← The HC Commodities Podcast
The HC Commodities Podcast
Chip War: how semiconductors became the new oil with Chris Miller (summer repeat)
00:55:12
Semi-conductors, microchips, are the new oil - the scarce resource on which the modern world depends. Today, military, economic, and geopolitical power are built on a foundation of computer chips. This has only accelerated with the emergence of AI. Yet, the supply chain is incredibly concentrated with just a handful of countries and companies dominating. As with oil in the previous century, chips sit at the center of geopolitics, great-power rivalry and trade, especially between the US and China. How did chips become some crucial and their supply chains so concentrated and what does it mean for all our futures? Our guest is economic historian Chris Miller, author of Chip War - The Fight for the World's Most Critical Technology, the New York Times best seller.
Loading summary
Transcript45 lines
- [00:00]
Chris Miller
Foreign.
- [00:06]
Paul Chapman
Welcome to the HC Insider Podcast, a podcast dedicated to the commodities sector and the people within it. I'm your host, Paul Chapman. Today we're talking semiconductors. The chip semiconductors are absolutely crucial to the global economy. They're in our phones, our fridges, and our cars. Covid laid bare what happens when there's a disruption in the supply chain, and as we'll discover, that was a consequence of the wrong chips being made as opposed to a cessation of manufacturing. The supply chain itself is uniquely concentrated. Only a handful of companies and countries around the world design and manufacturer semiconductors. The pace of change is frenetic, and keeping up with it is crucial to companies and countries alike. As the world starts to understand the power of AI, both in civilian and military applications, it's unleashed a whole new wave of demand for the latest GPUs. Yet since inception, semiconductors have been at the center of geopolitics, geostrategy, espionage, cold wars, and fierce economic competition, making for a fascinating story and a consequential one for world security and economic development. Our guest is Chris Miller. Chris is the Associate professor of International History at the Fletcher School of Law and Diplomacy at Tufts University. He's also the author of Chip the Fight for the World's Most Critical Technology, published in 2022, a phenomenal book on the history and current state of how semiconductors are the new oil of the 21st century. I also want to note that we have an upcoming HC Insider podcast live event, this time on September 14th in central London. Hosted by Onyx Capital Group. We're discussing the future of oil derivatives and who really prices oil today. The panel consists of myself, moderating Greg Newman, CEO of Onyx Capital Group, Savas Menusos, former Head of Trading at sepsa, and former guests Kurt Chapman and Tor Sveland, Founder and CEO of Sveland Capital. The event is free, but invitation only and spaces are limited. So you have interest in coming along and seeing the panel, Please do email me or reach out via LinkedIn. If not, you'll be able to hear the panel discussion on a future episode of the podcast. As always, you can really support the show by leaving a positive review on the platform you're listening on, or just spread the word. Please let your colleagues know about the podcast if you enjoy it, which only helps us to continue to get great guests on to talk about powerful topics in and around the commodities sector. And finally, as always, I hope you enjoy this episode. Chris, welcome to the show.
- [03:02]
Chris Miller
Thank you for having me.
- [03:04]
Paul Chapman
I'm excited to have this Discussion. We're basing it on your book Chip War, which is all about semiconductors, the integrated circuit, and essentially how it has become. They have become the crude oil of the, the late 20th and start of the 21st century, and now so pivotal in geopolitics, but crucial to the world's economy, which, as we'll discover and as we all know, Covid laid bare. But there's potentially worse in store, given the, the narrative that is playing out. Before we sort of start on this, this history, can you just get us all on the same page as to what we mean by a semiconductor? A chip.
- [03:43]
Chris Miller
So semiconductor, chip, integrated circuit, they all refer to the same thing, which is a small piece of material, usually silicon, with millions, or today often billions of tiny circuits carved into them. And these circuits flip on and off very rapidly. They're switched on and off via device called a transistor. And in doing so, they provide all of the ones and zeros undergirding all digital computing. Because today all data is remembered and processed by turning it into strings of ones and zeros. And each one is just a circuit that's on, each zero is a circuit that's off. And semiconductors are what provide all these digits undergirding all this computing.
- [04:24]
Paul Chapman
Perfect. Okay. And this all starts right at the end of World War II, when the utility of computers was obvious. But they were using vacuum based transistors at the time, and there was a need and an insight from some key people that you could. There's another way of doing this. You could miniaturize a computer and thereby gain in function and shrink in size. Can you just. I mean, this is the story, or the start of the story about how Silicon Valley became Silicon Valley, where, you know, can you start off with Texas Instruments and the first chips and how it ended up that Silicon Valley is and was the epicenter of this entire technology.
- [05:06]
Chris Miller
Well, it was the US Military that was actually the first big source of demand for chips when they were invented. The US military and the space program wanted to miniaturize computing power, to put it in the nose cone of missiles or into rockets that they were launching into space. And there were a relatively small number of advanced electronics firms in the US that were capable of producing this type of really sophisticated electronics at the time. Some in the east coast, some in Texas like Texas Instruments, but a number in the towns south of San Francisco that would later come to be known as Silicon Valley. And they were founded there partly because there was a pretty deep reservoir of expertise in aviation, radios, and over other high tech industries that were focused in, in the Bay Area. So there was partly some structural factors that were attracting these early tech firms to Silicon Valley, but there's also some pretty idiosyncratic factors, like the fact that the inventor of the transistor, William Shockley, his mom lived in Palo Alto, which was at the time a very small suburb of San Francisco. Today, of course, is the epicenter of the tech world. But it was as early as the late 50s that Silicon Valley was on course to become the center of the world's chip industry and therefore the epicenter of all of the world's most advanced computing.
