it?

I think there are two reasons why, or two parts of the answer. I think Part One is that, in the chip industry, the economies of scale are just fast, which means that if most of the industry is in Taiwan, and you’re going to try to start a small outpost in a different region, you’re not going to benefit from economies of scale until you manage to scale up over a decade, which is the time horizon. And in the interim, you’re going to be facing higher costs for every chip you produce, and that’s just the reality. So there are real economic factors that disincentivize diversification.

I think the second factor is that, for the Big Tech companies that are the consumers of many of the most valuable chips that are produced in Taiwan, they benefit from a sort of moral hazard. They assume that, in a geopolitical crisis, they’ll get bailed out somehow by the government, because, in wartime, government steps in and helps private companies, and I think the boards of directors think that their shareholders will punish them for losing margin by shifting the higher-cost manufacturing outside of Taiwan, but their shareholders won’t punish them if they, like everyone else, find themselves in a very difficult situation amid a geopolitical crisis because they can say, “Who could have predicted? That wasn’t our job to predict. We’re in the business of business, not in the business of geopolitics.” And so I think the entire tech sector in the US has—I think irresponsibly, but nevertheless—has washed its hands of responsibility for their supply chains on the hope that, if something goes wrong, government will be standing there to solve the problem for them.

I think the Biden’s administration’s forecast, or their metric, is that 20 percent of advanced chips would be made in the US by decade’s end, by 2030. Is that a good metric of success? Would that be your metric of success?

I think that sounds ballpark about right. That’s where most private-sector forecasters are as well, and that is testament to the deep structural factors that make Taiwan such an attractive place to make chips, that even after almost $50 billion in support, the US will have gone from zero to 20 percent market share. Now, I think that’s meaningful diversification, and, in a crisis, that would give us room for maneuver that we did not previously have, but I think it also speaks to the challenges of diversification in an industry with such vast economies to scale.

Should we be spending more?

The mental math that I do is the following: “What’s the cost of a Taiwan crisis,” times, “what’s the probability you think it’s going to happen.” The cost is easily in the low trillions of dollars. That’s where the best forecasts that have been done place it. The likelihood, nobody knows for certain; and different people have different estimates, and I’m struck by the extent to which, in the private sector, estimates are meaningfully lower than when I talk to folks in the Pentagon or in the intelligence community, but I think that’s the mental math for how much insurance we ought to be buying.

Well, we’re buying $40 billion of insurance. It seems like that’s on the low end, based on your back-of-the-envelope calculation.

I think that’s right. If your probability of things going wrong are 10 percent, or 20 percent, or 30 percent, we’re probably not buying enough insurance.

Cutting checks is one thing. Of course, we need to be able to build these plants, they need to be able to be built in a timely manner, and we need to have the workers to run them. When you were talking about how important the workers are in Taiwan, I was thinking, there’s a couple of news stories about problems duplicating the kind of work environment here that you’d find [in Taiwan], forget about just finding the qualified people, once they’re here, working under the kinds of conditions, and the pace, and the rules they have in Taiwan. That seems to be a problem.

How would you rank your concerns about getting to that 20 percent number by 2030? Would it be regulations? Would it be workforce issues? Again, cutting the checks is sort of the easy part, even though that’s not so easy.

That’s absolutely right. I think that the challenge is going to be less in actually operating facilities when they’re up and running — these facilities are highly automated, and so you need some workers, but actually not that many, because it’s mostly machines doing the job for you. The place where the US has faced a lot of challenges is in the construction phase, where, due to the way federal contracts work — CHIPS Act construction has to use entirely union labor — and so that has driven up costs, but also because there have been relatively few people in the US who have built CHIP fabs recently, and so a lot of the skills needed build them have had to be learned on the fly. Whereas in Taiwan, there are construction crews that do nothing but build chip factory after chip factory, year after year.

And so, in addition to wage differentials in the construction industry and regulatory differentials, there’s also know-how differentials. And so I’ve been struck, if you talk to chip companies that are building factories and ask them, “What’s the type of labor you’re really facing shortages?” People think they’re going to look for material scientists or PhD physicists, but actually it’s welders who have previously worked in building chip plants, because the productivity differential between a welder who doesn’t know a chip plant and a welder who does is apparently vast, substantial, because you’ve learned how to weld in certain ways, certain types of equipment. And so that’s the type of labor that actually we need is in short supply, and that’s where you see real cost differentials between building the US versus building in Taiwan or Korea.

How about the regulatory aspect? That certainly comes up in a lot of news stories, not just about chip plants, but building other things in the United States. Is that a significant problem? Environmental regulations, whether it’s the Clean Water Act,NEPA [National Environmental Policy Act], what have you.

