On March 25, 2025, Chris Miller visited Taiwan at the invitation of DSET, marking his first public analysis in Taiwan of technological policy shifts under Donald Trump’s second term. The forum brought together industry leaders, government officials, and major international and Taiwanese media to explore the geopolitical and industrial transformations that a second Trump presidency could bring.

The event was moderated by DSET CEO Jeremy Chih-Cheng Chang and featured a distinguished panel of discussants: Liang-Rung Chen, Editor-at-Large at CommonWealth Magazine; Ting-Fang Cheng, Chief Tech Correspondent at Nikkei Asia; Hung-Ta Lin, Deputy Editor-in-Chief at Wealth Magazine; and Wen-Yee Lee, Taiwan Technology Correspondent at Reuters.

Opening Remarks

Chris: Well, let me begin by extending a great deal of thanks to Jeremy and all the organizers at DSET. As Jeremy mentioned, it’s now been two years of collaboration between us, and I think there’s no think tank in the world that, in just two years, has managed to have a real impact—not just in their own country on the issues they focus on, but also internationally, on issues of semiconductor supply chains, export controls, tariffs, research, security, and the like. I’ve learned a lot from their research, and so it’s a real honor to continue our collaboration tonight.

As Jeremy mentioned, this is a critical point for the United States and Taiwan when it comes to semiconductors, but also when it comes to the bilateral relationship in general—perhaps as critical a point as we have in the last half-century in US–Taiwan relations. And so what I did in my opening remarks is to both touch on semiconductor-specific aspects of that relationship, but also to place them in, I think, the necessary context of the overall bilateral relationship—the overall political and security architecture in Asia and around the world—which I think, to a substantial degree, is in a state of flux that we have not seen in decades.

When it comes to semiconductor-specific issues, there are a number of specific topics that are on the agenda right now in US–Taiwan bilateral relations. There’s the question of tariffs, which, of course, President Trump has embraced as a policy tool in a way that we haven’t seen a leader of a major economy embrace in several decades. He’s promised to impose tariffs on major trading partners of the United States, which are the world’s other largest economies—Europe and China chief among them. And I think this presents substantial shifts in the structure of the national supply chain in general, and therefore, shifts in the semiconductor industry.

Now, of course, it was the case that President Trump proposed substantial tariffs on China in his first term, but in the first two months of his second term, we’ve already seen the average tariff rate on China double. And it’s doubled at a time when he’s been imposing tariffs on other countries. And so we’ve barely even noticed the fact that tariffs on China are now twice as high as they were when he took office just two months ago. And I think that’s a good data point for the scale of shifts that we should expect to see over the subsequent three years and ten months that remain in President Trump’s term in office—and that has vast implications for many countries, but especially for a trade-dependent country like Taiwan.

And even if it’s the case that President Trump decides not to impose tariffs on Taiwan in particular—and I have no crystal ball that lets me predict what he will do or won’t do when it comes to tariffs—any major shift in the world trading regime is bound to have very important implications for a country like Taiwan that is highly trade-dependent and deeply integrated into US supply chains. I think tariffs are going to be one key issue of the bilateral relationship that we can discuss this evening.

I think the second key issue is going to be the ongoing technology competition between the United States and China, which has several different aspects to it — export controls on advanced semiconductors and some of the manufacturing equipment, regulations in terms of accessing cloud computing services by both Chinese and other countries, access to US cloud providers, access to R&D capabilities and know-how, which the Trump administration also targeted as a policy aim, trying to restrict the transfer of activities and knowledge to China.

All these are areas where, in theory, there’s a lot of alignment with Taiwan’s government, but I think in practice, there’s also a fair amount of scope for friction, as Taiwanese companies, as well as companies from other countries, interact with the controls that already exist and new controls which I think are likely to be put into place. And so, the escalation of technology competition is going to have an additional impact on technology supply chains, and therefore impact on the structure of the US–Taiwan economic relationship.

I think the third challenge is going to be when it comes to security, because, of course, it’s impossible to look at the US–Taiwan economic relationship solely through the lens of trade or technology. Since the founding of modern Taiwan and since the origins of modern US–Taiwan relations, there’s been a deep and complex security relationship between the US and Taiwan. I think I’m not going out on a limb to suggest that security relationship is at a point of tension and stress right now in a way that it really hasn’t been since perhaps the 1970s — a half-century ago — when we saw a major change in the nature of US–Taiwan relations.

We’ve seen President Trump, over the past two months, be very forthright in his desire that other countries that are partners or allies of the United States spend more on defense. And we’ve seen the Trump administration be, I think, aggressive in its pressure tactics to get other countries — especially in Europe thus far — to spend more on defense. But I don’t think we should expect the Trump administration to stop at Europe.

And Indeed, I think when the Trump administration looks at the change in European defense policy over the past few months, they believe their tactics are working. Germany has passed almost a trillion euros of additional spending in infrastructure and defense, just to give one data point. And with results like that, I think we should expect the Trump administration to continue using similar tactics to pressure other allies and partners to spend more on their defense too.

And so I think that has implications for Japan, implications for Korea, implications for Australia, but I think Taiwan also has substantial implications for. And it would be, I think, wrong to assume the discussion about the defense relationship can be separated— or will be separated — from the discussion about trade. When it comes to US–European relations, we’ve seen the Trump administration intermingle economic, political, and security discussions — they’re not separate tracks, they are all combined.

And I think we should assume a similar conversation starts out in earnest at some point between the Trump administration and Taiwan — intermingles security, intermingles trade, intermingles technology and broader economic concerns. I think that’s going to be a hard conversation to be had.