- [06:28]
Paul Chapman
And it's fascinating, isn't it, and kind of instructive as well, considering DARPA's budget today is so minuscule compared to its impact back then. But how pivotal the us, the military, as well as NASA and the broader Cold War was in actually getting it going. Right, because this is before personal computers came along. So really they were the major consumer of this technology. And already we're seeing this as a technological edge that would give them advantage over Russia's hordes of tanks and men, as it was seen at the time.
- [07:01]
Chris Miller
Yeah, you know, that's right. And it's interesting when you put it in the context of what we normally think of as R and D budgets, research and development. And the reality is that universities do a ton of R, lots and lots of research, but most of it's not very practical. And so it informs technology, but it doesn't directly produce it. And then companies do a lot of D, they develop new products. But usually companies are willing to invest in development when the science is basically already sorted out and they just need to make some tweaks and test new market applications. But there's actually generally a small number of institutions that are willing to translate R into D, take cutting edge technology that hasn't been fully proven and build prototypes to see if it'll work once or twice or a dozen times. And the military was an example of a institution that was willing to do that at a very large scale in lots of different technology areas during the Cold War, because they had a budget that was in all practical terms, unlimited. They had demand for really exquisite capabilities when it came to cutting edge technology. And so for lots of startups that were capable of producing a small number of very highly priced high tech products, the military was their only potential customer. And that's exactly where early chip firms found themselves. They knew they could make a number, a small number of chips. They could do so at A very, very high price point. And it was only the military that was willing to buy the initial chips. But the fact that they were able to find an initial customer in the Defense Department department or in NASA gave them the Runway. They needed to learn how to drive down the price, hone their manufacturing processes, and pivot into a mass market device. And so that's really the critical role that defense procurement played in, in setting off the chip industry and its, its trajectory of becoming a product that today is mostly used in civilian applications.
- [08:50]
Paul Chapman
Yeah, And I just want to spend a quick moment on this because then we're going to expand it in a minute when we talk about the supply chain today. But even at an early stage, it wasn't just the complexity in the design and theory and all, you know, the physics and mechanics of these things. It was also the complexity in actually manufacturing. Right. It was actually translating these transistors in ever increasing numbers onto slithers of silicon, which required masking. And the book does a wonderful job of helping you understand the manufacturing process and all these amazing characters in the 50s and 60s behind it. And they were shifting companies. The Shockley's team all went to Fairchild and that was Noyce and Moore, more of Moore's Law, who then went on to found intel and so forth. And there's a fascinating study in incentives and equity and all that good stuff there. But can you just give us a. You know, when we're talking in the 50s and 60s, it was also that capability in manufacturing that was so critical
- [09:48]
Chris Miller
even then, you know, that's right. And the reason is that chip manufacturing has always been very, very difficult. And the best data point for understanding this is that one of the key variables in differentiating different chip companies and different products is what's called the yield, which is the share of chips that are produced that actually work. And in most spheres of manufacturing, like cars, the car companies don't produce that many cars that don't work. If they don't work right off the assembly line, something's gone horribly wrong. But in chip making, it's common to have yields that are 90% or even early stages of technology, 70%. And that's because the technology is constantly being developed and constantly being pushed forward. And so it needs to be honed to turn it from a prototype into something that's possible to manufacture at a, at a high volume. And what that's meant is that technology is not just something you do once in a lab. The chip industry, it's equally important to have something you can replicate by the thousands or tens of thousands, or in some cases by the tens of millions, because that's the scale at which you need to produce chips, functional chips that work in real life once they come off your production line. And so learning how to scale has been as important as any sort of advance in physics or material science.
- [11:07]
Paul Chapman
Yeah. And staying in this period, those middle decades, this is also already it starts to become seen as geostrategic. Right. And the Russians adopt this strategy of essentially, or the Soviets, I should say, adopt the strategy of copy. They unleash thousands of spies, an incredible sort of buying off the market operation to try and copy the US advantage in chip manufacturing. Even at this stage. They set up Zelenograd, I think, as their answer to Silicon Valley. But crucially, it doesn't work. And that plays out vital towards the end and in fact the winning of the Cold War. Can you just talk about that strategy and why didn't it work?
- [11:50]
Chris Miller
Well, the key to understanding why it didn't work is to go back to the concept of Moore's Law. So Gordon Moore, one of the founders of intel, in 1965, set out the prediction that the computing power of chips would double every two years or so. And that's held true all the way up to the present. That's a rate of improvement, doubling in productivity that's unmatched in any other sphere of modern economies. And what that's meant is that there's no industry in which it's so hard to stay at the cutting edge or to catch up to the cutting edge if you started out behind, because the cutting edge is racing forward at this exponential growth rate. And so if you started out behind as the Soviets did, and tried to copy what your competitors were doing, you'd end up copying technology that by the time you finished copying, was old. And that's exactly what the Soviets found themselves doing. They had a lot of success in getting their PhD students trained at places like Stanford, at smuggling back semiconductors into the Soviet Union, at smuggling in entire pieces of semiconductor manufacturing equipment. They had lots of smuggling going on. The problem was that that was an awful strategy in an industry that was changing so rapidly. Smuggling works in places like nuclear weaponry, where there haven't been any technological advances in half a century. But if technology is changing so rapidly, if you're copying, you're copying yesterday's technology.
