Q: I am going to begin with what, given so much emphasis around it, sounds like a basic question now, but your book shows how complex it is. What is a semiconductor and how is it produced?

A: A semiconductor is a small device, in most cases the size of your fingernail. And they are used to manipulate electric currents to provide computing. And so today there are different types of Semiconductors. Some of them remember data and some of them process data. Some of them turn real-world signals into ones and zeros. But all of the computing that happens today happens thanks to semiconductors. And although most of us have never actually purchased a semiconductor because they are embedded deep in the devices we rely on, we touch dozens, hundreds, thousands of chips every single day as we go across our daily lives.

Chips are generally made of silicon and into the silicon are carved millions, often billions, of microscopic switches called transistors, which flip on and off, turning circuits on and off. And today, if you go to the store and buy a new smartphone, just the primary chip — and there are many semiconductors in a new phone — but the primary chip will have 10 or 20 billion transistors into them. Each one of these transistors is roughly the size of a coronavirus. And so the production of millions and billions of virus-size transistors requires the most complex machine tools that humans have ever made. Machines that are capable of moving materials at almost the atomic level produce the level of precision that temporary chip-making requires.

Deconstructing the supply chain

Q: Take us through the supply chain that goes into making a chip. And who are the big players in the supply chain?

A: So, to start, you need to design a semiconductor. And given that chips have millions or billions of components, you need to use very complex and specialised software. In the design process today, there are just a couple of firms that have a dominant position in the production of chip design software, almost exclusively based in the United States (US).

Once you have got your design software, you need to undertake a design, and chips are designed differently based on what they do. Some chips are in phones, others are in data centres. Some chips are designed to connect to cell towers, for example, to send signals back and forth from your phone. Others are designed to process images. So there are a whole wide variety of types of chips. And, generally, there are a small number of companies that have expertise in a given type of chip design. Most of those companies that are good at designing chips are based in the US, Taiwan, and Israel. India is another country with a large number of chip designers. But the US is the dominant player in chip design.

Once you have got your design done, which is essentially a computer file, you have got to begin the manufacturing process. And there are two different facets of the manufacturing process that are worth emphasising. The first is the production of the machine tools that are needed in manufacturing. These tools are among the most complex machines that humans have ever invented, tools that can deposit thin films and materials or etch tiny canyons in the silicon, just a couple of atoms wide. And there are just a couple of companies in the world capable of producing the most advanced types of these tools — one in the Netherlands, a handful in California, and a couple in Japan. You can’t make an advanced chip without acquiring tools from each of those three countries.

And then finally, once you have got your design, you have got your tools, you need to undertake the manufacturing process. And manufacturing requires, first, acquiring the materials that are needed. There is an ultra-pure silicon wafer that is required for each semiconductor, then a whole suite of purified chemicals and gases, which is the speciality of companies based in Japan. Then finally, once you have got the materials, got the software, got the design, you manufacture your chips, and this is something that a couple of companies, Samsung in South Korea, and above all, TSMC (Taiwan Semiconductor Manufacturing Company) of Taiwan, are the most skilled at doing.

The final step, once you have got the manufactured chip is to put it in a package and assemble it in a device, and this is a part of the supply chain that usually happens in Taiwan and China, or Southeast Asia.

So you can see there is not a single country that, on its own, can produce cutting-edge ships, and every semiconductor requires tools and expertise, the software from multiple different countries and dozens, if not hundreds, of companies.

Q: You started working on this in 2015-16, but the world began to hear a lot more about semiconductors in 2020 after the pandemic. The general understanding was that it showed the crisis of the supply chain. But you counterintuitively argue that the supply chain worked quite well, given the circumstances. Could you take us through the pandemic and what made semiconductors so salient?

A: Well, what most people don’t realise when talking about the chip shortage that we faced during the pandemic was that the world produced more chips each year of the pandemic than it had previously. More chips in 2020 than in 2019, more chips in 2021 and then in 2022. The issue was that demand grew even faster than supply.

And that was driven by a couple of factors. The most important was that the pandemic induced lots of people to work from home and induced lots of companies to upgrade their data centre infrastructure. And that coincided with the expectation of certain large users of semiconductors like auto firms that demand for new cars is going to slump because of the pandemic. And that didn’t happen. And what that meant was that as many customers of chip firms, such as auto firms, went back to their semiconductor suppliers halfway through the pandemic and placed orders for new chips, they found that that capacity had already been allotted to companies that were making PCs or smartphones or data centres. And that caused a vast disruption to a whole suite of manufacturing segments. That wasn’t really about a deficit of chips per se. It was this unexpected surge in demand, with an abnormal pattern as to what types of chips were being demanded.