I think President Trump has illustrated that he is keenly focused on questions of power and leverage — the most insightful aspect of his conversation with Ukrainian President Zelensky from several weeks ago was when he told President Zelensky, “You don’t have the cards,” which I think was, to many people, shocking because they expected — they had come to expect — US–Ukraine relations to be discussed through concepts of shared values and shared interests. And President Trump was much more transactional — focused on power imbalances and keen to use what he perceived as power imbalance in his interests.

And I think we’ve seen President Trump pursue similar tactics with regard to other countries that are smaller than the United States. And so when I think about the future of US–Taiwan bilateral relations over the next couple of months and the next four years, I think we should expect some pretty hard elbow diplomacy from the United States.

I think it will be a real challenge for the US and Taiwan to manage both the domestic politics of the United States and, I think, whatever domestic political impact is in Taiwan when those conversations begin in earnest. And semiconductors will be part of that conversation because they’ll be inseparable from global security.

When I talk to my friends in the chip industry, most of them wish they could get in a time machine and go back to 2015 — which is a time when they didn’t think about politics, they didn’t have to get photographed next to presidents, they didn’t have to analyze their supply chains to see which countries they intersected with. But I’m afraid we’re not going back to that world anytime soon. If anything, the chip industry intersects even more with politics today than it did when I started researching chip work. For better and for worse, that’s the world we live in. 

And as I think we stand at the cusp of another major change in the international trading system — another major change that structures supply chains — I think the politicization of the chip industry is going to ratchet up one level higher, with implications for the US, for Taiwan, for China, and for Japan — all the countries that have a big role in the chip industry. And for the United States and Taiwan who share a whole lot of interests in both the chip industry and, more broadly, managing this new level of politicization — it’s going to be a real challenge, which is why I’m excited to be back in Taipei to learn about how the Taiwanese government and society, and industry, are thinking about these challenges. And I appreciate all the work that DSET is doing in engaging in these complicated discussions. Thank you.

Panellist Discussion_1st Round

Jeremy: Thank you, Chris, for the impressive remarks. You mentioned many points which are critical, and I think there are huge insights that remind not only of like-minded countries but also especially here, on the island nation — the tech community here. In fact, we released a report on President Trump’s inauguration day. The report is titled “Walking a Tightrope: Navigating Taiwan-U.S. Semiconductor Security Under Trump 2.0” We also covered some of the points you mentioned. While working on this report, we interviewed Chris many times. You gave us a lot of hints, information, and thoughts on how to navigate the future of this uncertain era.

We’re really looking forward to today’s discussion. Therefore, we’ve invited four intelligent leaders from the media—senior journalists who have covered tech issues and the industry for many years. They are representing Taiwanese people who are working and living in this community to have a discussion with you. 

First, we have Liang-Rong Chen, Editor-at-Large at Commonwealth Magazine. I’m also a big fan of his podcast. Chris, you recorded an episode with him a couple of weeks ago. So I’ll pass the microphone to Liang-Rong.

Liang-Rong: Thank you for the introduction. Chris, your lecture provided a great foundation for my questions. Since I’m asking on behalf of the general public in Taiwan, I’ll have to ask easier questions.

At this stage, when we look at America, the first question we ask is: What’s wrong with your president? It’s strange for Taiwanese because during Trump 1.0, we admired him. Without Trump 1.0, all the AI servers would not have been manufactured in Taiwan—maybe many would be in China.

I’ve spoken with people in the server supply chain recently. Most of them say their largest competitor is Chinese. Without Trump, they would’ve lost. Thanks to Trump, they now have many orders from Nvidia. Some even joked about building a temple for Trump. 

But now, things are totally different. As we discussed before on my podcast—at the beginning of Trump 2.0, I asked you: Trump said Taiwan stole the chip industry from the US. What should we do now that he’s president again? Should we give the chip industry back to him? You told me: “Don’t worry. That’s just something he says during the campaign. Everything will be normal, like Trump 1.0.” But you know, it’s different.

Can you give us some insight into what really happened? Why did everyone predict wrong? And if the trend continues, what will Taiwan’s semiconductor or tech industry be like in 2030? That’s my first question. Thank you.

Jeremy: Thank you for expressing the anxiety for Taiwanese people. Our next panelist is Ting-Fang Cheng, the technology correspondent from Nikkei Asia. I’ve watched many of your online dialogues with Chris, so Ting-Fang is another representative of Taiwan. 

Ting-Fang: Thank you. So after Liang-Rong asked about Trump 2.0. I think the entire supply chain is also very worried about the new chip war between the US and China, as it could affect broader supply chains around the world—especially in Asia, including Taiwan, Japan, and South Korea.

One thing I especially want to ask is: after China’s initial success in EVs, they now identify humanoid robots as their next major area of development. This has already attracted a lot of Chinese investment. Many companies—including Xiaomi, Xpeng, and others—have said they are developing humanoid robots.

We’ve also heard that China might have the upper hand because they produce many precision components for humanoid robots. So we want to ask: how do you foresee the next phase of the US–China tech rivalry unfolding, and what’s at stake?

Some people say humanoid robots do not require cutting-edge chips. How concerned should we be about the potential disruption given China’s growing capabilities in mature chips and DRAMs? How concerned should Taiwan or the US be? And whether we are welcoming this new mature-chip shock from China?

The second question that everyone hopes to ask Chris, is about China’s localization campaign. They hope to build the entire supply chain themselves. For example, Huawei wants to build from chip production to chip equipment. And we see a lot of progress that they are now saying they can build lithography machines or something. In your book, Chip War, you discussed the history of lithography. So we’d like to ask: do you think China could build its own lithography machines and other machines? Would that disrupt the global semiconductor industry? Should companies like Applied Materials and Tokyo Electron be worried about that? Thank you.

Jeremy: Thank you. Liang-Rong focuses on the Trump issue and Ting-Fang focuses on the China issue. We have 10 minutes for you, Chris, to respond.