- [13:13]
Paul Chapman
Yeah. And on the flip side of that, the US was actually also, in the interest of securing allies and keeping friends close and so forth, was pushing manufacturing to Japan as well. Right. There was this sort of. And in part this has always been a tandem of sort of government allowance and companies seeking more efficient lessons, less costly manufacturing. And it was seen that Asia provided a really good labor force for that, in particular Japan, subsequently Singapore and Taiwan, which kicks that whole story off. Can you just take us there?
- [13:45]
Chris Miller
If you think of the landscape during the early Cold War, 50s, 60s, 70s, what you had is, on the one hand, the US interpreted one of the causes of World War II as having been that the Japanese economy was insufficiently focused on success in economic terms. So the US wanted to make Japan economically successful to reduce the risk that Japan would turn again towards an aggressive foreign policy. And at the same time, in Japan, also in Southeast Asia, you had a lot of governments who had massive young workforces moving from countryside to city looking for factory jobs and had to find industries to employ them in. And simultaneously, you had US firms that were looking for the cheapest labor force possible to assemble electronic devices. You had a confluence of interests between the US Government, governments in East Asia and US companies, all of which agreed that having workers in East Asia play a bigger and bigger role in the chip industry and the electronics industry writ large was in their interest. And so the chip industry played a really substantial role in binding East Asian countries to the United States at the same time that East Asian governments utilized their growing role in the chip industry to bind the US to them. Because there were a large number of East Asian governments that said, well, we need US Military production. We can't be sure that the US military will protect us for our own sake, but if we've got a Texas Instruments plant or a Fairchild Semiconductor plant, we're going to be a lot more likely to get the protection that we want from Communist China. And so that's the alliance between Asian governments, U.S. government and U.S. firms that put Asia at the center of the chip industry from the very earliest days of the industry's foundation.
- [15:31]
Paul Chapman
By the early 1990s or the 80s, it was suddenly seen that the US had offshored most of the manufacturing. There's a, you know, there's fantastic, again, characters in the book, like the AMD guy who's, you know, one of the few that holds onto their own fab plants and so forth. And this was when the first idea that chips were the crude oil of the crude oil of the 1980s at the time. But there was serious concern about that being, you know, the weakness of having offshored all this manufacturing. Kind of, you know, a story that has come right back round again. But can you sort of Give us some sense of firstly how to think about the value chain of the creation of a semiconductor from design the tools, lithography. And then can you, I guess, tell us how? I think probably the easiest way to talk about this is TSMC in Taiwan. Give us some sense of how and why it ended up that very few US chip companies had their own fabrication capabilities.
- [16:32]
Chris Miller
Start with the value chain. If you want to make a semi hunter today, you need to begin by acquiring design software. Software that's, that's specialized in chip design. If you want to lay out 10 billion transistors on a piece of silicon the size of your fingernail, you need very automated processes to do so. There's basically just three companies that have an almost oligopolistic position in chip design software, all which are based in the US Then you actually need to design the chip to do what you want it to do. And there's lots of different chips that do different things. Some chips manage WI FI connections, other chips train AI systems, other chips remember data. Each of these types of designs involves a lot of really unique know how, such that there's only in most cases, a couple of companies that can design chips at the cutting edge in each of these different spheres and a lot of unique process knowledge that's retained inside of these companies, as well as a lot of unique intellectual properties. So chip design is actually the most value additive part of the entire chip industry. Companies like Nvidia only design chips, manufacture nothing, just do the design. Once you've got your design, you need to actually undertake the manufacturing of a chip. But this requires not just manufacturing itself, but also acquiring the machine tools that are needed to produce the microscopic circuits on, on your chip. And today there's a small number of companies that can produce this type of equipment. So you need equipment, for example, that can deposit thin films of materials, the thinnest of which are just four atoms thick, with basically perfect uniformity across a silicon wafer. Machines that can etch tiny canyons into your silicon wafer to fashion the shapes that become the transistors on your chip. Or the most complex of all are the extreme ultraviolet photolithography machines, which pattern the transistors onto your silicon wafer. And these machines involve the flattest mirrors humans have ever made. The most powerful laser ever deployed in a commercial device, and have an explosion inside of them happening at all times at a temperature of around 40 times the temperature of the surface of the sun. These are extraordinarily complex. Machine tools have hundreds of thousands of precision components, cost $150 million a piece. And so the precision tools themselves are just an extraordinarily difficult thing to produce. And they're produced as a result of this complexity by just a couple of companies. There's basically five companies that make up the lion's share of revenue in the machine tools segment of the business. Three in the us, One in Japan, one in the Netherlands. And if you don't have tools from all of these companies, you can't make an advanced chip. Full stop. Simply impossible. But none of these companies that make the tools can actually make any chips. And in fact, there's different companies like TSMC in Taiwan, the Taiwan Semiconductor Manufacturing Company, which is the world's largest chip maker, the world's most advanced manufacturer of processor chips, which brings together all these different ingredients to actually manufacture chips. And they've got unique know how that no one else right now can replicate. And so that's the value chain of chips from the design to the manufacture. And each of the firms that I mentioned only does its specific segment of
- [19:48]
Paul Chapman
the supply chain, which is just indicates just how fragile this is. Right. That one Dutch company, EUV Lithography, were that to have a collapse or whatever it might be, it would just have this profound impact on the, on that value chain. But let's stay with TSMC because it's going to be important a little bit later. How did that get started? And can you give us some sense of what their strategic goal was and how they came to acquire such a dominant share of manufacturing?
