One of the central arenas for competition in the 21st century is the battle over computing power—and the chips, data, and energy that go into it. In his first term as president, Donald Trump kick-started several restrictions on Chinese technology that Joe Biden has since expanded. What will Trump do in a second term? How will tariffs play into his strategy? And how will Beijing respond?
On this week’s FP Live, I spoke with the preeminent expert on semiconductor competition: Chris Miller, the author of Chip War: The Fight for the World’s Most Critical Technology. Subscribers can watch the full interview on the video box atop this page or follow the FP Live podcast. What follows here is a condensed and edited transcript.
Ravi Agrawal: Let’s start with a basic question: Why do we care where our chips are made?
Chris Miller: I think there are two reasons you might care where chips are made. The first has to do with supply chain security and whether you can get access to the chips that you need, because chips are in everything—including in cars, microwaves, and missile systems—and you need thousands and thousands of chips per person to make an economy function. Today, there is great concentration in the supply chain in Korea, but above all in Taiwan, which produces many of the most advanced chips, including 99 percent of the chips that are used for training AI systems. If you lose access to Korea or Taiwan, the entire global electronics supply chain will freeze up.
The second reason you might care where your chips are made is that data security really matters. It doesn’t take a frequent reader of spy novels to imagine ways that governments might try to use access to semiconductors as a means of gaining access to networks and devices for both espionage and sabotage purposes, as we’ve seen with Israel’s recent pager operation vis-à-vis Hezbollah.
RA: You mentioned Taiwan and South Korea there, but which other countries are influential in chip supply chains?
CM: Different countries specialize in different types of chips. When you’re looking at the processor chips that go into smartphones or PCs or AI systems, Taiwan is far and away the world’s leader. But processors alone aren’t enough. You also need memory chips that can store data. Here, Korea is the world’s leader for certain types of communications chips or sensors. There are companies in Japan or the United States or Europe that also have important roles to play.
China is the world’s biggest buyer of semiconductors. It spends as much money each year importing chips from all the other countries that I mentioned as it does importing oil. But it’s trying to catch up. So the biggest change in the chip industry over the last decade has been China’s very active industrial policy as it seeks to move up the technology ladder and become self-sufficient in an increasingly large share of semiconductors.
RA: Tell us more about where the United States fits in here.
CM: The United States is the leader in chip design with companies like Nvidia, Qualcomm, and Apple, and also big tech companies like Microsoft, Meta, and Alphabet. They are some of the most capable chip designers in the world. If you look at valuation, Nvidia is the largest chip company by far, but Nvidia has never manufactured a single semiconductor. Most of the firms that I listed work with the world’s biggest chip manufacturer, the Taiwan Semiconductor Manufacturing Company (TSMC), to actually produce all of the chips they design.
TSMC is the world’s largest chip maker and also the world’s most advanced chip maker, which means that there is a deep interdependence between Silicon Valley’s biggest tech companies and Taiwan’s biggest manufacturer. Neither one of them can operate without the other. That’s been a great partnership in terms of driving technological progress and enabling new types of products. But it also entails extraordinary risk if something were to happen to Taiwan.
RA: How far behind the United States is China when it comes to chip design and chip manufacturing, at least at the highest end of it?
CM: I think here you’ve got to differentiate between chip design, chip manufacturing, and then the production of the machines that are used to manufacture chips. This is an entirely separate level of the supply chain where the United States has a much bigger footprint than when it comes to just the manufacture of chips themselves. There are five big companies around the world that produce the most high-end tools that are used to make chips. Three are based in California. One is based in Japan. And perhaps the most important one, a company called ASML, based in the Netherlands, makes a type of tool called a photolithography machine, without which you can’t make the most advanced chips.
If you go to TSMC, you’ll find machines from each of these five companies inside of it. This is also true if you go inside of China’s most advanced chip-making facilities, because China is even further behind when it comes to making the machines that make chips than it is in making the chips themselves.
We had a great data point come out recently on this. Huawei, the leading Chinese telecommunications and AI hardware firm, just released its newest smartphone. The chip inside it was produced by China’s leading chip maker, a company called SMIC, using the manufacturing process that [Taiwan’s] TSMC pioneered around 2018. This means that the leading Chinese firm, SMIC, is around five or six years behind TSMC. That might not sound like a lot, but because of what’s known as Moore’s Law, the computing capability of chips doubles roughly every two years. So if you’re six years behind, you’re three doublings behind the cutting edge. That’s a position that China does not want to be in, and that’s a position that makes Taiwan so extraordinarily important.
RA: You’ve just described how China is behind. I’m curious what impact over the last few years U.S. export controls have had on that? How successful are U.S. restrictions on China?
CM: The United States has identified AI as a critical sphere in which it wants to maintain as large a lead as possible over China. It also noted that the key to training bigger AI systems is more capable semiconductors, in particular the chips that are designed by Nvidia and manufactured in Taiwan by TSMC. These are the chips that undergird almost all of the world’s leading AI models, whether trained by Google or Facebook or whether trained by Tencent or Baidu in China. Nvidia is by far the largest player. The United States has restricted both the high-end AI chips from being sold to China and the tools that are used to make high-end AI chips that China can’t manufacture on its own at scale.