Chris’ Response_1st Round

Chris: On understanding the Trump administration policy is in “changes” and “continuities”, I think it’s easy in the media environment we live in to focus solely on changes. So let me start by focusing on continuities first, and then I’ll talk about changes.

On “continuities”, TSMC announced its first investment in Arizona under the first Trump administration. That is a continuity: from the scale of the investment test, increase the number of fabs planned to be built has grown over time. But it was now six years ago when TSMC first announced its investment. And so that is the continuity now across three different US administrations, across two different leadership teams at TSMC, across a whole lot of change in the chip industry, across the invention of ChatGPT—a lot has changed—and that’s great. There’s a continuity. And I think if you look, and you watch the press conference between President Trump and C.C. Wei (TSMC CEO) several weeks ago, people were surprised by the $100 billion number, but I don’t think anyone was surprised to see the US president and the TSMC CEO because we got used to that type of event. So there’s a major comment on tariffs. 

When it comes to semiconductor tariffs, President Trump imposed tariffs on imported semiconductors from China in his first term. President Biden doubled the rate of tariffs on semiconductors from China. So there’s another continuity. Even higher under Trump’s second term. So when it comes to export controls, we haven’t yet seen any new export control moves from the Trump administration, but there was a whole lot of continuity, I would argue, from the first Trump administration, and the Biden administration.

And I think the next move on export controls will be to make them more restrictive, not less restrictive. There will be continuity on that issue as well. And so I think it is important for us, as we look at the contours of US policy, to not overly index on the drama of today’s headline—there will always be dramatic headlines—but also to take continuity seriously.

I think we need to focus on the continuities as well as the changes, because the continuities help us understand what’s actually driving policy. It’s easy to over-index on any given precedent. When you’ve got the continuity of policy across three different presidencies, with different cabinet secretaries, and different political parties in the White House and Congress, it shows you that there are broader forces in US business and politics that are driving some of these shifts. And so I think those forces persist even the presidents change.

I think, first off, many of us—myself included—in the time when Trump was not president and Biden was president—forgot the extent to which President Trump has a unique skill in dominating the media environment. In his first term, every headline was about Trump, Trump doing good things, Trump doing bad things, Trump doing great, Trump doing terrible. It was all about Trump. And it’s not only in the United States, every country I go to, all they want to talk about is Trump too. We forgot about that when Biden was president because Biden wasn’t very good at dominating the media environment. Now that Trump’s back, we’re surprised at just the extent of headlines about Trump. I think part of what we’re reacting to now is that every headline we see is about Trump, and they’re all talking about and focusing on Trump.

The second thing that’s changed is that President Trump does not have to run for reelection. And it’s often said to be the case in US politics that in their second terms, presidents focus on their legacies. They’re less constrained by day-to-day politics and are thinking more about what they want to accomplish before the term ends.

And I think President Trump has been pretty clear that tariffs and a restructuring of open industrial policy are his priority. So I think that does suggest an increased willingness to take economic risks on the tariffs front to watch the stock market decline, to have companies be frustrated with him. This is why we’ve seen tariffs on Canada, which I think very few people would’ve expected just a couple of months ago. That said. I think we’ve also seen the Trump administration already, just two months, be responsive in the economy and now back to tariffs. When all the company CEOs call and say your tariffs on Canada are going to drive up car prices, we’ve seen the Trump administration step back. It appears to be a case that the decline in the stock market has already been impacting decision-making in the White House. So exactly how that process works is impossible to bring certainty, but I would suggest that we need to focus on the factors that are shaping the White House’s decision-making is muchly focusing on the tweet of the day.

If you look back to the first Trump term, were you better off focusing on the tweet of the day, or were you better off trying to understand the structure of politics? The structure of politics is a much better guide to what matters than the tweet or the headline of the day.

That’s my main takeaway from the first two months of the Trump administration. This doesn’t mean we can predict exactly what will happen next week or the week after. But I think structure does matter—in addition to agency or any individual president.

I think since the first Trump administration, the strength of Taiwan’s chip industry has grown. TSMC’s market share is higher today than it was when Trump left office, is higher when Trump first took office. So sometimes when I come to Taipei, I hear people talk about the hollowing out of Taiwan’s chip industry. And I ask them: what chip industry are you looking at? Because the rest of the world looks at Taiwan and says Taiwan’s position is stronger than ever.

So I don’t see any reason to think that those trends will dramatically change, because the reason Taiwan’s chip industry is so central in the world’s supply chains is not because of any sort of political factors—it’s because of technology leadership that other countries and other companies can’t match. Unless that dynamic changes, it seems to me that Taiwan’s position when it comes to semiconductors will remain very strong, like it is today.

The open question is, when it comes to the rest of the AI supply chain—everything above the AI hardware—what role will Taiwan play there? I think there’s been opportunities, but there’s also no guarantee that Taiwanese firms will manage to monetize above the chip level. And I think we should expect that a lot of the biggest businesses that will be built out of the economic profits will accrue not only to the chip level but also to all the software built on top. That’s why I’m not sure to what extent Taiwan will be a player there or not.

When it comes to Ting-Fang’s questions, the president we’ve not thought enough about over the past two months is President Xi, who of course is a major player, perhaps the major player, in the trends that we’re seeing.

I think people often believe President Trump started the chip war. I can assure you that there was almost no one in Washington who could pronounce the word “semiconductor” to say nothing of extreme ultraviolet lithography in 2016 when President Trump was first selected.

The reason semiconductors emerged on the radar screen of US policymakers was because of concern about China’s efforts to build up its own indigenous capabilities. And I argue this in Chip War that there was a learning process—first, it was industry and government in Taiwan who were watching what was happening in China and registering concerns, then it was Japan that began to look at what was happening and getting concerns. Only after that,  was the case that policymakers in Washington came to focus on semiconductors.

