35
Are Microchips the New Oil?

Hugo Scott-Gall: Today, I’m delighted to have with me Chris Miller. Chris is an associate professor of international history at the Fletcher School of Law and Diplomacy at Tufts University. He’s also the Jeane Kilpatrick visiting fellow of the American Enterprise Institute and Eurasia Director of the Foreign Policy Research Institute.

His research examines key shifts in international politics and economics. His latest book, Chip War: The Fight for the World’s Most Critical Technology is an epic account of the decades-long battle to control what has emerged as the world’s most critical resource, microchip technology. Chris, thanks so much for being here.

Chris Miller: Thank you for having me.

Hugo Scott-Gall: Okay, so let’s get going. I think we need to at the start set a few things up. So, the first thing I’d like to ask you, really, is why now? Why is it so important to write a story, which I think has not been told enough or told as well as you bring it all together? So, why did you decide to do this now?

Chris Miller: I’m actually a historian of Russia by training. And I started out wanting to write a book about technological change in military power. And the question that puzzled me was why was it that during the Cold War the Soviet Union could produce atomic weapons, could produce rockets that shot people into space, but they couldn’t produce the types of military technology that was changing the balance of power by the end of the Cold War.

Above all, the development of munitions that could strike targets with precision hundreds of miles away. And the more I dug into this shift in military technology over that time period, the more I realized that at the core of this change in global power was semiconductors, the tiny devices that let us miniaturize computing power. And during my research, I learned that the first semiconductors that were invented were created for missile guidance systems during the Cold War.

They needed a small type of computing that could fit on the nose of a missile as it flew through the air. And since that point, there’s been a really deep interrelationship between the development of military power and the development of the computing capabilities that today we take for granted.

Hugo Scott-Gall: So, I completely understand why you’ve written the book. But why is it that–I think most people can get their heads around geopolitical power being quite correlated with control of energy. I think that’s quite well known and well established. But it isn’t so well known and well established around microchips. So, why do you think that is? Is it because the war is tangible and microchips are still tangible, they’re just very small and people are just less familiar? So, I’m just wondering why–do you agree with the assertion that actually they are a source of geopolitical power. If you agree with that, which I think you do, why are they given much less prominence as a source of power than say energy is?

Chris Miller: Well, I would go so far to say that chips are even more politically influential than oil is. And if you look at the production process of chips, which we’re going to dive into, there’s much more concentration on the chip industry than on the oil industry. Saudi Arabia produces 10-15% of the world’s oil. Taiwan produces 90% of the world’s most advanced semiconductors. And in other parts of the chip supply chain, there’s even more concentration, which gives rise both to pricing power for companies but also to political power for the countries that can control them.

But we haven’t realized it because most of us never, ever see a chip although we rely on them for all of the devices we use on a daily basis. Unless you take apart your smartphone or your dishwasher, you don’t see the chips inside. And so, most people, myself included, when I started the research, had no idea the number of chips that we touch over the course of our daily life. When you wake up in the morning and your alarm clock rings, there’s a chip inside of there. You turn on the coffeemaker, there’s a chip inside of there. You sit in your car, there are hundreds of chips inside of your car. And that’s even before you’ve opened your smartphone, or your PC, or logged onto the internet, which is basically just a bunch of datacenters scattered around the world.

And a datacenter is nothing more than a building full of semiconductors. But we never see them, so we don’t really think about them. And we haven’t realized the extent to which they provide the building blocks for all of the modern economy, even though most of us never actually think about it.

Hugo Scott-Gall: Let’s do some definitions. In the simplest form, what are chips and then what are the different types of chips?

Chris Miller: So, to start, the word chip, semiconductor, and also integrated circuit can be used basically interchangeably. A chip is a piece of silicon in most cases that has tiny circuits carved into it. And you can measure a chip’s computing power, roughly speaking, based on the number of transistors that are carved into it. And so, if you go to an Apple store today and buy a new iPhone, the primary chip in the iPhone will have 15 billion transistors on it, each one the size of a virus. And these transistors turn on and off on a regular basis.