However, enforcement is tough on its own, and the way the rules have been written allow, for example, the shipment of certain tools to certain facilities in China that promise not to produce cutting-edge chips, but ban the shipping of the same tools to other facilities that are producing cutting-edge chips. There are real questions to be asked: How much should we really trust the Chinese firms that say they are not producing AI chips when right next door there’s a factory that’s producing the cutting-edge versions that are prohibited? There’s probably a fair amount of subterfuge going on there, with Chinese firms lying about the types of use cases they’re actually undertaking. The rules have been full of loopholes, but even then they’ve had a real effect.
If you listen to the CEOs of Chinese firms, they’re very public about the comparison with U.S. AI firms—they face a much greater challenge accessing the chips that make it possible to train big AI systems. The fact that you see smuggling of chips into China illustrates just how dependent China still remains on chips that are imported from Taiwan and designed by U.S. firms.
RA: One of the things that the Trump administration will reportedly consider is tariffs on components of chips. Can you explain what that is and how that might work?
CM: One of the concerns that both the Biden administration and also many Republicans have had is the segments of the industry where China has got the capabilities that it needs (i.e., midrange and low-end chips). China is building out tremendous production capacity, which, over the next couple of years, is going to spill into global markets and drive down prices. Lower prices might seem like a good thing, but it’s bad news for Western firms that will face competition that’s not dictated by market dynamics but by the scale of government subsidies that Chinese firms are benefiting from.
We’re already seeing Western firms declining to invest as much as they otherwise would because they’re afraid future investments won’t be profitable as they look at this large surge of Chinese production coming online. To respond to that, the Biden administration has already increased tariffs on chips that are directly imported from China to United States. The challenge is that China doesn’t actually sell many chips directly to the United States. They’re often sold to Malaysia, packaged into a different component, and then assembled into a final device in Vietnam before they’re brought to the United States. So if they’re sent to the United States from a country other than China, or if they’re sent in a final good rather than at the chip level, they pay no tariffs related to China or related to chips.
So the U.S. tariffs that solely impact chips coming straight from China will have a very limited effect. That is why one of the debates in Washington right now is whether you should look inside of a good and ask how many Chinese chips it has in it. And if it has above a certain threshold, should it pay a tariff on its chip content inside? This would be a major change in how the global trading system works and a major administrative burden for companies that haven’t had to detail what components are inside of every washing machine or smartphone they sell. But it would also be a way to get at the challenge of how you impose trade measures in an era of very complex supply chains in which most trade is not bilateral—between two different countries—but rather crossing multiple different national borders.
RA: Chris, you know many of the players who are making these decisions, both at the company level and at the country level. What is your sense of how a Trump White House would game out what kinds of tariffs to prioritize?
CM: I think if you talk to companies, they’re quite concerned about blanket tariffs because they fear that this would slow economic growth in general, make products more expensive, and ultimately hurt their bottom line. When it comes to more targeted component tariffs, it depends on which type of company you are. Companies generally welcome any sort of measure that reduces their competition for China, but they oppose any measure that would increase their cost. So for every company that would welcome a tariff on a component coming from China, there’s a different company that would prefer to access the lowest-cost version possible, even if it’s from China. There are significant debates within industry about how to best target any sort of trade response to Chinese industrial subsidies in this sphere.
If you turn to the new administration, it’s clear that Trump has signaled that he believes in tariffs, and he’s focused on rebalancing the international trading system. He thinks that blanket tariffs are an important tool to do that. But I also think there will be voices in the Trump administration that will warn about its economic costs, the impact on inflation, and the impact on the stock market, and will be pushing for more targeted and impactful measures that can address the trade concerns in a way that will minimize the broader costs to the economy. That’s a debate that we saw playing out during the first Trump administration. I think we’re going to see that same dynamic playing out in the new administration as it forms.
RA: Another debate to watch is where to get the energy to build the data centers that are powering a lot of the development in AI. The places that have the money tend to be, for example, Gulf countries like Saudi Arabia and Qatar. Where do you think the future of AI is headed in a geopolitical sense? Is the next decade going to see the United States and China looking to other countries and places that can help them build energy sufficient data centers?
CM: One of the reasons we’ve heard the Trump administration, as it forms, talk so much about energy is that they don’t want to be reliant on other countries to provide the energy that AI will require in the United States. This is a point of comparison with the Biden administration, where they’ve had roughly similar views.
It’s been worth watching, for example, the requests from the United Arab Emirates to import large volumes of high-end chips to build a capable data center in that country. It’s been a real controversy for the U.S. government. On the one hand, the UAE is obviously a partner in many different spheres, but it’s also a country that has close ties with China.