I remember interviewing one member of the National Security staff in the first Trump administration. I asked, “When did you start focusing on photolithography as an important topic?” And he told me, “Oh, my Japanese counterpart told me I should look into it.”

So if you want to understand how the United States got involved in the chip war, it was by learning from partners who were following work closely than Washington was.

I think this speaks of one of the key drivers of the chip war: China’s effort to build its indigenous supply chain is stronger than ever. The amount of resources they are putting into, these efforts is as large as they’ve ever been. And the fact that China’s made steady progress in many different segments of the supply chain over the last decade. I mean their capabilities are, again, greater than they’ve ever been. 

And we see this impacting companies in Taiwan, in the United States, in Japan, in Europe—perhaps South Korea, which is facing China’s head-on in the memory market, for example.

I think President Xi, for a variety of reasons, believes that self-sufficiency shifts is a core economic and national security goal. As a result, China is willing to spend vast resources with the aim of building self-efficiency.

I don’t really believe that policy shifts and other countries are going to change dramatically. China policy is broadly speaking set in the direction try to succeed in building a whole lot of indigenous capacity when it comes to both manufacturing and design. And I think China has made some progress in beginning to replicate the tools that are needed in semiconductor manufacturing.

I think there’s still a long way to go. We’ve seen headlines about Chinese lithography tools. but that is only headlines. But I think we should expect a country that is as large as China can eventually indigenize more and more, which means any sort of technology controls will be temporary rather than permanent in their efficacy.

Ting-Fang: How long do you think they can do it? It’s very challenging.

Chris: Well, this is a key question. Nobody knows exactly how long. It’s almost impossible to predict. We know that there have been public reports of numerous ASML staff being hired by a variety of Chinese firms. We’ve seen media reports saying that Chinese researchers have a limited salary to work on photolithography. One assumes that the Chinese government has been very helpful in hacking into designs of various tools and components that are needed. 

But we also know that it took three decades of research to build the first EUV (Extreme Ultraviolet Lithography) tool that worked in a commercially viable manner. And we know that every generational type of tool gets better and better. And so in lithography, like in every other thing, in the supply chain, you’re not trying to catch up to a static technology, you’re trying to catch up to a technology that’s rapidly moving forward.

It’s already been six years since China’s been trying and unable to acquire ASML EUV tools. Six years is a long time to be trying to acquire a tool and not be able to access it. My best guess is that it will take a decade to indigenize photolithography tools, which—if that’s true—means China will not have its own EUV tool in four years. But now I think it might take even longer than a decade, because four years seems to be pretty soon, given that right now China doesn’t have its own capability to build DUV (Deep Ultraviolet Lithography) tools. I think you probably need a DUV tool before you can make an EUV tool. So it may take even longer than a decade to fully indigenize. Thank you.

Panellist Discussion_2nd Round

Jeremy: Yeah, thank you so much to Chris for responding to Liang-Rong and Ting-Fang’s questions. I think this is super critical for watching the trends in Washington, DC. Although there is a continuity you mentioned in the Trump administration, on the other hand, there is huge anxiety — we see the former threats in Trump 1.0 not be nominated again in Trump 2.0. Matt Pottinger, who visited us last month, Mike Pompeo — traditionally friends of Taiwan. But we see figures like Elon Musk or other leaders who are not traditional policymakers are having more influence.

On the other hand, as you mentioned, we are in the hub of manufacturing activities in this region. So people here have more information on tech issues, especially on competition. The stakeholders here are capable of addressing issues or raising alarms first when China is accessing or surpassing our capacity for technologies.

So, we are trying to do methodology work and publishing reports on this part because we want to bring the agenda to Washington, DC — to make sure everybody is on the same page. We are in the center of manufacturing activities, and not many people in Washington, DC are capable of accessing this critical information.

So, in the second round, we’re going to have Hung-Ta and Wen-Yee. Hung-Ta is Deputy Editor-in-Chief at Wealth Magazine and a veteran journalist in the semiconductor industry. Recently, Hung-Ta is co-authoring a report with DSET focusing on legacy chips. So I’d like to pass the microphone to Hung-Ta for your thoughts with Chris.

Hung-Ta: Hi Chris, I hope to discuss questions we are looking for answers for a long time about mature nodes.

When it comes to capacity, Taiwan holds the number one position in mature node capacity. If you want to build an AI server, a car, or even a toy, you need that. Without mature nodes, you cannot build an AI server. That’s important for Taiwan’s economic security because many companies’ products include mature nodes — including TSMC.

But we saw China investing a lot in mature node capacity, and they are trying every process to produce by themselves. Meanwhile, the United States is not investing much in mature nodes. By 2032, I see the United States will only have a 10% market share in mature nodes.

So my question is: If a chip shortage comes back, how can the United States achieve economic security? And what policies will the United States use to enhance mature node security?

Because Taiwan has significant capacity, how can Taiwan and the United States cooperate to enhance supply chain resilience?

That’s my question.

Jeremy: And the second round’s second panelist is Wen-Yee. Wen-Yee from Reuters has just taken the position of Reuters Taiwan Technology Correspondent. Wen-Yee, it’s your turn to ask the questions.

Wen-Yee: Thank you so much. Following up on Hung-Ta’s questions — he’s focused on mature node chips — but I want to say that actually TSMC’s advanced nodes still have a high market share. So, at the White House, President Trump called TSMC a monopoly. My question is: Do you think TSMC’s $100 billion investment is going to be enough to ease the pressure from being called a monopoly and keep it from being targeted by the US government? That’s my first question.

