Semiconductors are for the 21st century what oil was for the 20th — the material resource that fuels the entire modern economy. And, much like oil in the Seventies, our leaders are waking up to the fact that microchip manufacturing is a rivalrous geopolitical issue; in the mounting Sino-American cold war, both power blocs are competing for the military and economic advantages they confer.
Chris Miller, author of Chip War, spoke to Freddie Sayers about how semiconductors became such a vexed issue, the purpose of America’s protectionist CHIPS Act, and the chances of military confrontation over Taiwan, the world capital of microchip manufacturing, which he places at 20%. Below is an edited transcript.
Freddie Sayers: Why have microchips become a geopolitical issue?
Chris Miller: If you look at the production of advanced processor chips — the types of chips that process data that you find in a smartphone and PC, or a data centre — around 90% of them are made in Taiwan, which also produces over a third of the new computing power in the world each year. Taiwan’s importance is really extraordinary when it comes to all of the world’s technology. And it’s managed to acquire this position largely thanks to the efforts of a single company called Taiwan Semiconductor Manufacturing Company (TSMC), founded in 1987. Since that time it has grown inexorably: today, it’s both the world’s most advanced chip-maker and also the world’s largest manufacturer of silicon chips.
FS: And this pre-eminence can be traced back to TSMC’s founder, Morris Chang. Is he still involved?
CM: He’s officially retired, although he still plays a big role in the company’s decision making and in the company’s culture. He played a critical role in building the entire chip industry. He was in Texas when the first chip was invented in 1958. He played a major role in Silicon Valley’s decision to globalise the industry and build assembly facilities in East Asia, places like Hong Kong and Singapore, and especially Taiwan. More than anyone else, he can claim to have made possible the semiconductors on which the modern economy is critically dependent.
FS: Do we know anything about his politics — his contemporary relationships with China and the US?
CM: He’s done such a good job in business, in part by managing relations with multiple different governments. He was born in mainland China; his father was an official in the nationalist government. When the Communists took power in 1949, he and his family fled. He moved to the US and enrolled at at Harvard, then he spent 30 years working in the US chip industry — he held the US security clearance for working on specialised chips that went into defence systems. And then he got a job offer to start this company in Taiwan, backed by the Taiwanese government, which put up three quarters of the money needed to found TSMC. And since then, he’s lived in Taiwan. He’s still a US citizen. But he’s one of the most influential people in Taiwan because he founded and for a long time ran a business that contributes around 10% of Taiwan’s GDP. But he has generally stayed out of politics. Most of TSMC’s facilities are in Taiwan, a couple of small facilities in China, and a couple of small facilities in the United States. But it’s predominantly focused on production in Taiwan.
FS: How did this history lead to the conflict we see today?
CM: The problem is that TSMC has grown its capacity in Taiwan over the last two decades, just as America’s military advantages around Taiwan have been deteriorating, and China’s have been growing. And so it’s in some ways really the fault of the United States for failing to keep deterrence in the Taiwan Straits effective. Two decades ago, it was obvious, in the event of world war, who would win. Today, it is not at all obvious who would win. That uncertainty created by our decision to let our military advantages over China deteriorate — that’s why we’re concerned about production in Taiwan.
FS: Do you think chips specifically are making the possibility of conflict between China and the West more likely?
CM: I think when Chinese leaders assess the situation in Taiwan and think about their goals vis-á-vis Taiwan, they’re not primarily thinking about semiconductors. The reality is, the Chinese Communist Party has wanted to control Taiwan since before the first semiconductor was invented. When they talk about Taiwan, they talk about the rejuvenation of the Chinese nation, these broad ideological slogans rather than anything specific, technological or economic in nature. If we’re trying to understand policymaking in Beijing, leaders in China realise that Taiwan produces a lot of chips, but that’s not at the centre of their thinking. Now, some people say that Taiwan’s role in chip-making deters conflict, because China knows that if it were to attack Taiwan, it would destroy the chip-making facilities and the entire world, including China, would pay a massive economic cost. And it’s true, the cost would be tremendous. But there’s lots of examples of wars in history that have been waged by leaders who have decided the economic costs were worth it, or underestimated the economic costs — look at Putin, for example. I think anyone who is betting on “mutual-assured economic destruction” to maintain the peace doesn’t have much grounding in history on which to base that faith. And if we were to lose access to Taiwan’s chips through a war or a blockade, the impact on global manufacturing would be as dramatic as anything we’ve seen since the Great Depression.
FS: Biden’s CHIPS Act is a response to this threat — what does it do and is it making the situation more or less dangerous?