When they’re on, they create a one. When they’re off, they create a zero. And these are all the ones and zeros that undergird all software, all operating systems, all computing. You couldn’t have modern computing without the chips that actually are doing the computing for us.

Hugo Scott-Gall: And there are different types of chips. There are logic chips, memory chips. You’ve got CPUs, GPUs. What are all of them?

Chris Miller: So, you can divide chips into three different categories broadly. There’s types of chips that process data, which are the main chip in your smartphone, for example, your PC, often called a CPU, a central processing unit. There are types of chips that remember data and store it over short or long periods of time. And you’ll find these in your phone or your PC as well. And there’s a third more diffuse category of chip often referred to as an analogue chip, which take real world signals, like optical signals, images, or radiofrequency signals, and convert it into digits, ones, and zeros.

So, any wireless device you’re using, whether Bluetooth or a cell phone, will have a chip that is managing the conversation of ones and zeros into a radio wave that can fly through the air. So, there are lots of complex chips needed for that analogue to digital conversation as well. And within these three broad categories you can talk about many more nuances, but these are basically the three main types of chips that we use.

And in a smartphone, you’ll have multiple examples of memory chips, multiple processors, and a fairly large number of different types of analogue chips that you need to make a smartphone work.

Hugo Scott-Gall: And obviously, to make chips you need a number of different things. Can you talk a little bit about the equipment?

Chris Miller: So, to make an advanced chip with say 15 billion transistors on it, each one measured in a number of nanometers—that’s billionths of a meter in terms of their size—you need the most precise manufacturing equipment that humans have ever made. So first, you start the process of making a chip by designing it.

And to design a chip with tens of billions of transistors you need really ultraprecise design software that is produced largely by three firms based in the U.S. that have the unique software capabilities. Then, you need to acquire the tools to actually manufacture chips and you manufacture chips by taking a circular piece of silicon called a wafer. You shine light at it in a certain pattern. The light reacts with chemicals to carve shapes in the silicon.

And these shapes, after repeating this process multiple times, become your transistors. That’s a very simple articulation of the process. In fact, it’s really, really complicated because the precision needed is at the level of billionths of a meter. And so, there’s only a small number of companies that can create the tools necessary to produce these chips. And the segment of toolmakers, there’s only five companies that stand out—one in the Netherlands, one in Japan, and three in the United States—that have a real strong market position in the production of the most advanced tools.

And then, you need to use these tools with a bunch of ultra-specialized and purified chemicals to actually produce chips. And this is extraordinarily complex and expensive and there’s really a small number of firms that are capable of producing cutting edge chips. And one that I’m sure we’re going to talk about when it comes processor chips is the Taiwan Semiconductor Manufacturing Company, which is the world’s largest chip maker and also the world’s most advanced chip maker when it comes to making processor chips.

Hugo Scott-Gall: Yeah, so absolutely. I definitely wanted to get there, on how hard this is to replicate because that’s almost the key question sitting in the middle of it. But I think if it’s okay with you I’d like to do a bit of history because reading your book, one of the things that I just didn’t know, I didn’t appreciate, was one of the motivations to improve chips was the military and lessons learned from the U.S.—the Vietnam War and the U.S.’s participation and the inaccuracy of its bombs.

And so, that was a surprising motivation. So, if you could maybe go back to the very start, how did the industry come about and what were the motivations? And then, we can do who and where because the people, I think, are actually almost as interesting as the technology.

Chris Miller: That’s right. Before the first semiconductors were invented, the computers that existed—which were extraordinarily simplistic and the capabilities relied on devices called vacuum tubes, which were like light bulbs. They turned on and off.

But they regularly burnt out. They attracted moths because they were shining with visible light. And they were impossible to miniaturize. So, computers at the time were the size of entire rooms. And even still, they had a tiny fraction of the processing power of a contemporary iPhone. And so, there was a real push by the military to find ways to shrink computing power so it could be put in weapon systems distributed across a battlefield.