- [20:16]
Chris Miller
TSMC was founded in 1987 by a businessman named Morris Chang, who I think is one of the most underrated businessmen of the last hundred years. He was initially born in mainland China. He fled when the communists took power in 1949, enrolled at Harvard. At the time, he was the only Chinese student in his class, and then was hired to work at Texas Instruments just at the time that one of his colleagues at Texas Instruments was inventing the first integrated circuit. And he spent the first half of his career at TI and became known as a manufacturing expert, driving up yields on TI's production lines, finding efficiencies that made it possible to drive down the cost of chip making. And he was hoping to become the CEO of ti, but was passed over for the job, second in line and so left and was looking for something else to do. And he'd been thinking during his final years at TI about a new business model for the industry, because at the time, unlike today at the time, almost all chips were both designed and manufactured at the same company. That's what TI did. But Morris changing, realized that both design and manufacturing were getting much, much more complex. And he projected into the future and saw this complexity growing at a literally exponential rate. And he realized that in the future there would be extraordinary economies of scale that would accrue to the biggest chip manufacturers. And so it was better off to just focus on one part of the industry. And so sort of like what Gutenberg did for books. Gutenberg didn't write any books, he only printed them. Morris Chang wanted to do for chips. And so he founded TSMC with the business model of not designing any chips. It never has, never will. It's only manufactured chips, and that's let it manufacture for a huge array of different companies, for Apple, for Qualcomm, for Nvidia, for amd. And as a result, it's the world's largest chip maker today. And not coincidentally, it's the world's most advanced maker of processor chips. Because just as Morris Chang predicted in the 1980s, the more chips you produce, the more data you gather about your manufacturing processes, the more you hone your technology, the more advanced capabilities you develop. And so today, 90% of the world's most advanced processor chips, the types of processors in your iPhone, in PCs, in data centers, in telecoms infrastructure, 90% of these chips can only be manufactured at TSMC in Taiwan.
- [22:46]
Paul Chapman
We've actually got an episode coming out hopefully after this one on Taiwan's power grid, and we'll be talking about silicon shields very shortly. But Meanwhile, in the 90s into the 2000s, chips have moved from just not just the domain of the military and being a technological offset and a key strategic requirement to suddenly being ubiquitous in society, right from cell phones through to computers, PCs, the rise of the PC and have, you know, just became absolutely dominant and the scale of manufacturing and the types of. We had developments of new types of chips. You've mentioned some. Obviously we moved from logic and memory chips onto the analog chips that go into what Nvidia is up to and so forth. So suddenly these things become absolutely crucial to society and baked into everything that we have. The story here kind of leads on to China. And in 2010 you say in the book that China was spending more on semiconductors than on oil. And they kind of had this, I think in the wake of Edward Snowden's releases, had this epochal moment when they suddenly realized how reliant they were on US technology. Can you take the story on from there?
- [24:03]
Chris Miller
Yeah, well, that's right. And the data point that you Referenced that China spends as much money each year importing chips as it spends importing oil. It's not just for one year, that's every year for the past decade, China's spent that vast sum of money. And the reason is the following. It's that today almost every device of an on, off switch has at least one and often dozens or hundreds of chips inside. So just to give you an example, a typical new car will have around 1,000 semiconductors inside, some doing very simple things like moving the window up and down when you press the button, others doing very complex things like managing the autonomous driving systems. And policymakers in China began to realize, first off, that the entire world, the entire manufacturing infrastructure was becoming more dependent on semiconductors and that this trend was going to continue for a long time. That's the first thing they recognized. The second thing they recognized was that the economies of scale and the unique technological know how required for cutting edge production was driving concentration in the industry at every step of the supply chain. In the chip design software, in the design methodologies themselves, in the machine tools, in the chemical production. We haven't gotten to that, but we can in the manufacturing processes themselves, such that the chip industry was becoming more consolidated. And it was becoming consolidated almost exclusively outside of China, because China at the time, like today, doesn't play a big role in any of the cutting edge segments of the chip industry. And so Chinese policymakers realized their dependency was only going to increase just as their need for chips was also going to increase. And, and finally China looked at the map and realized that all of the world's advanced semiconductor technologies are produced in countries that are at best competitors and at worst geopolitical adversaries of China, Taiwan, the United States, Japan, South Korea chief among them. And so it's completely understandable that Chinese leaders looked at the situation and said this is completely unacceptable to have our economic future depend on a tiny number of firms located in countries that we can't trust.
- [26:12]
Paul Chapman
President Xi called forth the assault. Right. And I mean, it's a multi pronged strategy, all hands on deck, to essentially try and grab parts of that supply chain. You know, from, well, from espionage all the way through to what now looks quite, you know, naive to say the best, if not cynical at the worst of many U.S. manufacturers selling off IP to China along with other industries as well, in the commodities industry, we saw that as well. To gain access to those markets. Can you tell us what China unleashed to try and catch up?
