Paul Diggle

Hello and welcome to Macro Bytes, the economics and politics podcast from abrdn. My name is Paul Diggle, Chief Economist here at abrdn.

Luke Bartholomew

And I'm Luke Bartholomew, Senior Economist at abrdn.

Paul Diggle

And today we are talking to Chris Miller. He's an Associate Professor at Tufts University. He's an academic and historian and he's the author of Chip Wars: The Fight for the World's Most Critical Technology. It's a Financial Times Business Book of the Year. It's a really riveting account of the history and the geopolitics of microchips. Chris, welcome to the podcast.

Chris Miller

Thank you for having me.

Paul Diggle

So, Chris, you write in the book that semiconductors have defined the world we live in, determining the shape of international politics, the structure of the world economy and the balance of military power. Our world is defined by quintillions of transistors and a tiny number of irreplaceable companies. And it's that contrast, the technological marvel that is fitting an ever-increasing number of transistors onto a chip and the concentration of that expertise commercially, geographically, that's really the theme at the heart of the book. But I want to start by asking you a little bit about the news of the day, which is, of course, the enormous performance of a small number of technology companies in the U.S. – Nvidia first and foremost amongst them, and perhaps you could tell us a little bit about the role that Nvidia plays in the chip ecosystem as a way into this topic?

Chris Miller

Well, Nvidia is the world's leading producer of a type of chip called a GPU, a graphics processor unit, which was a type of chip that was actually invented for computer graphics. But it became clear around a decade ago that the same underlying math for computer graphics was also good for training AI Systems and Nvidea bet very heavily on converting this type of graphics chip to AI uses. And so today, when you look at sophisticated AI systems from the various GPTs and across the board, most of them are trained on Nvidea chips and the company has not exactly a monopoly, but something pretty close to one in this market, both because its chips are extraordinarily high quality, but also because it's got a software ecosystem around its chips that provides a really strong competitive moat. And so it's become the first trillion dollar chip company thanks to its absolutely central role in training AI systems. In the past couple of months as we've seen a boom in interest in generative AI there's also been a huge shortage of servers with Nvidia chips inside as even the world's biggest tech companies like Amazon and Google struggled to get their hands on all of the GPUs that their customers are demanding.

Luke Bartholomew

So, there's a lot of themes in there, Chris, that we want to get to in the course of the podcast. But perhaps a good place to start is something that Paul was alluding to there in terms of the number of tiny transistors on the silicon chip. And I guess that's the technological marvel that is the modern semiconductor and the now subatomic engineering that's required to make Moore's Law hold.

So perhaps you can just talk us through how advanced chips have become and what is now required to keep that Moore's Law thing working.

Chris Miller

Well to fit the billions of transistors on a chip the size of your fingernail, each one of them has to be roughly the size of a corona virus. And so manufacturing them requires a level of precision that's completely unparalleled in basically any other industry, any other type of mass production. And to do this has been the struggle of the past 70 years to keep Moore's Law alive by doubling the number of transistors per chip every two years, which means shrinking the size of them dramatically every couple of years. And today there's a very complex and extensive ecosystem of companies that produce the software, the ultra-specialized materials, the precision machine tools, and then know how to bring all these different factors together in manufacturing to do this manufacturing with high levels of precision and accuracy. And this supply chain stretches from Europe to the United States to parts of East Asia, like Japan and Taiwan, where you have companies that have specialized in just one facet of this production. And each of these companies, because of Moore's Law, has to roll out a new set of products every couple of years, a tool that's even more advanced or materials that are well-suited to the next manufacturing process. So they all face this relentless pace of innovation, which only a couple of companies have managed to keep up with.

Luke Bartholomew

So the way you describe the supply chain arrangements there and indeed the way you explained it in the book brought to mind a parable that the economist Milton Friedman used to tell - you know, hold up a pencil and say, no-one in the world knows how to make this pencil. It has the word and the graphene and the brass and the rubber, and it relies on this complex division of labour and globalized supply chains and I guess Friedman's point is that even something as humble as the pencil requires this degree of complexity. But, you know, it seems to me that the silicon chip really is the acme of that kind of globalized supply chain process. So maybe you could talk a little bit about the history of how it came to be that we have this highly intense division of labour across the global economy?