And that was the initial impetus for semiconductors. Shrinking computing power, making it more energy efficient in the process, and therefore allowing it to be deployed not only to better computers in buildings but also in all sorts of devices that could be used in a more mobile fashion. And that prodded the initial funding for the first semiconductor firms. The first major order for semiconductors was for the chips in the Apollo spacecraft’s guidance computer that sent astronauts to the moon. The second major order was for the guidance computer on the Minuteman II intercontinental ballistic missile, which was designed to send nuclear warheads to the Soviet Union.

And since then, there’s been a real deep relationship between the Pentagon and its R&D and the chip industry in the U.S. and globally.

Hugo Scott-Gall: That relationship is deep, but it hasn’t been consistent. Is that fair to say?

Chris Miller: That’s right because the industry has shifted since the earliest days. When the chips were first invented, the Pentagon was buying almost all of them. Today, however, only a couple percentage of chips made globally end up in defense or aerospace uses. Most go to civilian applications today. Smartphones consume around a quarter of chips produced by value, PCs around 20%. And so today, for most chip firms, its consumer uses that are the most important. And that’s made it more difficult for defense planners to shape the industry simply because they’re less important customers. That’s beginning to change, though, as geopolitical tensions rise. I’m sure we’re going to discuss a bit of that dynamic.

But it’s forced a shift in the industry in recent decades because since the 1990s or so, the chip industry hasn’t really thought much about defense uses because it’s much more profitable to sell to smartphone or PC markets.

Hugo Scott-Gall: So, I’ve jumped ahead, and I shouldn’t have done that. Let’s go back. The industry is founded in the U.S. and it’s by some fascinating people, which you could describe as a brilliant crazy bunch doing some far-out things. Can you talk a bit about how these people found each other. I guess some of these people were brilliant scientists. They’re all PhDs, but they’re also visionaries, but they also got the job done as well—which I think that blend makes them super interesting.

Chris Miller: Yeah, one of the things that really stood out to me over the course of the research is that brilliant scientists alone are not enough to create new industry. You need brilliant scientists and engineers, but you also need effective managers. You need people who understand supply chains. You need, above all, people who understand where the market’s going to be. And if you look at the people who invented the first chips and built the chip industry in its early hubs in Silicon Valley and then also in Texas, where Texas Instruments was headquartered, what you find is that the real leaders in the industry married expertise in the science with business thinking.

So, if you look at, for example, Bob Noyce, one of the founders of Fairchild Semiconductor, and then Intel later on, he was one of the people who invented the first semiconductor. But his real expertise wasn’t so much in the science, although he was an MIT physics PhD. His real expertise was in understanding the way the technology could develop and meet new markets. And so, although when he started producing chips he was selling mostly to NASA and the military, he’d already envisioned a future in which most chips were going to civilian computers because he realized the cost was going to decline as the technology improved and it could sell far beyond simply government customers.

And similarly in Texas, in Texas Instruments, which was really a cutting-edge firm in the early days of the chip industry, there too the CEO Pat Haggerty realized that because chips had become more powerful and smaller, they could be used far beyond mainframe computers.

And so, he was envisioning pocket calculators, mobile devices, decades before they were actually implemented. And that’s really what differentiated the good research labs from the transformative businesses, was that ability to combine the business thinking with the technological capabilities.

Hugo Scott-Gall: And so, to get these things off the ground, you need capital and so the government’s role cannot be underestimated. And the main motivation was initially military. And I guess, as per the Space Race, there were all sorts of spillover innovations that come out of government funding. Is it fair to say that motivation didn’t necessary shift from military, it was more that there were ways of monetizing by finding new end markets? And that was the next phase of growth. But then, the industry began to shift with the rise of Japan. But, I guess, before we get to Japan you mentioned earlier at the start Russia.

So, Russia realized, I think—correct me if I’m wrong—that militarily it was falling behind and it needed to catch up. So, it tried to replicate the industry of the U.S. in Russia, but it failed. Why did it fail and why did Japan then succeed?

Chris Miller: So, the failure of the Soviet Union is actually an interesting puzzle in some ways because the Soviets had brilliant physicists, an extraordinary educational system, a ton of capital investment, and a big military market just like the United States. What they didn’t have is a consumer market. And they didn’t have an international supply chain.