NEPA is certainly something you hear a lot about from people building new chip plants, not so much the regulations, but just the time it takes to get through them is the key barrier. Companies say, “We’ll meet the regulatory standards, we just want to know what the timeline is so we can get our plants under construction.” If you compare regulation in the US versus in Taiwan or Korea, there’s a whole lot of clean-air and clean-water regulation that’s relevant to chips facilities, but actually, there’s not much difference between regulation here versus in Taiwan.

The outlier here is actually in China, where regulation is much, much looser, and so part of the cost differential that you get in China is that there’s just no environmental regulation to speak of, and that creates real cost differentials compared to building in the US.

A big part of the US manufacturer aspect of this would be Intel, a company which seems troubled. I think in the most recent earnings report there were analysts questioning their viability as a going concern. How worried are you about Intel, and, if it came to it, would we just bail them out like we did the auto companies back during the financial crisis, and would you be in favor of that?

So in the chip industry, there are three companies that produce at the cutting edge: There’s TSMC in Taiwan, Samsung in South Korea, and Intel of the US, and these are the only three companies that are really trying to produce the cutting edge, the only three that have any scope to do so, and in the past couple of years, TSMC of Taiwan has been far ahead, Samsung has been in the second place, and Intel has fallen behind the last couple of years, both in terms of their manufacturing technology has slipped behind TSMCs, but also, as you say, their business model has faced a series of really tough issues as the types of products they’ve produced have fallen out of favor. And this creates, I think, a real dilemma for the US. The US would love it to be the case that the world’s leading chipmaker is a US firm, but if you look at who’s getting support from the US government, it’s a mix of TSMC, and Samsung, and Intel, precisely because I think the US didn’t want to bet on only a single horse if that horse looks like it might not win the race. I think that’s a reasonable strategy: Diversify your bets. But as you say, Intel faces ongoing challenges, and I think we’re going to see over the next couple of years whether it needs to take further changes to its business model to stay in business. The US tech sector, I think, is perfectly comfortable not having US-owned firms in the industry. If it’s only Samsung and TSMC, I don’t think US tech companies would see that as being problematic, but I think a lot will depend on how much these firms continue to invest in the US, or at least outside of their home markets, to provide the sort of geographic diversification that the US government is pushing for.

Silicon Valley may not care, but is it important, politically, that one of these companies is an American company, and if so, should we do whatever it takes to make sure that an American company is viable? Again, even if it would require nationalizing it, bailing it out. We have a fairly recent history of doing that to companies that we think are — it’s important that American companies manufacture cars, and if it’s important they manufacture cars, it has to be at least important that they’re manufacturing something that you say is as strategically important as oil.

I think the most important thing is that technology companies, people developing AI systems, get access to the best chips, and so we don’t want to be in a situation in which we’re supporting a second-best player solely because it’s an American firm. If you look at the auto bailouts, I think that there’s a lot of reasons why that is not a comparable situation to where we are today. The automakers were essentially viable firms that just ran into a major short-term financial problem during the 2008 crisis, but they knew how to produce cars, they knew how to sell cars. If Intel were to face more serious problems than it currently does, the question would be, can it get its manufacturing on track and keep up with TSMC? There would be a question of viability, and so simply giving the company five billion more dollars I think is not going to really resolve the challenges that they face. The challenges are in manufacturing technology, the challenges are in designing chips, and they’ve got to do both of those, it’s not fundamentally a financial problem.

You can hear people in Washington talk about, should we mandate that US firms buy chips from US companies? And I think that would be a dangerous road to go down, in part because it would mean mandating that the most successful tech companies in the US don’t get to choose who they buy their chips from, which means that they’re not able to buy from the best technology providers, and so I would be very cautious about going down that road, given the implications it would have for the competitiveness of the rest of the US tech sector.

It sounds like a bad next step would be what you just described, that kind of mandate. What would be a good couple of next steps by the federal government to ensure that a lot of high-end chips are made here? Or is everything done? We passed the CHIPS and Science Act, full speed ahead.

I think the government has done a lot on the funding side, and simultaneous to that, private markets are doing a lot on the funding side. Just look at NVIDIA and its valuation. So I think funding right now, that issue has been dealt with, at least for today. The next question is implementation, and that’s dealing with the regulatory issues that we’ve discussed, helping address some of the workforce issues that we’ve discussed, because, ultimately, we now know which chip plants are going to be built, and we need to make sure these plants can be operated profitably. That’s the next question, and that’s primarily a task for the companies, but governments set the environment in which these facilities are operating, and so we need to make sure they’re operating in the most efficient environment possible.

The other aspect for this is actually tax policy, because for companies making multi-billion dollar investments, the tax treatment of their capital investment is actually a huge variable in the profitability of these firms. And that’s a place where the US has, in the past, been less competitive than countries in East Asia, and I think it speaks to the importance of tax policy and shaping where companies will keep investing in the future.

(This post is republished from AEI.)