A few weeks ago, Biden hosted visiting Dutch Prime Minister Mark Rutte at the White House, where he urged the Netherlands to support a recent U.S. decision to tighten controls on semiconductor equipment exports to China. Among the restrictions, the restrictions include banning the export of ASML, the world's largest manufacturer of lithography machines, to China. ASML is currently the only company in the world that can manufacture the most advanced EUV lithography machines. Without the Dutch machine, Apple would not be able to produce iPhone chips. ASML doesn't just support the Dutch economy – it supports the economy of the entire world. And Biden's move, how did it happen, and what does it mean at the same time?
- 1ASML sells to customers all over the world, but only a few companies around the world can actually use EUV machines. But EUV lithography machines have few potential customers because the prices are so high and the level of precision manufacturing technology required to actually use them is really very high, so that ASML's customer base will always be only a few or at most a dozen companies.
- 2 The chip shortage has nothing to do with which country has the capacity to produce the most advanced chips. There is growing concern that China is making real progress in chip-making capabilities, and there is now a global government consensus that advanced process chips, especially those with high computing power that make their way into data centers, are critical to training the next generation of AI systems.
- 3 Right now, the United States is trying to cut off China's ability to make advanced semiconductors, judged to be critical to training AI systems. If you don't have access to state-of-the-art chips, then you can't make meaningful advances in AI.
- 4 Many links in the chip manufacturing process are capital-intensive, and manufacturing lithography machines is very expensive. This stifles global competition because newcomers have to spend billions of dollars to see if their products are working. And the expertise required to produce these types of tools is something you can't do on paper. You have to constantly grind in the manufacturing process. There aren't many training or doctoral programs that give you an idea of how these devices and systems work when actually manufactured.
Tufts University professor Chris Miller, author of "The Chip Wars: The Fight for the World's Most Critical Technology," has written a lot about this and delved into geopolitics and fascinating chip-making processes.
The Verge, the editor-in-chief of The Verge, and Chris Miller had an exchange, including foreign policy, high-power lasers, competent executives, monopolies, the fundamental limits of physics, etc., all in the following dialogue, Chris Miller gave his own interpretation.
Nilay Patel: A recent news story that has been in the spotlight is about a big thing happening in the tech world and the chip industry, and Biden is trying to pressure the Dutch government to stop exporting chip-making equipment to China. This is a very surprising geopolitical event. Can you explain what happened?
Chris Miller: Right now the U.S. is trying to cut off China's ability to make advanced semiconductors, and the logic of its judgment is that advanced semiconductors are critical to the training of AI systems. If you don't have access to state-of-the-art chips, then you can't make meaningful advances in AI.
To make advanced semiconductors, you need to buy lithography machines from the few companies in the world that have the ability to manufacture these precision chips. One of the most important of these companies is Netherlands-based ASML. ASML is the only company in the world so far capable of producing high-end lithography machine equipment that can produce EUV lithography machines, without which it would not be possible to manufacture advanced chips.
Nilay Patel: Most people don't realize that chip manufacturing is at the heart of a country's economy. From reading your work, (I learned) that the development process for EUV began with Intel in the United States, then completely broke away from Intel and eventually became the situation it is today. How does a Dutch company have this critical chip-making technology?
Chris Miller: The concept of lithography, the process of using light to create patterns on silicon wafers, was invented in the United States in the late 50s of the 20th century and deployed in the chip industry from the early days of the first generation of semiconductors. The chip industry originated in Texas and Silicon Valley in the United States, so the early users of lithography technology were mainly American companies. In the 1980s and 1990s, the chip manufacturing industry tried to move to a more advanced lithography technique called EUV (extreme ultraviolet) lithography, the so-called EUV, which refers to the type of light that is in this process of lithography.
ASML is an established manufacturer of lithography machines and has the ability to turn entirely new technologies into mass production. Much of the research is funded by Intel and several other U.S. chip companies, and is tested at U.S. national laboratories that have the same equipment and testing capabilities as the wavelength ultraviolet needed to actually manufacture them. Although science and technology are mostly done in California, no American company can commercialize such equipment. This also makes ASML the only equipment manufacturer in the world that can produce EUV lithography machines.