And so, whereas U.S. firms could sell to Europe, could sell to Japan, it could acquire equipment and materials from the best companies across almost all the industrialized world, Russia was trying to do it all on its own. And that was a bad strategy given that the Soviet economy was a small share of global GDP. And so, from the earliest stages, the Soviets decided they couldn’t really innovate on their own. They tried to copy instead. And copying sort of worked in that the Soviets were able to copy chips.

But copying is not very effective as the strategy in an industry where progress improves at an exponential rate. And because of the extraordinary increase in processing power, the Soviets copied chips but they were a half decade or a decade behind and not doing them very much good because the industry in the West had raced ahead and experienced exponential growth in the interim. And so, as a result, the Soviets were never able to build a viable chip industry despite pouring lots of money and lots of scientific expertise into it. And that’s what the Japanese did really well, is they took expertise. They took money. But they found market opportunities globally.

And so, Japanese firms succeeded because they were able to acquire the best equipment from the U.S. and Europe. They sold to global markets as well. And they showed that it’s possible to catch up in chips. But it’s only possible when you’re deeply integrated with global markets rather than trying to do it domestically or indigenously which was simply too hard to do in an industry that’s complex.

Hugo Scott-Gall: So, from a U.S. point of view, was Japan too successful and did that lead to the rise of Korea?

Chris Miller: Well, there was a lot of worry in the U.S. chip industry that Japan was too successful, especially in the late ’80s when Japan was peaking in its influence. And there were a number of trade disputes between the U.S. and Japan about alleged dumping of Japanese chips in U.S. markets. But ultimately, I think the key challenge that Japan faced is that it was able to catch up very effectively when it came to producing the specific types of chips that were the focus in the 1980s DRAM memory chips.

But Japanese firms were too focused on that particular roadmap and executing it very successfully. But they didn’t think about what was coming next. Companies like Intel, which were smaller companies at the time, were not focusing on producing that same type of chip better. They were focusing on different types of chips, like the CPUs that ended up being in every laptop computer. And as a result, Intel was able to out innovate the Japanese, even though the Japanese had an extraordinary capacity to produce very high-quality chips. They just couldn’t find a way to envision the next generation of chips that would sweep across global markets.

Hugo Scott-Gall: Is that a vision thing? Is that a—back to people, the people with vision plus execution drives innovation to keep you cutting edge?

Chris Miller: No, I think there is something unique about the atmosphere in Silicon Valley where companies like Intel were based that corporate leaders focused on what was coming next technologically. But I also think that in Japan there was too much of a focus on winning market share, that the leading companies saw as their primary goals. They spent extraordinary amounts of money in capital expenditure given to them by banks that offered below market rates on loans. And so, the CEOs at the time just focused on market share rather than focused on profitability. And they won market share. But they won market share in a way that made them all unprofitable, and it ended up being a very bad business model.

Whereas, in the U.S., there was extraordinary difficultly in raising capital in the 1980s given how high interest rates were. So, companies like Intel decided not to compete on their ability to raise capital, not compete on their ability to build larger factories and win market share per se, but rather on devising new types of products that they could charge very large margins for.

Hugo Scott-Gall: Let’s talk a bit about the role of Taiwan. As an aside, I still find it odd that Silicon Valley and Taiwan both are in seismically very dangerous places. It is odd that you get these two concentrations of expertise in essentially geologically unstable places, but also geographically unstable in the case of Taiwan. So, how did Taiwan rise? Because I think the story of Morris Chang is fascinating but it’s not just him. But the rise of Taiwan I think we should discuss because it gets us to the present.

Chris Miller: Well, the Taiwanese had identified electronics and semiconductors as a priority area since the 1960s. And at first, the Taiwanese government just wanted to attract some pretty simple electronics assembly jobs, like assembling TVs or assembling radios.