Chris Miller

Well, it's a key question because it's a way the industry has changed a lot over the last couple of decades. When the first chips were invented in the late 1950s, there really was no supply chain because chip companies tried to do it all in-house. They designed chips, they manufactured them, they fabricated their own silicon wafers, they built their own machine tools, they invented their own design software. But as the process has become more and more complex, there has just inevitably been specialization across the supply chain. And so today there's not a single company, nor is there a single country that can undertake every step of the chip production process. And you can go even deeper than that within the most advanced chip makers like TSMC in Taiwan, there's not a single person that understands TSMC's production processes because there's a thousand or more process steps involved in chip manufacturing when you're inside of a FAB, which is the what you call a chip manufacturing facility, and there's not a single person at TSMC that understands all 1000 of those processes because they're all themselves extraordinarily complex. So, yes, this is taking the pencil analogy and making it a thousand or a million times more complex. And the impact of that has been that for companies to enter the chip industry at any segment of the supply chain, the barriers to entry are really quite large because there's all sorts of really unique types of knowledge and often not intellectual property, but tacit knowledge that's built up inside of companies that is very hard to learn or to replicate outside of advanced companies. You can't really study it in universities. You’ve got to learn it from doing it inside a chip maker. And that has made many segments of the chip industry extraordinarily durable in their market structure because these tacit knowledge requirements present real barriers to entry.

Paul Diggle

So Chris, the comparison that people often draw is between oil and chips, you know, foundational to the economies of the 20th and 21st centuries. And with important political and geopolitical roles as well. Indeed we've done episodes of this podcast about the geopolitics of energy. A huge case of that is, of course, playing out in the world over the past 18 months in the Russian invasion of Ukraine. But I thought we could perhaps talk about the ways in which that analogy doesn't hold because that sheds some light on the uniqueness of chips as well. You know, the concentrated production, the specialism, the choke points. Why don’t you flesh them those out for us to tell us why chips are even more special than in their role than oil played previously?

Chris Miller

Yeah, and I think you're right that the oil analogy is correct in pointing out the economic criticality and the entanglement in geopolitics. But the oil industry is actually a lot less concentrated than the chip industry. If you look at advanced processor chips, 90% of them are produced in Taiwan by one company in just a couple of factories, whereas the largest players in the oil industry, Saudi Arabia, Russia, United States, produce, you know, less than 15% each of the world's oil. And so there's much less concentration in energy. The other aspect is technology. You know, on the one hand, oil companies are relying more than ever before on technology for their exploration, for optimizing drilling processes, for the refining process. On the other hand, there's just no comparison to the rate of technological improvement that the chip industry has seen. There's no other segment of the economy, no other aspect of human endeavour in history that has increased at the rate the chip industry has increased - the number of transistors per chip doubling for over half a century. Now, I like to think of other analogies like what if airplanes flew twice as fast every two years? We'd be at six times the speed of light by now. In most industries, put in a productivity improvement of 3 or 4% a year you've succeeded. In the chip industry, 3 or 4% would be an absolutely abject failure. And so that technological aspect is a huge differentiating factor from the oil industry, which is why there's a lot of companies that can more or less effectively drill for oil in many different countries with some marginal differences in efficiency, whereas there's just a couple of companies that can produce not just chips, but every aspect of chip supply chain has roughly comparable levels of concentration in the hands of just a couple of companies.

Paul Diggle

I suppose another way of expressing that is that there's fungibility in the oil market, you know, different grades of crude, while they may be different and you'll have to change your refinery to deal with them, ultimately you can replace, you know, one with another. But actually there isn't fungibility at the cutting edge of the chip industry because the smallest transistors on the most advanced chips really are kind of something different from further from the productivity frontier.

Chris Miller

You know, that's absolutely right. And even at the lagging edge, less technologically advanced chips is not the case. You can substitute one for another. If you look at the chip shortage the last couple of years in the auto industry and elsewhere, the problem wasn't a deficit of chips globally. In numerical terms, we actually produced more chips in 2021 than in 2020 and more chips in 2022 than 2021. The problem was that the specific types of chips that car companies needed, they couldn't get access to and many of the world's chip makers couldn't produce that type of chip because their factories weren't tooled for it. So, yes, the market is much less liquid because there's much more product differentiation. And you can't just take an iPhone chip and plug it into a computer and hope that it works because it won't.

Paul Diggle

And you've mentioned TSMC, of course, this Taiwanese company crucial to the chip ecosystem. And I want to talk then about the US-Taiwan nexus of chip interdependence because that's a crucial part of the story you tell in the book. And specifically do you think of that nexus, that interdependence as stabilizing or destabilizing? And I have in mind the idea of, say, weaponised interdependence. Does that actually lower the risks because the stakes are so high for everyone there? You know, is there a silicon shield over Taiwan or by contrast, to stretch that analogy, as silicon sword of Damocles? How do you think through that issue?