My second question is still focused on TSMC. Even with all the money TSMC is going to invest in the United States, we’re still seeing tariffs as an issue. It seems like Taiwan isn’t spared from Trump’s talk of tariffs. So I was wondering — do you think more investment from Taiwanese companies in the US could help Taiwan and make a difference? Or is there anything that you think the Taiwan government could have done better to negotiate? Thank you.

Chris’s Response_2nd Round

Jeremey: Okay, so Hung-Ta—legacy chips, and Wen-Yee—TSMC and other Taiwanese companies in the US. Okay, so those are two questions from our panelists. Please, Chris.

Chris: On the question of legacy chips, I think there’s been a recognition in the US, in Taiwan, in Japan, and to some degree in Europe—depending on which European country you’re talking about—that there is already a major impact on pricing from China’s expansion of legacy chip production, and that there are risks overdependence on Chinese components and supply chains.

However, the question of what to do about that is a real challenge. We’ve already seen a number of very well-documented Western companies decline to invest more in legacy chip production because they were afraid they wouldn’t make any money, given Chinese pricing. That’s exactly the opposite dynamic of what we’d like to see. We’d like to see more investment in friendly countries to work align us, but we are seeing the opposite.

So, what can be done? Well, a couple of policies have been discussed—none of which have really been acted upon at scale. We’ve seen the US impose tariffs on the import of chips from China. The problem is that the US doesn’t actually import that many chips directly from China. The complexity of supply chains means that if you wanted to use tariffs as a tool, you’d have to impose them at the component level—and that’s administratively very, very challenging.

There are a couple of precedents in US trade history where component-level tariffs were used, but those cases are pretty obscure and targeted products that were much less widespread than semiconductors. Most manufacturers say it would be costly—or maybe even impossible in the short run—to ascertain what are all the chips and raw materials and then to tell the customs agency, and to apply the proper tariffs. I think certainly companies have an incentive to play with the difficulty, but I think they’re right to say that the interface with the customs would be a real challenge.

In the US, we’ve had sort of a dry-run experiment with the Uyghur Forced Labor Prevention Act, which bans to use of any products tied to Xinjiang. Companies have found it extremely challenging to comply with that. For example, Volkswagen had a large shipment of cars held up at the border because they couldn’t certify that they didn’t have any Xinjiang components. I think that experience has given a lot of companies some of the tools they need to begin at the point of component tariffs. But it is the case that, even if we’ve never done this at scale before, applying component tariffs to a component that is in every product—often by the dozens, hundreds, or even thousands—would be a logistical challenge.

My view is that it’s still worth exploring. But if we’re going to do that, we should have a long time horizon before implementing any sort of component tariffs. If you imposed component tariffs tomorrow, it would be highly disruptive. If you imposed them starting in 2030, would it be disruptive? No, you’d have more time to adjust. So it’ll be interesting to watch how component tariffs evolve. It’s a powerful tool—but also a potentially disruptive one.

The second tool that’s been used in limited cases is simply banning Chinese shifts from certain sensitive supply chains—like connected cars in the US, for example. Soon, we’re going to see drone supply chains and perhaps in certain data center applications. But so far, it’s been limited to aspheres where data security is seen to be important—where there’s a risk of sabotage that has national security implicated. It’s not really a broad application. We’ve also seen no movement along this front in Europe and very little movement on it in Japan. 

And I think it’s worth recognizing that the US is only 20% of the world economy. So if the US acts alone, it risks not having the scale needed for global-level success. But if you have entire G7 acts together, you have more than half of the global GDP working together. You are scaled and you are sized. But if the US acts alone, the scale is on the side of everyone else who doesn’t act. And so assembling a coalition around this is tricky—especially given the difficulties of working with the Europeans. So this is the second tactic—banning Chinese components—that has limited effectiveness unless you can build that coalition. 

The third tactic is trying to subsidies. The US subsidized, Japan subsidized, Europe subsidized. But the unfortunate reality is that no one is going to win a subsidy race against China. So subsidies are a limited tool for short-run applications.

All of this means we’ve got a variety of tools—none of them perfect, and none of them have been tried at scale. That’s why China continues to win market share in legacy chips, and Western firms are investing less than ever. I think it’s a real problem. I think there’s a need for action, but I don’t yet see a coalition forming among enough countries to provide a real path to a credible response.

Hung-Ta: Can I have a follow-up?

Chris: Yes, please.

Hung-Ta: Because now you have the 301 investigations into China. So what will be the impact on semiconductor trade in the future? What are the possibilities out there?

Chris: Well, we don’t know what the results of the investigation will be. I’m sure the investigation will find subsidies in China. I’m not sure what remedies they’ll recommend. They could recommend even higher tariffs on chips coming from China—but that won’t matter, for the reasons we discussed. They could recommend component tariffs. They could recommend quotas. Once again, it doesn’t seem to me to be likely to have the effect we’re looking at.

I think the key question is: can you imagine a scenario in which the Trump administration uses tariff pressure on the Europeans, and then gets the Europeans to come back and cooperate? I think the track record shows that’s very hard to do—both because the Trump administration’s relations with Europe right now are bad, and the European Union is one of the hardest organizations to get to do anything.

So, I know that TSMC’s announcement—its new investments in Arizona—has been much debated in Taiwan. I think, from the perspective of someone watching the press conference between Xi Xiwei and President Trump in the United States, I would say that it came off as an extraordinary success. 

I think TSMC managed to put itself in a position of partnership with the new administration. President Trump was very supportive of TSMC in his remarks—he praised the company, its technology, and its centrality to US-AI supply chains. I think all of those are wins for TSMC. And I think they’re wins for the US–Taiwan relationship as well. So I do hope the announcement does help to stabilize both TSMC’s relationship with the US government and bilateral ties.