But over time, they focused on acquiring more advanced types of technology and identified semiconductor manufacturing—not just putting chips into devices but actually making the chips themselves as something they wanted to focus on. And Taiwan benefited from a number of unique factors. Although it’s a long way away from Silicon Valley in terms of number of miles, it was actually deeply intertwined with Silicon Valley in terms of individuals. There were many dozens of influential Taiwanese technology executives in Silicon Valley in the 1960s, ’70s, and ’80s.

Lots of Taiwanese students studying at Stanford and Berkely. So, there were a ton of individual connections that made Taiwan a lot closer in terms of its connection to Silicon Valley than almost anywhere else in the world. And that helps explain why the Taiwanese government was trying to find ways to help set up new companies to move forward technologically in the chip industry.

And they identified a businessman by the name of Morris Chang. He was born in mainland China but spent his career at Texas Instruments in the 1950s, ’60s, and ’70s where he rose up to almost become the CEO of Texas Instruments. But he was passed over in one of the great errors of American business history. And eventually, he left TI and was looking for another job and was approached by the Taiwanese government. And they said, “Would you like to start a new semiconductor firm,” and offered him a fair amount of funding to do so.

And Chang took up the opportunity and built this new firm called the Taiwan Semiconductor Manufacturing Company into what’s today the world’s largest chip maker. And he did it with a couple of advantages. One was the Taiwanese government was very supportive of TSMC in terms of tax policy, in terms of education, but also because TSMC had a really unique business model. Before its founding, almost all companies designed and manufactured chips in-house. But Chang wanted to split the design process from the manufacturing process.

He said, “I’ll focus on the manufacturing. I’ll put in the capital expenditure. I’ll hone the ultraprecise processes needed to manufacture chips. And I can sell these services to lots of different chip companies that will only have to design chips.” And because they don’t have to manufacture, the initial startup cost will be much lower. And so, he enabled a flourishing of chip design startups that didn’t need any expensive equipment, could just hire a couple of designers, design new chips, and take their design to TSMC for fabrication.

And that shift in business model proved remarkably successful and it explains why TSMC is today one of the world’s most important tech companies, because it produces the chips that firms like Apple are critically reliant on.

Hugo Scott-Gall: So, before we get into the present and the future, I want to just circle back a little bit on how hard to replicate—because if that history that you just gave us shows you the central importance of chips to the modern economy, but also the U.S.’s dominance. Then Japan. Then comes Korea. Then comes Taiwan.

[00:23:00] So, we know that chips are super important. We know they’re strategically important and we know that they can transform, or certainly accelerate, your military power. So, we know the motive. But how hard is it to replicate? Because this kind of gets us on to China in minute. But what is it that makes it so hard? And I guess it’s a combination of things. But is it know-how, people? Is it capital? Is it access to materials? Or is it disentangling very entangled supply chains? And I think the answer is all of those but I just wondered how—

Chris Miller: All of those.

Hugo Scott-Gall: —you would rank them?

Chris Miller: The answer is all of those. I think that the capital investment is massive when it comes to chip fabrication and also the tool manufacturing. But capital investment is something that governments can subsidize and often have tried to subsidize with differing levels of success. So, capital is far from enough.

The personnel is really critical. And what you find is that it’s not just a question of having smart people. You need smart people but it’s not enough. Nor is it enough to have people who are trained in physics or trained in electrical engineering, although you need a lot of PhDs. Because the really unique expertise is only built up in the manufacturing process itself, you need people who have worked in manufacturing for a long time, worked in one of the leading-edge companies, who understand how the processes work. Because there’s only a small number of cutting-edge chip making facilities in the world and only a small number of companies that make cutting edge tools.

And there’s nothing you can study in an academic environment that gets you the information you need. Simply impossible. So, what you find is that having worked for a couple of decades in a leading company is a critical starting point for having any hope of catching up. And the companies that have emerged as startups, that later on played a big role in the industry, were almost all founded by people who had already cut their teeth in the industry for several decades.

Morris Chang is a good example. He founded TSMC only after having served for three decades at Texas Instruments. In addition to that, though, you also need to find a niche in a very complex supply chain. And here, the supply chain makes it more difficult for new firms to enter because all of the companies in the semiconductor supply chain are planning their research and development processes over a five- or 10-year time horizon. And they’re planning their R&D already knowing who their customers will be because there’s a small number of firms in the industry.