Chris Miller

Well, it all depends on your assessment of how China's leaders make decisions. And I think it's always been difficult to know and it's gotten more opaque the last couple of years. I think the silicon shield thesis is most likely to hold if you think the most plausible military scenario is sort of a D-Day version 2.0, where China tries in a massive invasion to take the island. And if it were to do so, it would inevitably trigger a vast war, knocking your making facilities offline and plunging the world's manufacturing economy into Great Depression levels of disruption. That would be bad for everyone, bad for us, bad for China. But I don't think that's the most likely scenario. I think the most likely scenario if China were to decide to escalate militarily, and again, I hope it doesn't, I don't think it's guaranteed to do so, but I think it's a non 0% and probably above a 10% probability over the next decade. If you're sitting in Beijing, your very strong incentive is to try to take Taiwan without triggering a US response. Because if the US gets involved, you end up with a massive war. If the US doesn't get involved you might get away with small one or even no war at all if you can force Taiwan to fold without fighting. And that's where scenarios like blockades look much more appealing from the perspective of Beijing. They're not easy, but they're much more likely to induce Taiwan's surrender without fighting. And if you put yourself in the shoes of the US President, and suppose China implements a partial blockade of Taiwan or not even stopping the flow of products out of Taiwan, but just imposing, let's say, mandatory customs checks or phytosanitary checks by Chinese Coast Guard ships on shipping into Taiwan's harbour. What would the US do? Well, it's not going to sanction China because sanctioning China would have just as large costs on the US and Europe and Japan as it would on China. And I think the experience of the Russia-Ukraine war has actually helped some U.S. leaders realize that sanctioning China is just a non-viable option. Would it use military force to stop China from imposing customs checks on ships going into Taiwan? Well, it could, but that seems like a highly risky endeavour, because if you choose to use force and China doesn't back down, you've got a war and an economic disaster on your hands. And so you find yourself somewhat like in the position of Khrushchev during the Cuban Missile Crisis, except that now you're fighting over an island that's 100 miles offshore of your adversary, and you're in the position of having to decide whether to risk nuclear war and economic catastrophe, to break a blockade. In that scenario, I think the fact that the US coming to Taiwan's aid would entail massive economic risk of disrupting these supply chains actually ends up deterring the US from helping Taiwan.

It's not guaranteed to, but it would be a factor that would push the US to not get involved. And so actually I think that increases the risk because Chinese leaders are running through these scenarios in their heads as well. They realize that the US does not want to see a disruption in trade. And so if China can take steps that compromise Taiwan’s sovereignty and undermine the credibility of US security guarantees without disrupting trade and give the US the choice of choosing to disrupt trade. In response, the US might choose not to, and if so, the implications for Taiwan security would be dire.

Paul Diggle

And so Taiwan sits on a geographical fault line, you know, a geological fault line as well as a geopolitical fault line, as you've been explaining. What would be the economic consequences? You started to talk about them. What would be the economic consequences of Taiwanese chip production going dark, and disconnecting from the world economy?

Chris Miller

 Well the easy part of the answer is to look at high-end chips, the chips that go into smartphones, PCs, data centres, telecoms infrastructure. You know, around 90% of the world's application processors or smartphones are produced in Taiwan. So you basically have no smartphones produced anywhere in the world the next 6 to 12 months. PC production would fall by half because of the large number of PC processors produced in Taiwan. Data centre rollout would grind to a halt because many of the chips in data centres, both GPUs that we talked about for AI, but also networking chips in data centres. So in the tech sector you'd see massive impact. And if you look at all of the world's companies with trillion-dollar valuations, none of them can operate without chips from Taiwan. So you can write off Silicon Valley for the short run. But I think actually that the larger economic impact would be on the rest of the economy because Taiwan produces almost all of the world's high-end processor chips, but it also produces a ton of lower-end processor chips and these lower-end processor chips go in cars and dishwashers and airplanes and coffee makers. And as we saw during the pandemic, when a very small deficit of a tiny number of chips caused several hundred billion dollars of disruptions in auto production, if we were to have a major decline in the availability of low-end processor chips, the implications for world manufacturing would be, you know, I think it wouldn't be an exaggeration to call it catastrophic. Taiwan produces 20 or so percent of the world's processor chips. Every car has at least one and often dozens of chips made in Taiwan inside. And so actually the challenge we face if we're to lose access to Taiwan's semiconductors, would be to make dumber devices. We’d be pulling out the blueprints of the 1990s for refrigerators and dishwashers and cars and trying to find people who remember how to make them without more advanced computing capabilities. And it would take a long time to retool factories to do that. And in the interim, manufacturing production would plunge.