But I don’t think it obviates the need for action on all the other issues in the relationship that the Trump administration is going to push more. So I think the Trump administration has signaled that as part of the reciprocal tariff, the discussion that the president wants, there will be a discussion of tariff rates, of corporate tax differentials of non-tariff barriers to trade, and I think we should expect all of those issues to be brought up in the bilateral U.S.–Taiwan relationship as well. 

So I hope the TSMC announcement solves every issue, but I think it does put the relationship on a much stronger foot than it might have been. In that sense, I think it was a success.

You also asked a question about President Trump referring to TSMC as a monopoly, not once, but twice, if I recall correctly, in the press conference. I think this gets back to our discussion about the strength of Taiwan’s semiconductor industry. No tech CEO likes to hear their company described as a monopoly, but there are many tech CEOs who wish that their companies could be plausibly described as a monopoly, but haven’t achieved enough market share to reach that goal.

I think if you look historically, every technology company with such a level of market share has faced some degree of competition regulation or antitrust discussion, whether it’s AT&T in the US telecoms network, or IBM and major computers or Microsoft and the personal computer or Apple or Google. They’ve all faced some degree of investigation or regulation. And I think it’s no surprise that now that TSMC is in the ranks of the world’s great tech companies, it also faces these types of questions. 

What does it mean specifically? I think it’s hard to predict exactly how this discussion will play out, but I don’t think we should be surprised to see a tech company with a $800 billion capitalization and 90% market share having to answer questions about competition policy. And that’s inevitable when you’re so successful.

Jeremy: In terms of legacy chips, I just mentioned to Chris, we are going to release our newest report, which we call “The Great Siege: The PRC’s Comprehensive Strategy to Dominate Foundational Chips.” We are bringing this idea to Washington DC the next week as well.

On the other hand, so in terms of Wen-Yee’s discussion, TSMC going to the US, I think all of us believe that it’s only the phase one for the Taiwanese manufacturers going to the US. So my following question would be… How do you predict the end goal for the US? You previously mentioned that this is necessary for the US to gain the capacity of TSMC or other most advanced semiconductor manufacturing capacity, R&D, in the US as well. But what is your take on the end goal as a US strategist for the future of this dominating technology semiconductor manufacturing?

Chris: I think if you asked any world leader, “Would you like your country to dominate advanced semiconductor manufacturing?” they would say yes. So I don’t think we should be surprised to see US political leaders articulate simulatable having a bigger role in the advanced chips making or having cutting edge on R&D. But that’s not really a useful signal as to what the policy actually is.

So I think the key question is going to be: to what extent does the US government provide incentives or use leverage to attract or to push more investment in the United States?

And I think there are costs and benefits to the US, just as there are costs and benefits to companies. I think the benefits to local leaders are that, if you want more manufacturing investment — if you think this is within your political interest — and you get more manufacturing investment, then you can celebrate that by telling your voters. You can say you’ve delivered on your promises.

The cost of that, if it’s the case, that you’re attracting investment in ways that increase cost structures, and so downstream consumers pay more, well those are costs that will fulfill your economy over time.

I think for the United States there is a challenging set of trade-offs, in particular around advanced chip making and AI. I think the Trump administration has said it wants AI dominance as its strategy, partners in intelligence, and it wants as many AI models trained and deployed in the US as it can get. 

And there’s a pretty direct trade-off between your ability to capitalize on AI and the cost of AI hardware. If your AI hardware costs twice as much, you’re going to have a lot less. And so I think as the administration, in aggregate, thinks about how much chip-making it wants in the US versus other countries, part of the equation is: what’s the cost differential? And I think it’s pretty clear there will be a cost differential we can debate how large it will be.

So I think, next to the slogans of political leaders or next to statements like “we would obviously like more chip manufacturing,” the question is: at what cost to the budget, at what cost to your industries, at what cost to economic growth?

And here, I think there’s a recognition of real trade-offs. I think it’s notable that a month ago, it was possible to hear discussions about imposing tariffs on imports of advanced semiconductor products in the United States. But at least this month, that has died down—in part because, I think, of a realization that: who would pay the costs of tariffs on importing AI chips? The US AI industry would be the largest payer of tariffs.

And so I think those trade-offs are real. They don’t often translate directly into the political rhetoric under the current president—or, I think, under many presidents. That doesn’t ever exist. And so I think that’s a key part of understanding US policy.

Q&A

Emcee: Thank you all for your valuable comments. Let’s give them another round of applause.

Now we open the floor to questions to the audience, please raise your hand, state your name and affiliation, and keep questions short for Professor Miller, as the main respondent. Questions for the panelists are also welcome. Director Chang will select three questions for the first round. Thank you.

1st Round Q&A

Q1 | I-Chung Lai, Vision Foundation

Thank you for giving me a microphone. I’m from the Vision Foundation. So good to see you again, Chris.

I’d like to ask about what we should expect from technology in conversations with the administration. We see that during the Biden administration, they talked about technology from the standpoint of economic security — with science acts and “small, high fences” policies. But Trump seems to be in a different world — enacting industrial capacity-building for a fully American-made supply chain and job creation policies.

Add to that the possible officials managing Trump’s technology policy — like Jacob Helberg, soon to take over either the State Department or the Bureau of Industry and Security, and David Sacks in the National Technology Council — all have software backgrounds, working on AI, not manufacturing. They’re also “pay-per-game,” unlike the Biden administration, where some have manufacturing experience in tech policy.

With these different backgrounds, what do you think the conversation will be different?

Q2 | Guermantes Lailari, INDSR

Hi. I’m Guermantes Lailari, the Visiting Researcher at INDSR. I have a quick question. given the fact that China has been able to develop this technology over the last 25 years at a rapid rate. Can you discuss a boundary of two technologies, where people thought they couldn’t do it and they did, and another they wanted to do but couldn’t? Can you bound that in the context of semiconductors? Thank you.