Nilay Patel: Can you tell us the basics of EUV lithography and how it makes chips?
Chris Miller: Before we talk about this, we need to understand what lithography is. If you want to make a pattern on a silicon wafer, you can do so by shining light with a mask. Masks block light in some areas and let light pass through others, which is how chipmakers etch microcircuits at silicon vendors. Today's advanced chips are embedded with millions or even billions of tiny circuits. They are usually the size of a virus, or even smaller. This requires ultra-precise processes.
EUV lithography uses light with a wavelength of 13.5 nanometers, which is an ultra-small light that is much smaller than the wavelength of visible light. This very small wavelength of light is needed because the engraved circuit itself is very small; Their own size is usually only a few nanometers. It is very difficult to produce this type of light because it is close to the X-ray spectrum, which makes the process of light generation complicated, and the mirror that reflects it is difficult to develop.
Nilay Patel: That's how the A13 chip (Apple) is made, right?
Chris Miller: That's right.
Nilay Patel: TSMC must buy an EUV lithography machine from ASML. ASML built the machine and then sold it to TSMC, which used it to make chips for iPhones or other things. Is ASML selling this machine to TSMC over? This sounds like a very complicated thing to do.
Chris Miller: It's very complicated. Transporting this machine alone would require multiple aircraft, each costing $150 million. ASML staff are on site next to the machine throughout the life cycle of these tools. ASML is the only company that knows how to solve problems when they occur, and they are the only ones that have spare parts to strike in case the lithography machine fails. Without ASML staff, you can't operate them.
They are so complex and precise that learning how to operate them in a mass production facility requires not only a lot of research on their use by semiconductor companies like TSMC, but also a deep partnership with ASML, which has a base of knowledge about how optics work and how light is reflected and refracted in different environments. You need to work very, very deeply with ASML to understand how to use these machines in actual mass manufacturing.
Nilay Patel: It sounds like ASML has a monopoly on this fab lithography equipment. Do they sell to other suppliers? Can Intel buy these machines? Can other foundries do? Can Samsung buy these machines?
Chris Miller: Yes. ASML sells to customers all over the world, and although there are some restrictions on selling in China now, there are actually only a few companies in the world that can actually use EUV machines. It is TSMC, Samsung, Intel, and some memory chip manufacturers such as SK Hynix and Micron. But EUV lithography machines have few potential customers because the prices are so high and the level of precision manufacturing technology required to actually use them is really very high, so that ASML's customer base will always be only a few or at most a dozen companies.
Nilay Patel: Why doesn't ASML make its own chips?
Chris Miller: Because ASML doesn't know how to make chips. They are a company of extraordinary value, but a company has a specialization in the art industry after all. The machine produced by ASML is just one of several ultra-complex machines involved in the complex and lengthy process of manufacturing chips. In addition to irradiating light at exactly the right wavelength through an EUV lithography machine, you need different machines to lay thin films of materials that are only a few atoms thick, or etch parts that are only a few atoms wide in silicon. These machines are produced by different companies, all of which have their own unique capabilities, but ASML knows nothing about the capabilities of devices other than lithography.
Chipmakers themselves have value in their existence. TSMC is better than anyone, including its suppliers, at using these machines to produce chips more efficiently. We need partnerships between all these different companies, tool makers like ASML, and chipmakers like TSMC to get the semiconductor done.
Nilay Patel: You can buy this device, but you have to know how to use it?
Chris Miller: Exactly. Understanding how to use it is a process that requires not only starting with a PhD in electrical engineering or materials science, but it does take years to use these tools. The process of developing the EUV tool took 30 years. This just gives you an idea of the precision required to actually utilize it.
Nilay Patel: I want to talk about it because that history with Intel is really interesting. In fact, we recently invited Intel CEO Pat Gelsinger on the show, and I asked him about EUV lithographs.