- [26:47]
Chris Miller
I think to understand China's strategy over the last decade it helps first to differentiate between the Chinese party state and the tactics that it employed and the Chinese private sector and the tactics that it employed, because a decade ago, there really was a Chinese private sector, especially in the electronic space, that was more private than public, that was globally integrated, that was low on the value chain, had to be responsive to foreign customers, but nevertheless was really part of the global computing electronics ecosystem. The Chinese party state did not have faith that this private sector could succeed in improving its technological capabilities at the rate China wanted, and didn't have faith that it would do so in a way that made China more independent. And that was probably correct, because the companies that were succeeding in China in this sphere were doing so not by becoming independent, but becoming increasingly interdependent, making themselves more deeply embedded in ultimately US Centric and Taiwan centric supply chains. And that was economically a smart move, but it was politically unacceptable. And so the Chinese party state responded by pushing an array of alternative strategies to try to acquire technology in the chip industry by stealing technology, by bringing bog cases in kangaroo courts against foreign chip firms, and trying to buy outright major foreign chip companies, and importing vast sums of money into the domestic chip industry in China, which some of which went to legitimate companies, many of which went to outright frauds, and hoping that this kind of combination of different methods to build up technological capabilities would succeed. And I think the striking thing about it is that actually the success stories in China's chip industry are actually still, generally speaking, companies that have leaned more to the side of being more integrated, more private sector oriented. And there's been actually a negative correlation between, I think, the amount of money companies have received from the Chinese government and their success in actually winning market share. But the impact on the rest of the world, which is looking at this huge wall of money flowing to the Chinese chip industry, looking at Chinese firms and often firms that were disguised so that they didn't look like they were under the control of the Chinese government, trying to buy up really advanced technologies outside of China. That set off a counter reaction in Taiwan and Korea, Japan, the US eventually in Europe as well. And government said, wait a minute, if China thinks this technology is so strategic, shouldn't we also be reassessing whether or not this is just a technology that the market should decide, or whether instead, perhaps government should play a role in setting priorities and in drawing boundary lines between what's acceptable and what's not?
- [29:37]
HC Insider Podcast Announcer
The HC Insider podcast is brought to you by HC Group, a retained search intelligence and advisory firm focused solely on the global energy and commodity sector, with six locations across Asia, Europe and the Americas and over 50 consultants. To find out more, go to our website, HCGroup Global. There you can also sign up for our HC Insider content for more interviews and white papers on relevant trends and talent impacts in the commodities world.
- [30:11]
Paul Chapman
Could you tell that story through Huawei? We spend a lot. You spend a lot of time on that in the book. And it's sort of a fascinating story that most of us have heard about the US Government's wranglings and trying to get Europe not to use it. And so can you help us understand Huawei and this interesting guy? We're going to get it wrong, But Ren Zhengfai?
- [30:30]
Chris Miller
Yeah, I think Huawei is a really fascinating company. I think it's a company that had Chinese politics gone in a different direction 10 years ago, we could be talking about Huawei the way we talk about Samsung or Sony or any of the other big electronics firms that have emerged in other East Asian countries over the last couple of decades. Because on the one hand, Huawei's been a very successful firm when it comes to devising effective products at competitive prices that win not only in China's protected domestic market, but also in export markets. But on the other hand, Huawei found itself deeply entangled in both Chinese and US Government efforts to assert control over the chip industry and the computing industry. And Huawei's primary businesses over the last decade have been, first, devising telecoms infrastructure, which is how most data is now transferred from your phone, for example, to data centers. And second, in making smartphones, where China, where Huawei was very successful in moving up the technology ladder and devising not only more and more successful smartphones, but also more and more sophisticated chips on those smartphones. And so in 2020, if you look at TSMC, the Taiwanese firm's two biggest customers, they were first Apple and second, Huawei, because Huawei's chip design capabilities have become really quite substantial. And this made the US Government, Japanese government, Australian government extraordinarily nervous. Not just that Huawei's telecoms equipment might be used for spying. The reality is history suggests that all telecoms equipment is used for spying. So it's, it wouldn't be unique if China were spying the Huawei's telecom equipment that would just be the standard. But I think that the bigger issue was that as it became clear that the political relationship between China and, let's say, the west was getting worse, the the chance that China would be able to have one of its firms be a core player in the world's technology infrastructure began to seem a lot more dangerous. And so the US Forced TSMC to stop producing advanced chips for Huawei and forced a number of other chip firms across the world to stop providing chip technology to Huawei. And because Huawei depended on the only supply chain that exists, the existing international supply chain, there was no alternative to turn to. And so Huawei was just forced to stop producing many types of advanced chips for phones, for example, and had to divest much of its smartphone business. And so it was the first large scale victim of the accelerating tensions we've seen in the chip industry over the past several years.
- [33:11]
Paul Chapman
Yeah, and the US was relatively late to this realization. Right. There was sort of latter part of the Trump administration. There was this idea that suddenly chips is everything we're competing on, which was the phrase used at the time. And a realization how crucial it was. That shows incredible stroke, right, to be able to tell TSMC to turn off Huawei. Can you give us a bit more insight into how that happened? That seems a lot of leverage.
- [33:36]
Chris Miller
Well, it is a lot of leverage. It's leveraged because the semiconductor supply chain is so concentrated. So there's in any semiconductor manufacturing facility anywhere remotely close to the cutting edge, not just this year's technology, but in the last 10 years of technology and probably in the last 20 years of technology, you're guaranteed to come across at least one and often multiple pieces of US made equipment, US made software, US implicated IP and materials. And that's true for Japan as well. Japan is just as central in the, in the manufacturing process of semiconductors. And because there aren't alternatives, or the only alternatives are other US or Japanese firms, the reality is that those two countries have immense say over who gets access to advanced semiconductor manufacturing. Because Taiwanese firms, Korean firms, any firms in the world that want to make advanced semiconductors need regular access to these types of tools, chemicals and software. If you can't get it, you just can't produce advanced chips. Which means that when the US says to Taiwan or anyone else, stop producing for Huawei, chip makers have no choice but to be responsive, because if they don't, if they're not responsive to the US regulations, they can't get the tools that their business depends on.