So, if you’re a toolmaker, you know you have to make your tools in a way that’s going to fit into Intel’s, or TSMC’s, or Samsung’s fabs because they’re going to be your biggest customers guaranteed. And so, what you find is that the toolmakers and the chip manufacturers essentially codevelop their technology. And the same is true for the chip designers as well. And so, they’re deeply interlinked, which makes it very hard for a new firm to jump into the ecosystem because they don’t have all the existing relationships, the credibility, or the knowledge to do so.

And as a result of that, we can say with pretty high confidence who the biggest chipmakers will be in five years’ time because we already know the R&D that’s underway to make the technology that will be cutting edge in five years. And we know who’s doing it. And it’s a small number of firms, all incumbent players.

Hugo Scott-Gall: So, as you think about this industry versus other industries, the entry barriers are super high, which I would argue for an ongoing status quo, as you just described. However, when you have a country that says, “I don’t like my external dependency. I want to create my own industry because I don’t like being externally dependent”—most companies don’t like being externally dependent on, whether it’s energy, food, water. And it turns out its chips in the case of China. China wants to build its own industry. How hard is that to do? Because everything you just said makes me think it’s very hard but China’s going to try anyway.

Chris Miller: Well, it’s going to be very hard. China has a couple of advantages. It’s going to spend a lot of money, which is certainly a prerequisite. It’s got a substantial electronics industry. So, a lot of the smartphones and PCs that are used are assembled in China, which means that China acquires a lot of chips in the process to assemble them into devices. So, it’s got some leverage in its half into the electronics industry. But when you look at the specific semiconductor product process, China’s a pretty small player. China does produce a fair number of chips but they’re almost all low end in terms of value and in terms of technological capabilities.

And when you’re looking at cutting edge production processes, you don’t really need any Chinese products to produce a cutting-edge chip. And you can’t produce a cutting-edge chip using exclusively Chinese products. It’s simply impossible. You can’t even produce anything close to cutting edge using exclusively Chinese products. So, it’s going to be quite hard for China to reach anything close to self-sufficiency.

And that’s sort of obvious because you can just look at the United States which is by far the world’s biggest player in the chip industry in aggregate, looking across the supply chain. The U.S. is also completely incapable of producing cutting edge chips from the design all the way to the end of the supply chain. So, if the U.S. can’t do it, it’s pretty hard to imagine the Chinese are going to succeed. But it does look like they’re going to try.

Hugo Scott-Gall: Which gets us back to Taiwan. So, Taiwan is incredibly, strategically important from the point of view of manufacturing. So, what you’re saying, I think, is it’s very hard to replicate what Taiwan is capable of and indeed in an efficient, harmonious system where everyone gets along fine, you don’t need to. But that could change. And so, from a Taiwanese point of view they have unique capabilities that they need to keep unique. From the rest of the world’s point of view the world is safer if you have a power competition between U.S. and China. The world is safer if what Taiwan does can be replicated elsewhere. How does that resolve?

Chris Miller: Well, that’s the big question that’s tearing apart the chip industry and the electronics industry right now. Over the past couple of months, the U.S. has taken new steps to restrict the transfer of chipmaking technology to China, trying to partially split apart the two countries’ electronics and computing industries. Simultaneous to that, a number of big U.S. electronics firms are meaningfully reducing their reliance on assembly and production in China.

So, for example, Apple is beginning to shift more of its iPhone assembly outside of China. Or Dell recently announced it’s going to stop buying chips produced in China, low end chips. They’re going to stop sourcing any chips from Chinese firms. These are really meaningful steps by large companies that play a big role in the electronic supply chain.

And Taiwan is, as you say, trying to defend its critical role in the industry, both for economic reasons—TSMC is a very profitable company—but also for political and strategic reasons. And the challenge that Taiwan faces is that the rest of the world is looking at the Taiwan Straits very nervously, concerned that the political tension might spill over into military tension. And as a result, governments from the United States, to Japan, to Europe are trying to subsidize the moving of some production capacity to their countries, or at least away from the Taiwan Straits.