Previously, Intel was known for shorting EUV, but now their attitude towards EUV has changed 180 degrees. Pat Gelsinger acknowledged that some of Intel's previous decisions in the past were problematic, arguing that Intel was foolish to win the chip-making battle by not taking EUV seriously.
Returning to the topic, President Biden is pressuring the Dutch prime minister to restrict the export of ASML devices to China. You said that there are only a few companies in the world that can use these machines, and none of the companies you named are Chinese companies. Why worry that the Dutch government will allow ASML to be exported to China?
Chris Miller: If you can use advanced chipmaking tools, such as ASML and a handful of other advanced tool manufacturers in the world, you have a great prerequisite to make advanced chips. Of course, there is no guarantee that they can produce more advanced chips, because these tools are one of the key bottlenecks, and as I said, there are only a few companies in the world that can make chips with advanced processes, and none of them are Chinese mainland.
The United States is considering building the next generation of military and intelligence systems that increasingly rely on artificial intelligence. These AI systems are trained in huge data centers filled with complex chips such as GPUs with advanced processes, which are key to training AI systems.
If you can't make high-end chips at scale, then you can't get the data processing power needed to train AI systems. The United States eventually tried to prevent China from developing advanced data centers, which is one of the main reasons why the United States wants to restrict the export of advanced lithography machines from Dutch ASML to China.
There are many Chinese chip manufacturing companies in the past few years, trying to keep themselves up with technology and build high-end chips, which requires them to buy EUV lithography machine tools from ASML. China's leading chip foundry (SMIC) is the best example. But now the U.S. wants the Dutch to exercise control over a wider set of lithography tools—not just the most advanced, but also the "sub-advanced" previous-generation DUV lithography machines. This is a new request from the U.S. government that requires extensive negotiations and discussions between the United States and the Netherlands on whether to allow this.
Nilay Patel: For what, against what?
Chris Miller: The rationale for this proposal is that even the previous generation of lithography tools (DUVs) could be used to produce some fairly complex chips. The argument against the proposal is that ASML and other related equipment manufacturing companies would lose significant market share and revenue as a result, as Chinese customers have been investing heavily in chip-making capacity over the past decade and are heavily subsidized by the Chinese government.
For many chip tool makers, China has become a very important market for non-cutting-edge chips. It would be a costly decision if the Netherlands were to implement export controls that restrict not only the highest-end lithography machines, but also the previous generation of lithography machines. For leading chip manufacturers, the cost may reach billions of dollars.
Nilay Patel: You're describing an escalating sanctions regime against China, including chip manufacturing, technology transfer from U.S. companies to Chinese companies, and technology transfers from other international companies to Chinese companies. Is that all? Is that a strategy? Or is it that now the Biden administration is aware of the chip shortage because of the epidemic? Are all these initiatives coherent and the context of the resolution is passive?
Chris Miller: No, I think it's a coherent strategy that I'm going to separate what's happening now from Trump's tariff trade war. This is actually a separate area. I also think this matter should be viewed separately from the semiconductor shortage. The chip shortage has nothing to do with which country has the capacity to produce the most advanced chips. Over the last seven years or so, you've found within the national security agencies, the National Security Council, and intelligence agencies that there is growing concern that China is making real progress in chip-making capabilities, and there is now a global consensus that advanced process chips, especially those that make their way into data centers, are critical to training the next generation of AI systems.
From the late Obama administration to the present, there has been a considerable degree of consistency in policies that limit China's access to advanced chip technology, including the Trump administration. Not only the United States has done it, but also Japan and South Korea. In advanced semiconductors, many different countries have taken steps to implement new investment screening mechanisms or restrict the transfer of technology or knowledge to China.
Nilay Patel: Does China have the ability to catch up on its own? Or does it really need technology transfer, it needs to buy these devices?