CM: The CHIPS Act is trying to make it less costly to build chip-making facilities in the US. The goal is to reduce reliance on Taiwan by building new capacity in the US and providing subsidies for companies to do that. So both US firms and also foreign firms that build in the US will be able to apply for subsidies to help defray the cost of production. And it’s already clear that there’s new investment coming in to the US that wouldn’t have happened otherwise, thanks to the subsidies. The challenge is that we’ve allocated $39 billion to subsidies, when one new chip-making facility can cost $20 billion. So $39 billion sounds like a lot of money, but it’s not that much. And our reliance on Taiwan is so substantial, that we’re going to need vast new construction not only in the US, but also in Japan and Europe and Singapore and elsewhere to really diversify the manufacturing base.
FS: The CHIPS Act also really tightens sanctions on China specifically.
CM: The CHIPS Act bans any company that receives US government funding from also investing in cutting edge facilities in China, and then, relatedly, the US government last year banned the transfer of cutting-edge chip-making tools and cutting-edge AI chips to China. And although Taiwan is the country that can manufacture chips most efficiently, the tools inside the facility in Taiwan and inside of every advanced shipping facility in the world are produced by US, Japanese and Dutch firms. The US wants to stop these tools from going to China, with the aim of stopping China’s chip industry from catching up. And this is going to cause dramatic problems for Chinese firms, because, for the past several decades, they’ve made real advances, but they’ve relied on imported tools. There are no Chinese firms that can produce comparable tools domestically. China’s ship industry faces a dilemma whereby they can produce not-cutting-edge chips, using tools they import, but they can’t produce cutting-edge chips because they can’t get the machinery that’s necessary from the US, from Japan or from the Netherlands.
FS: Is this a rare case of sanctions actually working?
CM: I think they’re certainly working in terms of having an impact on the Chinese ship industry. But that’s only the intermediate goal. The long-run goal of the export controls is to stop China from developing advanced AI systems that can be deployed for military and intelligence uses. And to train an AI system, you need access to the most advanced chips in a vast data centre. The US’s goal over the next decade is to make it harder for China to acquire the chips needed to train AI systems, so that China’s intelligence and military systems are less capable, and America is able to apply AI faster than China. That’s the goal. The success or failure of that will be measured over five or ten years; it can’t really be measured today.
FS: Do you think it’s the military application of this that makes it so central?
CM: The concern isn’t only the specific chips that are in missiles or in planes — more important than that is the chips that are in the data centres where AI systems are trained. So if you want to train a car to drive autonomously, or a drone to fly autonomously, you do it in a data centre. And these data centres are immensely computationally intensive. If you want to train a computer-vision algorithm to recognise a cat versus a dog, you need to show it millions and millions of pictures of cats and dogs before it learns. And you need ultra-advanced chips to do that processing efficiently. So for training drones for defence systems, advanced data centres are key, which means advanced chips are key and that’s what militaries are really focused on right now. If you want more autonomous, smarter military systems, you need to train them in advanced data centres. That’s the key use of chips in next-generation military systems.
FS: Given their value, might a black market for chips develop?
CM: It’s certainly going to happen — we see that with Russia today. It’s now illegal to transfer many different types of chips to Russia for military purposes, but Russia’s smuggling them in from Turkey and Kazakhstan, and from China. The types of chips the US is controlling for AI purposes are different, because they’re only used in a relatively small number of use cases. And the data centres that we’re worried about are vast buildings that can be seen from space. So it’s hard to say if we’re going to stop 100 types of a given chip from getting into Russia — I’m pretty confident the Russian security service services can smuggle most of those in. But if you’re talking about bringing in tens of thousands of chips to fill up an advanced data centre, that’s a much more tractable problem for intelligence agencies to keep an eye on.
FS: So it’s a race between the US and China as to who can onshore sophisticated chip production fastest?
CM: I think that’s how China sees it. I think the US sees it slightly differently, because the US is capable of producing advanced chips, not on its own, but in cooperation with Japan, with the Netherlands, with Korean, with Taiwan. It’s only by acquiring software designs, machine tools, and materials from all of those countries that you can make an advanced chip. The US strategy is not to onshore everything; it’s not to create a self-sufficient sphere. It’s to cut China out of the international supply chain, but keep all of its existing allies together. And the goal of that is that we collectively are going to produce chips and sell them to 80% of the world economy; China’s going to try to produce chips using less advanced machines and sell it domestically to 20% of the world’s economy. That puts China in a very bad competitive position: smaller market, worse machine tools, starting from a position of backwardness. It’s a race that China doesn’t look likely to win.
FS: In the next five years’ time, do you think there will be a major military standoff in the area around Taiwan, which would lead to major global hostility?
CM: I think the most likely outcome is that we have a period of tension, but we avoid a major military standoff. But if you asked me to put a probability on it, I’d give you a 20% probability over five years. And that’s a much higher probability than I put on it five years ago. I think — if you accept my probability of 20% — a 20% probability of both a very dangerous military situation and an economic crisis that would be equivalent to the Great Depression in terms of its shock to manufacturing: we ought to be willing to spend a lot of money, diplomatic attention and military resources to avert this type of crisis. My fear is that we’re not spending nearly enough given that the probability is far from zero. And the magnitude of the shock would be catastrophic.