Q3 | Jeremy Chiang, Executive Yuan

Hi, I am Jeremy Chiang, Executive Yuan. Building on Dr. Lai’s question, I was also really interested, because when you look at the US’s chip policy, it should not be viewed in isolation, but in relation to other Trump administration priorities. Also, we do see in Trump 2.0, that there seems to be a more centralized decision-making structure, centralized meaning maybe down to the president himself. So what we see is, at best, maybe Trump’s inattentive attention, or attentive inattention, towards chips. So he talks about chips, for example, as a monopoly. 

In contrast, with this first administration or Biden’s administration, there was a more technological focus on it. How is his dynamic with his team? For example, some describe his relationship with his team as more hierarchical. How does his relation with his team and also his other policy priorities affect his approach towards chips? Would it be more of a contractual approach? For example, we see in his first administration, that at one point ZTE was in the crosshairs, but the other second it wasn’t. Thank you.

Chris’ Response_1st Round

The question is about interpreting the Trump administration. Every US president will say their goal is to create jobs, whatever they’re doing. I think we can generally ignore political leaders when they’re talking about creating jobs because it means almost nothing. That’s not a statement of President Trump—President Biden would go to advanced manufacturing facilities that were entirely full of robots and talk about creating jobs too. So I’d ignore job creation as a meaningful statement. 

You’re right, though, that rebuilding industrial capacity has been a topic in Washington. I would know that manufacturing as a share of US GDP has not changed meaningfully in a decade, in which there’s been more and more discussion in the United States about rebuilding manufacturing capacity.

And I think if you look at this year through this actual action in rebuilding manufacturing capacity, it’s very narrowly focused on defense-related manufacturing. And there’s a reason for that. It would be very expensive to have a broader re-industrialization of the United States. Although that sounds good as a political phrase, I think once one begins to think through the economic implications of what would be necessary to meaningfully increase manufacturing with a share GDP, it’s not at all clear to me that the American body policy is willing to pay the costs of that effort.

So my guess would be that in four years’ time, we see manufacturing as a share of GDP would be roughly comparable to what it is today because the shift in it would require major economic shifts.

And so I would focus, in particular, with actual action taking on the reindustrialization of manufacturing, and there is, again, a narrow sphere, drones, shipbuilding, defense in the industrial relevant spheres.

On the question of China and catching up, I think there’s a long history of people saying China can copy, but it can’t innovate. I think that’s all very clearly wrong. However, there are spheres where China still imports the outside world.

I think it’s insightful to look at the goods that China imports, that is net imports of manufactured goods.  We find that there are very few products that China’s a net importer when it comes to manufacturing goods. China mostly imports commodities: oil, natural gas. Semiconductors are top of the list by far.

Semiconductor manufacturing equipment is the second largest category of manufactured goods that China imports, and there are small volumes of a lot of things: jet engines, certain types of vehicles, machine tools, and certain types of biotechnology equipment. What links together all of these manufactured goods that China still imports, so therefore China can’t produce massive quality. That means that there’s a high number of components and a very, very low margin of error than all of these. 

And so this is the type of manufacturing that’s the hardest to do. It’s therefore the highest margin. And it’s where China’s naturally focusing its efforts on both reasons, self-efficiency and for economics, I think that’s what unites all these different areas, which doesn’t mean that they’re going to be impossible to propagate, just means it takes much longer than other spheres of manufacturing, which have fewer components and larger tolerances.

2nd Round Q&A

Q3

You talked before about how surprised we all were to see Trump so openly transactional with Zelensky, telling him he didn’t have the cards. Do you think Taiwan has given too many cards too early to the US with TSMC’s $100 billion investment, and how does that impact the silicon shield in the European Union?

Q4 | Jikkie Verlare, The Netherlands office in Taipei

My name is Jikkie Verlare, the Economic Security Coordinator at the Dutch office in Taipei. We’ve talked about the near monopoly of TSMC. There’s another company in the semiconductor supply chain with a full monopoly on EUV machines—ASML. Do you think ASML’s monopoly could end up in the crosshairs of the Trump 2.0 administration?

Q5 | Thomas Schrøder, KTH Royal Institute of Technology

Hello, I’m Thomas Schrøder from KTH Royal Institute of Technology in Sweden. My question is simple: How strong is the semiconductor shield? To what extent does it protect Taiwan? I want to qualify that—So I’m from the Division of History of Technology. One thing you sometimes see is that very advanced technology very often relies on very basic technologies. And the other thing about this is that whenever you take any technology, even very simple technology, is always cooperation between many different actors, many different places in the world. There’s the famous TED talk which says something like nobody in the world knows how to make a computer mouse. 

So, if we’re talking about those basic things, one thing that seems to be a challenge for TSMC is also that it keeps growing, and can it keep growing in this way, If you’re thinking about the there might be problems with electricity? Taiwan has some challenges in the environment of electricity for TSMC. There are also challenges with manpower, and there are not many young people any longer. So yeah. By the way, your book is really awesome. Thank you.

Q6 | Yu-Jie Lee, National Security Council

You claimed the cloud war of the AI age began in January, earlier than the DeepSeek release, DeepSeek v3, and R1. What’s your observation from the chip war to the cloud war? 

Second, on China’s components, the Taiwan government has already banned DeepSeek, but we do worry about a lot of application services using DeepSeek as a foundation model. Not only that, there are some automobile manufacturers that are embedding this technology as an assistance. I think that’s a big risk in the future. What’s your comment?

Chris’ Response_2nd Round

I think in January, when DeepSeek first attracted international attention, it attracted attention for the wrong reason. The standard headline in the second week of January was about how the cost of inference was collapsing. We were surprised by that. Or the cost of training was collapsing, we were surprised by that. When you dig deeper, the inference price decline, the expected curves and the training price decline was less dramatic than DeepSeek advertised and also fit the expected efficiency improvement. 