Chris Miller: Well, that's a big question. If China cannot catch up on its own, the U.S. strategy will succeed. The U.S. is betting that China won't catch up, or at least not anytime soon. But there is some uncertainty around this. It's hard to predict whether China will find a way to produce some of the necessary technology for chips to involve and manufacture at home, or whether they will find a way to split the Western alliance and get some technology from countries that are unwilling to emulate U.S. export controls.
My guess is that the export restrictions that the U.S. and Japan are going to impose will be a real problem for China in the next few years, and that in the next 10 years or so, it will likely be extremely restrictive for China in terms of manufacturing advanced semiconductors. The more countries involved in these controls, the stronger the chain of export restrictions will be for the United States, which is why the Netherlands is so important.
Nilay Patel: How did we end up in a situation where we had to control a Dutch company to make sure China didn't acquire these capabilities? In the functional market, especially for something as important as chips, there should be multiple companies using many different methods to make advanced chips, but the reality is quite the opposite, there is only one ASML in the world, and it is in the Netherlands. How did this happen?
Chris Miller: Over the past few years, there has been a concentration across the chip industry, with only a few companies involved in many production and design processes, and in some extreme cases, only one company capable of producing software systems and hardware devices. There are two reasons for this. First, many links in the chip manufacturing process are capital-intensive, and manufacturing lithography machines is very expensive. This does stifle global competition, as newcomers have to spend billions of dollars to see if their products work.
The second reason is that the expertise required to produce these types of tools is something you can't do on paper. You have to constantly grind in the manufacturing process. There aren't many training or doctoral programs that give you an idea of how these devices and systems work when actually manufactured.
This means that the people who use these tools in a company have a truly unique value that makes it difficult for other companies and individuals to gain access to the relevant knowledge. It is for this reason that these equipment manufacturing companies provide a really strong moat, because for those who do not work in these companies, there is no direct way to obtain the necessary knowledge reserves. Capital-intensive and exclusive knowledge reserves make it difficult to build a few or even a dozen competitive companies in this industry.
Nilay Patel: A few months ago, I interviewed Pat Gelsinger about EUV, and now they're going to buy a machine from ASML, put it in Ohio, and try to make next-generation chips out of it. I asked him, "What's wrong with your EUV strategy?" "We've been gambling," he said. At the time we thought we didn't need EUV. We wanted to solve the problem of high-precision chip lithography through other technical routes, but these things did not succeed. Now it seems that at least we should have done a set of EUV parallel tests at that time. How stupid were we? ”
Do you think Intel was really stupid at that moment?
Chris Miller: EUV is a technology that should have been in production a decade ago. The development process was repeatedly delayed, with cost overruns of more than billions of dollars, and it was a near-complete failure for much of the late 2000s and early 2010s. It wasn't until 2015 that EUV was really developed. But at the time, it was very uncertain whether EUV was the "answer" to the next generation of chips, and whether it itself was far-reaching cost competitive.
Against this backdrop of uncertainty, some at Intel want to bet on EUV instead of betting on what they call unique technology. This means that using well-known lithography machines and conducting more tests of lithography work to engrave more sophisticated circuits will significantly reduce the overall cost input.
Trying to do so is in some ways a low-risk option. But in hindsight, Intel didn't do that. It's a bad bet, but you can understand why they made it. Engineers shouldn't be placed on the back, I think there's too much risk and uncertainty, and as Gelsinger said, Intel should prepare multiple avenues for R&D to see which one works. They didn't choose EUV, which seemed like a smart move to reduce costs at the time, but looking back at Intel's decision today, it actually cost them more later, because the technology Intel was advocating at the time turned out to be unable to produce the required accuracy for it.
Nilay Patel: Is that why Intel is lagging behind TSMC at every successive process node? Because their technology didn't work?
Chris Miller: That's part of the reason. I think this is a complex issue, but the latency of EUV is certainly an important part.
Nilay Patel: TSMC, which owns ASML lithography machines, is a formidable competitor in itself, although its business model is only to make chips; But all of its energy is focused on making chips. So now Intel must step up its efforts to become a chip foundry. If AMD wants to make chips there, he puts the AMD logo on the side of the Intel fab, but Intel has to go to ASML to buy an EUV lithography machine. And you must learn how to make ARM chips while developing Intel's next generation of truly own x86 chips. Is this possible? Does this seem like we're putting too much pressure on this company?