- [34:50]
Paul Chapman
Yeah, we haven't mentioned, as you said, but Japan obviously provides a lot of the chemicals and the gases that are so crucial as well. It's all very much interweaved. So then we're sort of coming right up to the present moment, but the COVID then strikes and that's really when it kind of enters the zeitgeist that, you know, how fragile these or how concentrated and therefore how fragile these supply chains are and the devastating impact it has on the global economy, even for short outages of production. Can you take us to what happened in Covid?
- [35:20]
Chris Miller
Well, I think a lot of people, he heard about chip shortages during COVID and thought there was a problem with chip supply. And that's not really actually what happened. The world produced more chips in 2020 than it had in 2019, more chips in 2021 than it had in 2020. What really happened was that the types of chips we demanded were unexpected. And as a result, certain types of chips were low in supply. And the reason was that in the early stages of the pandemic, a lot of consumer goods companies and car companies in particular, thought we were headed towards a new great Depression and so slashed their component orders because they thought they're going to be producing fewer cars. And that demand was largely made up by PC makers and data center infrastructure providers who, as everyone started working from home, got huge orders for new computers. And so chip makers began to reallocate their production processes in a way that when car companies went back to chip makers and said, hey, actually we're going to produce a fair number of cars this year. We need chips for them. There was no spare capacity available. And car companies, I think for a long time I thought about their key components as being their engines, their drive trains or other components when you think about cars that you think about. But in fact, as I mentioned earlier, a typical new car is a thousand chips inside, and it needs most of those chips to operate. Maybe you can operate without the chip that moves the window up and down, but you can't operate without the chip that manages fuel injection into your engine. And it was often because of a chip that just cost a dollar or two, often just one of the thousand chips that was necessary, that car companies had to leave $30,000 cars stranded in the parking lot outside of their plants. And so it's been estimated that the car industry alone suffered several hundred billion dollars in lost sales due to chip shortages in 2021, 2022, which I think brought home to most people the extent to which that chips aren't anymore just about smartphones or PCs. They're about basically everything. The entire world economy has become dependent on chips in a way that before the pandemic, hardly anyone realized.
- [37:28]
Paul Chapman
What Covid certainly did was catalyze the understanding at the political level, at the governmental level, and the people Level how crucial they are to the world's economy and how concentrated they are. Can you help us untangle, you know, the subsequent Chips act in the U.S. the trade wars that, you know, the entity list, the banning of key technology going to China, particularly, you know, these latest GPU chips that are crucial for AI. Can you just help us understand that whole landscape taking us up to where we are today?
- [38:00]
Chris Miller
Well, during the Trump administration there was a series of moves on individual companies or individual types of technology. So we mentioned Huawei ZTE is another Chinese company that was targeted. But during the Biden administration, these different measures were brought together in a more comprehensive way. And they weren't just done unilaterally, they're done trilaterally with Japan and with the Netherlands, which are the two other companies that are the major producer of machine tools to make advanced semiconductors. And so last October, October 2022, Biden administration released a new set of export controls that banned the export of semi hunter manufacturing equipment. Not only the most cutting edge, but what I describe as the second most cutting edge types of equipment to China. In addition, the US also banned the sale of the most advanced chips for AI training to China. And the rationale was the following. The US government believes, and I think justifiably so, that AI is going to be critical not just for ChatGPT, but also for defense and intelligence uses going forward. In the Russia, Ukraine war, we already see evidence of this. And the US government also concluded after studying the developments in AI that the key advantage that the US had, or at least one of the key advantages, was in the processing power needed to train AI systems. So if you look, for example, at the amount of data used to train cutting edge AI systems in the last decade, what you find is that it's been doubling roughly every six to 12 months. So you need twice as much data every six or so months to train the next generation cutting edge AI system, which means you need twice as much processing power to train these systems, which means having access to the most advanced chips is absolutely fundamental to being able to train AI systems in an economically competitive way. And so the U.S. looked at the landscape for AI chips and said, well, 90% of AI training uses chips produced by one company, Nvidia, California. All advanced AI chips are manufactured either in Taiwan or in South Korea. All advanced AI chips are manufactured using machine tools produced by the us, Japan, Netherlands. They're all designed using software tools produced exclusively by US firms. All the chemicals are largely, almost all the chemicals are made in Japan, some in Europe and the U.S. and so the entire supply chain for the production of the processing power which all AI requires is fundamentally a US or Western dominated supply chain. Why the Biden administration asked, would we give China this cutting edge technology when we think it's going to be critical not just for consumer uses, but also for strategic uses? And so they decided not to. They cut off China from the most advanced AI chips. And so today it's illegal to import cutting edge AI chips into China and only firms outside of China can access them. And in addition to that, the Biden administration went a step further and said in order to make sure our advantage is retained, we're going to stop selling the tools that make chips to China as well, so that China can't find an easy workaround by producing its own manufactured AI chips. And the administration convinced the Netherlands and Japan to implement roughly comparable rules. So today China can't access any of the cutting edge and really any of the second, second generation, most cutting edge tools. And therefore it can't produce cutting edge AI chips because it doesn't have the tools domestically. And foreign manufacturing houses like TSMC in Taiwan aren't allowed to produce these types of chips for Chinese firms.