So I think we sort of collectively got DeepSeek wrong in January and missede what perhaps is the broader and longer implication, which is that if China dominates the open source space, and companies have a lot of incentive to use open source models, if they’re good, China can embed its models broadly outside of China. 

And I do think that there are real risks here. I think there’s especially severe risks if you’re calling on the Deepseek AI, that’s obvious. And it’s striking to me that has not been restricted yet United States, but I think it will be, but has not yet been restricted. I think in Europe, it’s unlikely to be restricted, we might guess, but I think even reliance on locally running versions of the model, we should think very, very carefully about that, both because of the implications of running DeepSeek and the implications of not running a more trustworthy western model. Because market share matters fundamentally. And so yes, I think that’s the implication of DeepSeek. And I don’t think many governments have done enough think through what they ought to do about that question. 

ASML is different from TSMC—it has something everyone needs and no one replaces. That’s a good position. Countries may be unhappy about ASML’s 100% market share, but they need its tools. I don’t worry much about companies with irreplaceable technology.

When it comes to ASML, I think like TSMC, ASML has something that everyone needs, and no one can replicate. And so that’s a good business to be in if you have to negotiate. And so I would not be surprised to see countries unhappy about ASML 100% market share, but I also know that countries need their tools. So I don’t worry that much about companies that have irreplaceable technology, and 100% markets share, I think we’ll do fine.

Silicon Shield. Easy questions will last. I think I mentioned we haven’t spent enough time talking about President Xi, relatively what we talk about Presedien Trump. It seems to me that, if I were sitting in Beijing, I would look at the current moment with both a lot of concern, but also a sense of opportunism, that as the Western alliance center intents strain, there’s a possibility for China to take advantage. 

Now the risk for China is that Trump is difficult to predict, and that when he acts, he takes big moves, and so you would not want be on the wrong side of that, and so trying to escalate China’s activities in the Taiwan Straits, for example, could end up badly if Trump decided to respond to that in an aggressive action. 

But I think China is also looking at the current moment and thinking about opportunities as well, to split the US and Europe, to drive a wedge in the US and Taiwan. Looking at Korean politics, there are a lot of opportunities for China there to drive a wedge between the US and Japan. I think all this is happening to some degree of question is, how far will it go? And so I do worry a lot that although we haven’t heard much from President Xi over the last few months. That doesn’t guarantee we won’t hear from President Xi over the next couple of years. And I think we should not forget that Although President Trump is very good at dominating the headlines, President Xi has agencies too. 

I think that’s something for all of us to take seriously. And I think when I have conversations in Washington, it’s often forgotten that we’re not the only player on the chessboard and that our movements will have repercussions that we don’t always think fully through. That is something that I worry a lot about we don’t know what kind of next move is.

3rd Round Q&A

Q7 | Will Ripley, CNN

Will Ripley with CNN. If you’re a betting man, looking at China and the US, any chance of building an advanced chip manufacturing ecosystem that could rival Taiwan, in other words, being able to be right on on par with Taiwan in terms of getting these chips out of efficiently, sticking Taiwan decades to build that ecosystem. How many years would it take the US? How many years would it take China? And which country do you think could do it first? If it’s possible at all?

Q8 | Henry Lin, Ministry of Economics

Thank you, Professor Miller, for your remarks. I’m Henry Lin from the Ministry of Economics. My question is about the US economic industry. Currently, the US government wants to bring manufacturing back, but the software industry salary is three to four times higher than the semiconductor manufacturing sector in the US. Under this gap, how can you attract young people to manufacturing careers? Thank you.

Chris’ Response_3rd Round

Regarding the question of the the semiconductor manufacturing workforce in the US, I would say two things, I think, one, going back to our conversation about creating jobs, I don’t think the goal of policy should be to have more manufacturing jobs, per se, especially given the wage dynamics that you discussed. 

However, you need enough of a workforce to achieve whatever manufacturing goals you have, and so that’s where I think there’s going to focus on building workforce training programs and things like that so you can staff the fabs that are under construction.

I think what we’ve seen change in the last couple of years is that there’s a lot more excitement now in the US than there was in studying electrical engineering, studying material science. And I hear more and more on top universities about people double majoring in computer science and double E, for example, you’re at MIT University, or at other top universities. 

I think is a response not to the political demands or the workforce training programs, but a response to the fact that the world’s most valuable company is a semiconductor company. And so I think there you have economic incentives at work. Jenson has made chips cool again, and therefore you have people in the United States taking the semiconductor industry more seriously as a career path than you would have a couple of years ago. And that’s been, I think, a meaningful change over the past couple of years. 

But that’s, again, not primarily focused on manufacturing jobs as traditionally defined. And ultimately, our goal is to automate as much as possible, and that’s how we become more efficient. And so again, I would push back if anyone’s saying our goal should be to have more manufacturing jobs. 

US versus China. I mean, that’s impossible to answer any sort of certainty and number of years. You ask who has a broader base of the components that you need to have a complete semiconductor supply chain. The answer right now is in the US. A lot of tool making, for example, happens in the US, there is a larger workforce for invents like manufacturing in the US than there is in China right now. 

But if you ask who’s willing to sustain expensive policies over the long run to build an economically inefficient and unprofitable chip industry, the answer is China. I think that is the key question. For how long will China be willing to spend money building unprofitable companies? I think the track record suggests that China spend a lot of money over the long run and will be satisfied even if companies are not fully globally competitive and not even profitable, and that’s a policy challenge that I think all of the advanced economies right now are facing in different ways.

Related Research

Walking a Tightrope: Navigating Taiwan-U.S. Semiconductor Security Under Trump 2.0

The Great Siege: The PRC’s Comprehensive Strategy to Dominate Foundational Chips