Chris Miller: Well, I think there's no question that his job is challenging. Of course, if someone can do it, he can do it too.
Pat Gelsinger is confident about this.
Chris Miller: Well, I think there's no doubt that Pat Gelsinger has turned Intel's culture around, but I think the challenge that you said about Intel's process technology manufacturing, design, and business model is going to be very difficult to take the company to a new track.
I think tackling all three challenges at the same time is Intel's only option at this stage, but it's a tough task ahead of him. I think for the United States, whether the competition in the chip manufacturing industry can achieve a phased victory depends largely on Intel's success.
Nilay Patel: I specifically asked Intel because it was the only option. No other large U.S. chipmaker compares. Intel is the only bet we have, and if they succeed, the national security dialogue, the supply chain dialogue, and the export control dialogue could be very different. But because Intel is in this very dramatic moment of transformation, it has some downstream implications for the way we deal with China.
Chris Miller: I think that's true. I think of the three leading companies that produce processor chips, and Intel is the company that naturally invests in the United States, because it is itself a local American company. However, I think it's also worth noting that Intel doesn't have an off-the-shelf foundry business in the U.S. In terms of building foundry capacity in the U.S., our starting point is very low.
But when it comes to building scale and foundries, everyone starts with a very basic starting point. In some ways, this is why the United States may end up betting not on just one, but on all three, among which is Intel, TSMC and Samsung. The U.S. government is trying to get them all to invest more in the U.S. and then see who can develop the largest facilities and the most practical business models in the U.S. and win the race.
Nilay Patel: It's a very American approach, right? Like: we will subsidize the creation of the market, and then a winner will emerge. Doesn't it feel like it ends with someone winning the game, and then another strange monopoly appears in the United States? Ever thought that what is actually needed is diversification at all levels of the supply chain?
Chris Miller: The difficulty and challenge is that diversification is very expensive, and if you want to pay for additional capacity that the chip industry is not going to use, you are going to spend a lot of money. A new cutting-edge chip manufacturing facility alone would cost $20 billion to $25 billion, and its cutting-edge nature lasted only a few years. I don't think there are really a lot of people in the U.S. who are willing to take on huge capital expenditures and you need to build excess capacity. We're going to get some marginal capacity through the CHIP Act, but we're not going to get a lot of excess capacity.
We need companies that can have a functional business model after we put their foundry business in the U.S. on hold. That's why I think it really makes sense to bet on multiple companies and see which ones can produce that effect.
I don't think we'll end up with a situation where one company wins and others fail. It is likely that we end up with multiple commercially viable foundries in the United States, which will be a good result. There is no need to explain why this particular market must end up with one dominant company and the others lagging far behind.
Nilay Patel: Did you consider investments beyond OEM players? Should we fund ASML's competitors or should we look for the next technology beyond EUV and have the government subsidize it so that we can diversify that layer? President Biden always stressed one thing: "Please don't sell this machine to China." ”
Chris Miller: When it comes to "should we have a competitor to ASML and advanced lithography?" "I think the answer is, the cost-effectiveness out there is not obvious. We may have the Netherlands implement controls that are quite similar to those in the United States. In the coming months, I think we will see the outcome of these conversations.
Supplies in the Netherlands are not at risk, and no one is worried that the Netherlands does not ship to American companies. The cost of setting up an alternative lithography company would be very expensive because ASML has the unique capabilities they have built over 30 years. It will be very difficult to replicate this or build a competitor. So I think our production and R&D funds are best spent elsewhere.
But when you're talking about next-generation lithography and next-generation tools, that's a great place to invest. If you look at how the Commerce Department plans to use the CHIP Act funds, you'll see that three-quarters of the money will go to incentivize more manufacturing, and another 25 percent will go to fund research and development. Some of this will be used for next-generation tools, including potential next-generation lithography systems, which are needed in 5 or 10 years.