- [41:33]
Paul Chapman
We're going to come back to Taiwan, but just with the chips act though at the same time the US government is trying to incentivize and with hefty incentives, reshoring manufacturing. With everything that you've spoken, I mean, there's been some big announcements on various plants from intel and AMD and so forth. Is that realistic? And I'm assuming given what we're about to talk about, the Silicon shield for Taiwan, you know, TSM scene, the like, are probably not very keen to give up their manufacturing dominance given the security that it inherently provides.
- [42:06]
Chris Miller
I think if you, if you listen carefully to what the US government is trying to create, it's the following. First they want to a more diversified manufacturing base for semiconductors. No one in a position of decision making authority in the US is looking for independence when it comes to semiconductors. It's just not feasible given the dollar values we're talking about. There's so much manufacturing in Taiwan, Korea, so much of the assembly and packaging, the back end in Southeast Asia, that no countries can be independence but independent. But the US does want less concentration, especially considering China, Taiwan risk. And in addition the US wants to create a bit more of a level playing field. There have been a number of great studies done by the oecd, for example, over the past couple of years which found that many governments were subsidizing chip industries. And the US was actually far behind in terms of the level of subsidies provided. In fact, there were in an OECD study that was done in 2019 that was found that there were US firms that received more money in subsidies from foreign governments than they did from the U.S. government. And so in that context, it's easy to understand why you had offshoring of chip making if other governments were just offering more generous subsidies. And so partly the CHIPS act is just designed to level a playing field that was already not very level. It was not level in the US favor. But I think you're right to suggest that the rest of the countries that have major chip industries are looking at the US CHIPS act and saying this is an opportunity, but it's also a threat. And indeed, I think you've seen every other major economy, Europe, Japan, India, South Korea, respond with their own chip subsidy programs. But I think to just to nail down the chronology of how this was happening in 2014, China announced that chips were a core technology, and as you said, they would assault the foundations of technology dominance. And then everyone else's chip subsidy programs came after that point. I think it's critical to get the chronology right in our heads because it helps us understand the causality of the subsidies. The US didn't think this up. This was a response to even larger subsidies that have been in place in China for much of the last decade.
- [44:25]
Paul Chapman
Yeah, and again, it kind of comes back to the whole reason that the Silicon Valley was created was because of US military support and incentives in there to launch it right alongside phenomenal universities and entire, you know, research complex. Let's go to Taiwan. Before we sort of put our game theory hat on here for a moment, can you give us some sense? Because I just, I just don't think it's sort of widely realized or understood if tsmc, for whatever reason, were ceased to manufacture tomorrow. You know, whether the power's gone out or whether it's been whatever it might be, I don't even talk about it being destroyed. I mean, this would be the global financial crisis on steroids. Right? I mean, this would be an apps. This would cause a worldwide depression. Can you give us some sense of the impact on the economy it would be to lose tsmc.
- [45:18]
Chris Miller
Well, without chips made in tsmc, it'd be basically impossible to produce a smartphone anywhere in the world because almost every smartphone has chips. At least one chip that's made in Taiwan. Roughly half of PC processors pass through Taiwan at some point. When it comes to the types of chips that are in telecoms infrastructures like cell phone towers, Almost every cell phone tower has at least one, often many chips that are made in Taiwan. Today, all of the most advanced chips for AI applications are made in Taiwan data centers. Many of the networking chips in data centers are largely in some cases exclusively made in Taiwan. And so when you're looking at cutting edge production, the fact that Taiwan has 90% of cutting edge processor chip manufacturing illustrates that for high tech, I mean smartphone, PC, data center, telecom, losing Taiwan would be absolutely devastating. But I think actually in dollar terms and in percent of GDP terms, the bigger impact would be not in losing cutting edge chips, but losing the mature technologies where Taiwan's got capabilities other people have as well. But Taiwan's got more capacity than anyone else. And if during the pandemic we could produce more chips globally, but still, because the shortages have hundreds of billions of dollars in reductions in car sales, just imagine how bad it would be if we lost all of the mature chip making capacity in Taiwan as well. New cars, a thousand ships. Let's ballpark that. A typical new car, 15% of those ships are made in Taiwan. Well, it's 150 chips. Where are you going to get those ships? If Taiwan's capacity goes offline, there's, there's no spare capacity waiting to produce existing chips. Spare capacity is too expensive to maintain, so no one maintains it. I think we'd struggle to produce many cars in the year after a crisis between China and Taiwan. And it's not just cars, it's dishwashers, it's microwaves, it's coffee makers, it's refrigerators, it's airplanes. Just don't even begin to think of the number of chips in a 737. And then ask where many of those chips are manufactured. When you go through components for GDP and ask what percentage of them require semiconductors, the answer is almost all of them. Then ask what share the semiconductors are made in Taiwan. The answer is a large share, even for relatively low tech applications. I think what we find is that the impact on global GDP of losing access to Taiwan shipping capacity would be like the Great Depression in terms of its impact on manufacturing output.
- [47:59]
Paul Chapman
And key to this, just to re emphasize the point is you could not stand up an equivalent fab capability. I don't know what the timeframe would be, but it certainly would take a long time, right?