Luke Bartholomew

So, Chris, as a consequence of that strategic logic you laid out there and the dire economic consequences, we've had, the US Chip Act and increasingly other countries are trying to get in on the act, so to speak, with their own industrial strategies and ambitions to become major chip producers. And I know it's early days so far, but do we have a sense of whether the CHIPS Act is having that desired effect and more broadly, what the prospects are of chip design and production moving away from its highly concentrated current order?

Chris Miller

Well, I think we've certainly already seen a major increase in investments in chip manufacturing in the US and Japan, even in Europe I think we're starting to see that. But that's sort of obvious that if governments are going to fund chip-making, you get more chip-making as a result. I think the harder definition of success is will this continue after the subsidies are stopped. The US government, the Japanese government, others, are hoping that if they can catalyse a bunch of new investment in the next five or so years, that will take on a self-sustaining character and firms will voluntarily decide to keep investing over the long run, even if there aren't such a large scale of subsidies. And this is possible. We see that there's a lot of path dependencies and where companies decide to invest. And if you have a reinvigoration of the chip manufacturing ecosystem in the US, for example, you'll get a larger labour force and more ecosystem players and so that could lead to larger long run trends in chip investment. But I think it's far from guaranteed because the cost structure of making some manufacturing factories in the US and Europe and to a lesser extent Japan is just meaningfully larger than I mean, if we hired in Taiwan or in Korea 2030, 40% more expensive to make in the US versus in Taiwan. And so in the absence of long run subsidies, I think it remains to be seen whether the US or other countries can continue to have this current elevated level of investments in chip manufacturing in their countries.

Luke Bartholomew

So then finally, and coming back full circle to AI systems and I guess the performance of Nvidia, I wanted to get your thoughts, Chris, on whether the systems themselves as distinct from the physical chips, could also become an area of geostrategic competition. So what I have in mind is, for example, the dominance in the design of these systems or embedding a certain set of values inside these systems, is that going to become the next technological arms race?

Chris Miller

You know, I think we're already seeing political leaders in multiple countries trying to explore what levers they have over the AI ecosystem. Setting aside the hardware. And there have been some recent headlines you may have seen that Microsoft is closing its AI research centre in China and moving or offering some of the researchers visas to Canada. And that's a very clear symptom of this broader role that political factors are playing in AI. But I think what you'll find is that political leaders have much less leverage over the rest of the AI stack, if you will, and they do over the hardware because it's much more diffuse. It's much harder to regulate, whereas the chip industry is concentrated and easy to regulate because there's just a couple of factories, a couple of facilities in the world that you need to control, and a couple of companies that produce the machines that go in those facilities. And they don't really face short-run competition. In the long run maybe they do if you regulate them too heavily but in the short run, political leaders have realized they can sort of do whatever they want, because if you want advanced chips, you need to go to TSMC. And if you’re TSMC, you need to buy machine tools from a small set of companies in the US, Japan and the Netherlands, and that has given politicians a ton of leverage over these companies. It's made the companies themselves pretty unhappy. But these companies find themselves in interesting position. For a long time, their shareholders realized that they had extraordinary market power, which gave them very high valuations and then politicians realized that they had extraordinary market power, which opened them up to being used as tools in this geopolitical game. I think that's somewhat unique to the chip industry and the hardware aspect of the AI stack. I think the rest of the AI stack, if you will, will be much less prone to that level of concentration and therefore the regulation will be much less effective than planned.

Luke Bartholomew

So I think that is all we have time for this week. Chris’s book is ‘Chip Wars: The Fight for the World's Most Critical Technology’. It really is a fantastic read, and I think crucial to understanding so many aspects of what's happening in the world, both economically and politically, and we highly recommend it. And so all that remains is for me to thank Chris for joining us today.

Chris Miller

Well, thank you. I appreciate the invitation.

Luke Bartholomew

And to thank you for listening. So thanks very much and speak again soon.

This podcast is provided for general information only and assumes a certain level of knowledge of financial markets. It is provided for information purposes only and should not be considered as an offer investment, recommendation, or solicitation to deal in any of the investments or products mentioned herein and does not constitute investment research. The views in this podcast are those of the contributors at the time of publication and do not necessarily reflect those of abrdn. The value of investments and the income from them can go down as well as up and investors may get back less than the amount invested. Past performance is not a guide to future returns, return projections or estimates and provides no guarantee of future results.