Nilay Patel: What are those next-generation lithography systems?
Chris Miller: ASML itself is planning to develop two more generations of EUV tools. Now they have basic EUV.
Nilay Patel: The solder ball you described drops, gets hit by a laser, produces plasma hotter than the sun, and that's the basic EUV?
Chris Miller: Yes, that's basic. The next generation will be called High Numerical Aperture EUV, and it will have more precise optics that allow you to engrave more precise chips. These machines should be on the market within three years or so. They will cost twice as much as the underlying EUV tool.
In addition to this, ASML is also developing so-called supernumerical apertures, so more specific optics, it is unclear whether it will work. That's a decade away from production, but it's research and development that's already underway, and that's what we need if we're going to make smaller and smaller transistors on increasingly complex chips.
Nilay Patel: There's a chapter in the book about smaller and smaller transistors. The concept of Moore's Law is that the chip industry doubles the density of transistors on chips every year. We are already talking about having to shoot a laser through the flattest mirror in the world to a solder ball and then build superspecific optics to make it smaller. Are there restrictions? "Is there a limit to Moore's Law?" Anyone can say this at any time in the last 40 years and turn out to be a fool, but we are now talking about the atomic level. Is there a limitation in the chip industry that "we will not go beyond the level of individual atoms"?
Chris Miller: To some extent, the answer is yes, but we're not talking about individual atoms yet. We're talking about layers of material measured in individual atoms, but the transistor itself has a lot of atoms, even at the current microscopic scale. I think in the existing plans of companies like TSMC and Intel, we have a pretty clear vision – at least until around 2030 – of how they're going to continue to shrink transistors, stack them on top of each other, and use more tricks to get more atoms on the chip. After 2030, it will be even more difficult to say. It's always been hard to see too far, so I don't know how much that makes sense.
Nilay Patel: History tells us that people always find solutions to problems. I mean, that's why it's called Moore's Law. The reason I'm asking in this case is because we're talking about limiting China with advanced chipmaking equipment, and we're talking about next-generation GPUs or other AI-accelerated chips. Even if all these restrictions are made, will China inevitably catch up? Are we just buying time, or are we really creating lasting and lasting advantages?
Chris Miller: I think it's inevitable that China will catch up with the status quo at some point. I don't know if it's in 2027 or 2035. It will not appear in 2024, but years later. Will China's technology catch up? I'm not sure of the answer to this question either.
Nilay Patel: Even if Moore's Law fails, it will catch up, right?
Chris Miller: It depends on what we mean when we say "Moore's Law fails." At some point, shrinking transistors further is impossible, but that doesn't necessarily mean that the computing power you can get from a single chip necessarily stops. You can package them in different ways, you can bring memory closer to processing power, you can improve your interconnects, you can put photonics on chips. There are many different technologies that are still in their infancy in many cases and are creating new ways to get more computing power out of chips, all of which require creative design and truly precise machine tools to manufacture.
Even if you tell me that transistors won't shrink by a nanometer after 2030, I'd still say that we'll get more computing power from a square inch of silicon well throughout the 2030s and beyond. So if you take this broadest view of Moore's Law and name all the different things you can do to tune the chip to get more out of it, I think there's still a long way to go far beyond what we can do in 2040 to generate more computing power. For this reason, I very much doubt the argument that we will hit a wall.
Nilay Patel: I think the three companies that do the best job of pushing the computing power of a particular chip forward are Apple, Nvidia (to some extent AMD, so probably three and a half companies) and TSMC, which is the manufacturer of two and a half companies there. Apple is very good at packaging, very good at optimizing software for its own hardware, and very good at pushing the limits of ARM. NVIDIA is clearly a leader in the GPU space. AMD is not going to outperform the average Intel chip, but is doing better because they use TSMC's manufacturing power to improve the ratio of battery life to performance.