And you had TSMC, the biggest Taiwanese chipmaker, begin to build new facilities in Japan and Arizona, which will be online in the next year or two. And this is intended to assuage American and Japanese fears about chip supply, but it also increases concern in Taiwan that in fact its most important industry is being hollowed out by the U.S.-China decoupling.

Hugo Scott-Gall: I think it’s worth reiterating or underlining just how important Taiwan is. So, if there was a meaningful disruption for a period of time to Taiwan’s ability to export chips, that would mean no new iPhones for a couple of years, or it would have serious consequences in lots of ways that perhaps the average person wouldn’t be able to imagine.

Chris Miller: Yeah, and it’s certainly impactful for Apple or smartphone producers. You’d see smartphone production globally fall to something not too far from zero. But it’s not just PCs and smartphones. It’s almost all sorts of electronic and manufactured goods. If you look at automobiles, for example, they have hundreds and in some cases thousands of chips inside. Dishwashers, microwaves, coffeemakers, all sorts of industrial equipment, airplanes—the world economy today is critically dependent on chips.

Almost anything with an on/off switch has a chip inside. And you don’t need to produce all of these chips in Taiwan per se, but Taiwan today produces almost all the most advanced processor chips. And it produces in aggregate over one-third of the new computing power the world adds each year.

So, in addition to the most advance chips for AI and datacenters, Taiwan produces a ton of chips that go into dishwashers, too. And if we lost access to a third of that computing power, we produce a lot fewer automobiles and dishwashers in addition to a collapse in smartphone production that coincided with it. That would be the worst decline in global manufacturing, I think since the Great Depression.

Hugo Scott-Gall: I think it’s fascinating to think about the world in terms of these huge pots of economic value add sitting on these small chokepoints. And again, back to the start, we know it with energy, but you don’t necessarily know it—and I think there’s an example really around say the manufacturing output of Germany versus its reliance on a chokepoint of gas coming from Russia certainly pre-2022. So, economic value add dependency versus chokepoints.

But in a world that is—and I think this is an overused expression—but deglobalization, or certainly a change in global geopolitics that is perhaps accelerating the desire to reduce external dependences, to improve security of key things, the logical outcome of that is for more localization. Localization of manufacture, greater control of resources required. But everything you just said around how hard this industry is to replicate would make trying to localize harder.

But if you think of a country like the U.S., which when motivated by the original Sputnik moment, was capable through government funded research, capable of some very, very important primary R&D and innovation. Do you think that is the inevitable endgame here, that you will see trusted economic blocks trying to generate as much as they can within the geographic blocks and that this industry, even though it’s hard to imagine, now have some real hard to replicate skills, that that’s where it will have to go.
And therefore, you might see a big increase in allocation of resources including people. So, you might well—the number of high IQ people in the country, you probably need a high percentage of them working on these kinds of problems that perhaps we’ve had in the last 20-30 years.

Chris Miller: I think the dynamic you’ve sketched out is right, but I think the economic block the U.S. is trying to put together goes far beyond just the U.S. It also includes Europe, includes Japan, includes South Korea, and even parts of Southeast Asia and India. And what you find when you dig into how the investment patterns of tech companies and electronics companies are changing, is not that they’re onshoring a lot to the U.S. but that they’re moving from China to almost anywhere else.

So, if you look at electronics assembly, like assembling smartphones, you see a lot of capacity moving to Vietnam and to India. If you look at chip investment, there’s a lot of chip investment in the U.S. but also a lot in Japan and Europe.

So, I think that deglobalization is the wrong framework. After thinking about this, I think de-China-izatoin is actually what the tech industry is doing. We’re having a bifurcation of supply chains, one focused on China, and one focused on most of the rest of the world, especially if you look at the world in GDP terms. And from the U.S. government’s perspective and the U.S. industry’s perspective, that’s a much preferable outcome because if you’ve got not a global economy anymore but an economy that includes Europe, and Japan, and Korea, and other major players, you’ve still got a large number of customers and a large number of potential suppliers that can give you the components that you need.