We’re also seeing the United States responding and saying that if energy is the challenge, we can address that by permitting reform and by encouraging more energy production. This is going to be a challenge, but the United States will try because it doesn’t want a situation where it’s reliant on Middle Eastern countries to provide the energy required to build American AI systems.
RA: There are clear climate consequences to this discussion.
CM: It’s a very complex picture. If you look at the trade aspect, imposing tariffs on solar panel imports and increasing the price of solar panels is bad news for the solar buildout and therefore cuts against efforts to address climate change. But I also think that one of the factors that has made China so cost-competitive as a manufacturer is a differential in environmental rules between Western economies and China. In certain segments of industry, you see environmental arbitrage shifting manufacturing to China because it’s cheaper and there aren’t the same types of rules in place. In those segments of the economy, you could imagine tariffs having a positive environmental impact if they reduce the cost differential and make it equally cost-efficient to produce in jurisdictions that have better environmental rules. How that balances out on net is very difficult to say, but I don’t think it’s a fair assumption to say, in general, that restrictions or a rebalancing of the trading system would necessarily have negative effects on efforts to confront climate change.
The AI aspect is really quite difficult to deal with because there is going to be a great increase in demand for electricity that AI requires, at least in the short and medium term. There are intermittency issues—solar and wind power alone can’t do it. Nuclear power is a great idea for the medium or long run, but it’s going to take years to build out serious nuclear capacity. At the end of the day, it does seem to me like there will be some increase in use of natural gas to meet the power demands of AI. This will be a complex tradeoff for many different countries to deal with and certainly countries that have been trying hard to decarbonize their power system. They are going to face some difficult choices about how expensive they want AI to be.
RA: Let’s play Washington’s favorite parlor game right now, which is what to make of Trump’s picks so far in terms of what they mean for industrial policy, trade, and chips. There’s been a wide range, from China hawk Marco Rubio, nominated as secretary of state, to Wall Street financiers running commerce and Treasury.
CM: I think you’re right to suggest that the key positions in the cabinet really run the gamut from quite protectionist to quite free-trade oriented. This means that the president himself will have wide latitude to make decisions but also that if you want to understand the dynamics, you should look not just at the personalities but also at the interests at play. Look at Congress, which has been very stable in terms of its preferences over the past couple of administrations regarding what it wants vis-à-vis Taiwan and China. Look at business interests, which will lobby regardless of who the president is. Look at the interests of the U.S. economy—no president wants to preside over a slowing economy, inflation, or a sinking stock market. And so the president is going have to think about all of these different interests as he devises policy toward chips in particular, and tariffs in general.
RA: It’s very clear that the Biden administration was seen by investors as being tough on business. And Trump has promised to pull back on that. He’s obviously close to Elon Musk and to the tech optimist Marc Andreessen. But, on the other hand, you have Vice President-elect Vance, who publicly has supported the Free Trade Commission chair, Lina Khan, who was pro-regulation. What is your sense of whether the tech skeptics or the tech accelerators are going to win out in a Trump administration?
CM: I think they might each win in different spheres. When it comes to what are now the legacy big tech firms, companies like Facebook, there’s reason to think they’re going to face some challenges. We’ve heard people like Vance take a pretty tough line on Facebook in the past. We’ve seen some picks for key antitrust positions who have also been quite skeptical of some of the big tech companies.
But I think you would have a different view if you looked at the types of investments that AI will require. Here, the Trump team has been very vocal—not just at the level of Elon Musk but also the new energy secretary—about trying to find ways to streamline the construction process, streamline the permitting process, and facilitate major investments into energy and data centers in the United States.
It’s possible that you simultaneously get some victories for those parts of the administration that want to take on legacy big tech and some victories for those who want to facilitate the buildout of AI infrastructure, which benefits some of the legacy big tech firms in certain ways but also benefits new tech players like OpenAI, Anthropic, and xAI.
RA: How is China preparing its tech industry and its desire to build supremacy in chip production? How are they preparing for a second Trump term?
CM: There are a couple of strategies that China is pursuing. One is to continue trying to domesticate as much as possible the entire technology supply chain. They can’t do it yet in advanced chips. They certainly can’t do it in the tools that are used to make chips. But wherever they can, they are building up domestic producers and trying to become self-sufficient.
Strategy two is preparing to hit back against the Trump administration. We’ve already seen China threaten limitations on the export of critical minerals like gallium and germanium, where China has 90 percent market share in the processing of these minerals. These are minerals that are critical in chip manufacturing. So China’s threats to cut off U.S. firms are something to be taken seriously. In addition to that, there’s a second retaliatory method that China can use, which is to restrict U.S. firms’ access to the Chinese market. Because China is 20 percent of global GDP, it’s a key market for many big U.S. firms and certainly chip designers like Nvidia and Intel but also Apple and Tesla. Hitting back against big U.S. firms can be a politically powerful way to retaliate against any measures that Trump takes.
The post How Trump’s Return Will Impact the Chip Wars appeared first on Foreign Policy.