- [48:09]
Chris Miller
It would take years, multiple years. Because you need not only to build the building, the buildings are the easy part. You need to make the tools inside of them. And it turns out that because these tools are so complex that actually most of the tool makers have already booked out their 2024 capacity. Here we are in August 2023 and they can't just ramp up production of the flattest mirrors humans have ever made on a whim. This is extraordinarily complex production. It's actually worse though. During the pandemic, two of the largest toolmakers, ASML and Applied Materials, both publicly reported that they were facing delays in the production of the tools that make chips because of the shortage of chips.
- [48:53]
Paul Chapman
Okay, so let's go back to Morris Chang. I mean, therefore, has he effectively provided Taiwan with this silicon shield that China just cannot risk losing, or has the inadvertent outcome of it no longer making chips for China make it irrelevant to China and somewhat weaken that silicon shield? Because I mean, I'm just trying to understand the. I just can't. I mean, it just seems to me that it would be just so devastating to everyone that any kind of invasion of Taiwan or disruption of, of TSMC and other manufacturers in Taiwan would just be the biggest Pyrrhic victory since Pyrrhus himself.
- [49:32]
Chris Miller
Well, I think that's, that's the optimistic view. And, and I hope that, that you're right about that. I think it depends on, on several factors. One, to what extent do we think Chinese policymakers are properly informed? And I think I'm relatively confident that they know that if a large scale war broke out, the economic consequences would be very, very bad. Maybe they don't appreciate the full scale of downside risk, but I think that risk is roughly appreciated. Second, do we think that they're maximizing GDP or something else? I think history is replete with lots of instances of leaders that don't maximize GDP and instead try to maximize territory or glory or national interest or their legacy defined by other metrics. And 2022 has a good example of, of authoritarian leader who did just that. And so I don't think we should over index on GDP as the primary metric that drives policymakers decision making. But I think the third facet is that it's not impossible to imagine scenarios in which China thinks it can get away with a temporary blockade or a temporary pressure campaign. And maybe it can even show that the US is not willing or hopes it can show the US is not willing to defend Taiwan. And then we find ourselves in a very dangerous situation of the US having either to show that no, in fact it is willing to risk World War III to defend Taiwan or it folds in which case China takes control of Taiwan, the US geopolitical position in Asia and really in the world collapses and we've still got this tremendous uncertainty over what happens to Taiwan's chip making capabilities. And, and I think the fact that the military balance has shifted so dramatically over the past decade in China's favor just creates this new uncertainty that never before existed. And here's, I think that the key fact is that until a decade ago, if there were a war in the Taiwan Straits, everyone knows who would have won in Beijing and Washington and Taipei, everyone everywhere. Today, if there were a war on the Taiwan Straits, nobody knows who would win. And that uncertainty, I think is the key driver of increased risk around the Taiwan Straits.
- [51:55]
Paul Chapman
You know, as China has, you know, over the last 20 years acquired a lot of know how and technology to re offset that technological edge around jamming and AI equipment and so forth. Missile technology that you've been talking about. Right? And they now have J25 stealth fighters and so forth, a lot of which that know how ultimately came from US technology being sold and licensed there back in the 2010s and the 2000s. Okay, well just, I mean it's an absolutely fascinating book. I can't put our crystal hat on. It's just scary talking about Taiwan anymore. But one of the big stories at the moment obviously is Nvidia and AI and so forth. Can you just, you know, I guess the, the book stops in, in 2022. You know, what has happened since you think that just, I guess really just continues the thesis that you outline in the book?
- [52:43]
Chris Miller
Well, I think the really, the really big surprise of 2023 has been the extraordinary boom in generative AI, which has created ChatGPT for us, which is, which is great. But the bigger implication is that it has driven up demand for GPUs produced by Nvidia, turned Nvidia into a trillion dollar company right now. But also it's in some ways proven correct. The thesis of the US government officials who were betting that chips would be absolutely critical to the future of AI. It's the shortage of GPUs that exists right now. So most of the chip industry is actually in a period of surplus because the slowing global economy means fewer people are buying smartphones and PCs. But for GPU, there's a severe shortage right now. And it's gotten so bad that there are now VC firms in Silicon Valley that are not handing out checks to companies they're investing in because cash is easy to get. If you're an AI company, they're handing out GPUs instead, and that's a sign as to just how severe the shortage of GPUs is and just how critical it is to have the chips that provide the most advanced computing capabilities, capabilities for training AI. That was a surprise this year. I wasn't expecting just the severity of shortage and demand for GPUs that we've been seeing.
- [54:04]
Paul Chapman
Yeah, $40,000 chips and you know, more and more, as you say, getting added to the the entity list. Well, it's been an absolutely fascinating discussion. Your book is Chip war, published in 2022 the Fight for the World's Most Critical Technology, published by Scribner and available on all good bookstores and online. And an absolutely fabulous read. And I can't recommend our listeners to garden and get a copy enough. So thank you very much Chris for your time.
- [54:29]
Chris Miller
Thanks so much for having me.
- [54:32]
Paul Chapman
Thank you for listening. If you enjoyed this episode and want to support the show, please give us a positive review on Apple Podcasts or Spotify. To find out more about HC Insider and HC Group, a search and advisory firm firm dedicated to the commodity markets, visit our website at ww.hcgroup.global. there you can find out more about our services and our offices around the world. There you can also find more content, from interviews to insight pieces to more podcasts focused on the commodity value chains. Thanks again for listening.