And that’s going to be a lot more efficient than if we were to break up into a European block, a North American block, and a number of different Asian blocks. For China, that’s a real challenge, though, if in fact that vision fully materializes because China’s a big economy. But on its own it’s a small share of the global economy.

And if Chinese tech firms are really constrained to just operate in China and a small number of other emerging markets, that’s not a very attractive future for them. And it doesn’t suggest that they’re going to be able to scale in a way that Western firms will be able to.

Hugo Scott-Gall: I agree with you. I think any change from how the last 20 years have been is automatically labeled deglobalization and I’m not sure it’s actually the right way of describing it. So, before we finish, a few more questions, one really about the future. And let’s be optimistic here. Things like AI, chips are central to the rollout and the diffusion of AI. Secondarily, power of chips has driven a huge change in capability. We talked about our smartphones, servers—the whole cloud is powered by chips.

So, are you a pessimist that actually we’re now talking tiny, tiny amounts that it’s hard to improve capability of chips and we’re at the end of what people call Moore’s law, named after Gordon Moore, one of the original crazy geniuses. Or actually, if we were doing this conversation in 10 years’ time, would we be still marveling at just how much more chips are capable of than they were 10 years ago when we were still quite naïve perhaps technologically.

Chris Miller: Yeah, I’m on the optimistic side of that debate I think for a couple of reason. First off is the shrinking of transistors that has enabled most of Moore’s law over the past 60 years still has some way to go. We have a pretty clear line of site until 2030 or so as to how transistors can be shrunk even smaller and therefore more can be packed on the chips. And that’s going to enable on its own a pretty substantial advance in computing power. But next to that, we’re, I think just in the early stage of a pretty revolutionary moment in chip design whereby we’re getting more effective and efficient ways to design chips that let them use the same amount of computing power but do more with it.

So, for example, in the AI space, we’ve seen a shift in running AI algorithms away from traditional CPUs and datacenters towards GPUs, which are structured and designed differently. It’s not that they have more transistors per se, it’s that they use the transistors in a smarter way and are able to accomplish tasks more efficiently. And I think we’re just in the early stages of this effort to understand what our future architecture is to design chips that we haven’t even though of yet and what are the use cases that are enabled by that.

And that makes me optimistic that even if Moore’s law eventually does slow in terms of the ability to make transistors even smaller as they approach atomic limits, we’ll still find ways to design them more effectively and get increases in computing power from that aspect.

Hugo Scott-Gall: I agree with you, although I know far less than you. In terms of the application, a shift in capability can lead to disproportionately beneficial changes in all sorts of different industries, whether you’re talking healthcare or whether you’re talking finance.

I think the benefits as they trickle down to various industries are almost unknowable, certainly unforecastable, because you just don’t know how one thing changes another thing changes another thing. So, final question to you is, you published a book. It’s very good. It’s excellent. It’s been very well received. Have important people in important places called you up saying, “Come and talk to us?” When you think about decision makers, people in positions of power, do they understand all of this? Are they asking you questions? If you have met them, what’s your perception of their understanding and their vision of the future around this?

Chris Miller: I think the baseline level of knowledge about the chip industry of a typical government official has improved dramatically over the past couple of years. If you’d asked me this question three years ago, I’d have said the baseline level of knowledge was pretty low. I can speak with confidence about the U.S. government, the Japanese government, and some European governments.

There’s a pretty substantial reservoir today of expertise in the industry. Not that everyone is an expert, but they’ve really done their homework over the past couple of years to understand how the industry functions. And so, I came away fairly impressed with what the U.S. government and other governments have done to learn about the industry given how complicated it is and given how hard it is to stay up with how rapidly the industry’s changing. I think the policies that have been made over the past couple of years have been made on the basis of accurate information and in pretty close consultation with industry.

Hugo Scott-Gall: I can tell you’ve been in high places and secret places by your carefully crafted answer. But I think it must be fascinating. And all I want to say now is thank you very much for coming on the show. It was a pleasure and treat to have you on and I really enjoyed it.

Chris Miller: Well, thank you for having me.