Nilay Patel: For these three-and-a-half companies, I thought, "Well, they're relying on TSMC. If TSMC can't push manufacturing, their technology to make and design better chips will actually come to naught, right? They are completely dependent on TSMC, which in turn is completely dependent on ASML. What does this relationship look like? Tim Cook wakes up in the morning worried about Dutch restrictions on ASML exports? Or is he a TSMC customer? Or is this just an API, he places an order, and the chip comes out?
Chris Miller: I think most of TSMC's customers are used to the fact that TSMC has an extraordinary track record of managing its own supply chain and ensuring that issues are resolved before they actually happen. One of the reasons customers love working with TSMC is that they don't have to wake up in the morning worrying about what will happen upstream of TSMC's production.
Because of these constraints, are people thinking about upstream supply chains more than ever? There is no doubt about this question. But if you're Apple or Nvidia, they don't really work for you, because all the inputs that TSMC production relies on are made in the United States, Europe or Japan. These countries are unlikely to control their transfer to Taiwan anytime soon, so you are actually very safe upstream. But your downstream – the chip assembly and the assembly of the final product that is usually done in China, this is risky.
Nilay Patel: How did we get to the point where the most important chip factories in the world are in Taiwan? How do we end the TSMC era?
Chris Miller: TSMC was born in 1987, thanks to the company's founder Zhang Zhongmou. In fact, he spent his career at Texas Instruments (TI), and before that he lived in Texas for most of his life. He was the visionary who wanted to create a foundry that didn't design any chips, just made them, which seemed like a crazy concept at the time because there were no fabless chip design companies. He started out with no customers, but he began to convince the company that he would do all the manufacturing work for them and take all the production risks; All they have to do is give him the chip design and he will return to the functional chip.
The model proved to be very successful because it allowed TSMC to scale up by serving many different customers. That scale, in turn, allowed TSMC to hone its production process, because the more chips they produced, the more they learned from actually making each chip. TSMC is both the world's largest chip maker and the world's most advanced chip maker, and there is a direct relationship between the two, both of which stem from the foundry model invented by Zhang Zhongmou.
Nilay Patel: Before we started recording, you told us that Zhang Zhongmou is your favorite character throughout the book. There is a sentence in the book: He can be said to be a Texas and not a Taiwanese. Why is he your favorite character?
Chris Miller: I think he's the most underrated businessman in the last 100 years. Most people have never heard of him, although we rely on the products made by his company every day. I think his life is a fascinating microcosm of the entire chip industry. Born in Chinese mainland, he moved to the United States after the revolution, attended Harvard University, was the only Chinese-American student in his class, and then built the chip industry with his own hands on the Texas Instruments production line, and then founded TSMC.
For all the major shifts in the chip industry and computing technology over the past 75 years, he not only illustrates those shifts, but he actually achieves them. We are all very grateful to Zhang Zhongmou, and I hope more people have heard of him, because I think his importance is really underestimated.
Nilay Patel: I think TSMC is grossly undervalued, it's a very opaque company. They are very proud of themselves, and they are very opaque and it is difficult to know how they work. How do you feel about TSMC? I mean, Zhang Zhongmou is no longer in place. How does TSMC culture continue? What does its new leader look like?
Chris Miller: Yes, Zhang is officially retired, but he is often present in TSMC's offices and at TSMC events, so I'm not sure we should really say that he is no longer there. I think the culture he's created is willing to bet big on R&D decisions, capital expenditure decisions, and TSMC's relentless efforts to hone its manufacturing processes.
Zhang's colleagues in the 1950s would talk about the ferocity with which he found inefficiencies in the manufacturing process and then pushed them out of the assembly line as quickly as possible. The pursuit of manufacturing excellence has led TSMC to what it is today, in large part due to Zhang Zhongmou and the culture he instilled.
*Disclaimer: This article is the original author's creation. The content of the article is his personal opinion, we reprint only to share and discuss, does not mean that we agree or agree, if you have any objections, please contact the background.
![The $39 billion "chip manufacturing subsidy" in the United States is not easy to take![fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)