The US-China battle

Q: Let me shift to geopolitics and how the US-China competition has shaped up in the chips domain. On October 7 2022, the US imposed severe export restrictions on China, and just last week, the US imposed restrictions on new investments in China. Why is Washington DC doing this?

A: When you talk to companies, they tend to think of semiconductors as being about smartphones or consumer devices or PCs. But when you talk to governments, they think about semiconductors first and foremost in the context of their defence and intelligence applications. And it’s been true for the past 70 years, and it remains true that the most advanced defence and intelligence applications are critically reliant on cutting-edge semiconductors for processing power, memory, the communications capabilities, for the advanced sensors. It’s as true as ever because all of the world’s major militaries and intelligence agencies are increasingly trying to apply artificial intelligence (AI) to their systems. And if you want the most advanced AI, you need cutting-edge semiconductors.

US policymakers came to realise over the past several years, first, that AI was becoming more important. Second, just a tiny number of companies produce the chips that make AI possible. And third, the entire world’s AI infrastructure is developed on US design ships. And so the US believes that, first off, currently, the status quo is that China’s AI capabilities, including for defence intelligence, are advancing on US hardware. And there is good evidence for that. And second, the US can develop an advantage over China if it’s able to cut off China’s access while retaining the most advanced capabilities for itself and its allies. So that’s what’s ultimately driving Washington to embrace this new restricted policy. It’s a desire to stop US technology from supporting China’s military modernisation.

Q: So how has this affected Chinese capabilities in the last few years, particularly in AI?

A: The honest answer is that the controls that have been put in place over the last 12 months will have a bigger impact over the next couple of years than they have had thus far. The reason is that most data centres in operation today are using chips that were designed several years ago because there’s a bit of a product cycle that needs to be worked through. And so the controls, I think, will have their biggest impact as we get towards the middle of this decade. Thus far, I don’t think it’s fair to say that they have had an immediate impact, but the impact is likely to grow over time.

Q: In the book, you point out that these US restrictions have galvanised the Chinese state into pumping more money into building its semiconductor ecosystem and seeking to attain a degree of technological independence, if at all that’s possible given the complexity of the supply chain that you just laid out. But what can China do and what will it not be able to do?

A: Well, there is a lot of uncertainty as to what progress China can make. I think that the challenge that China faces is that it’s starting from a position of being meaningfully behind in almost every segment of the chip industry supply chain, and quite far behind in a lot of key segments. And now the US and Japan and the Netherlands have made clear that they are not going to let China access not just cutting edge, but also one or two generations behind the cutting edge when it comes to tooling and software. And China is going to try to build these tools domestically, and over some time horizon, it will succeed, but it’s going to take a long time. So I think we should assume that for much of this decade, China is going to face a deficit of capabilities relative to what can be done by the supply chain that I described earlier, which involves the US, Japan, Taiwan and others.

And right now, that matters because the most advanced AI systems year after year require more and more data to train, which means that having access to the most advanced chips capable of training the most advanced AI systems will produce better AI now. That relationship might change in the future. But the trend in AI has been the more data you have got, the better system you can train, and to use lots and lots of data, you need advanced semiconductors.

The centrality of Taiwan

Q: What has made Taiwan so unique and why have others not been able to match the manufacturing capacity of TSMC?

A: TSMC has a unique market position. It’s not just the biggest, it’s half the foundry market, and many of the big fabulous firms, companies that design ships but don’t manufacture them, treat TSMC as either their only supplier or their most important supplier by far. And so I think the question is not who is going to match TSMC, because I don’t think TSMC is going to be matched in the short run. But I think there is a question of what other sources of supply come online. And for TSMC’s leading-edge production, the most advanced chips that are going into data centres or smartphone applications, it will be very, very difficult for anyone, but a small handful of companies to compete. But for more mature technologies, there is a lot more competition. And that’s where you see China, Japan, Europe, the US, and other countries, trying to develop their own capabilities in this more mature sphere.

Q: In the context of TSMC, you said that it’s not possible for others to up with what it has at the moment, but do you see other countries being able to integrate themselves into this ecosystem? And do you see a diversification of this supply chain from the few companies that have been controlling it so far?

A: I don’t expect a whole lot of diversification away from the critical companies involved. The reason is that when you’re talking about advanced fabrication or the manufacturer of these cutting-edge machine tools, the amount of really unique expertise that’s needed is extraordinary. And the places one can develop this expertise are only in the companies that are doing it. It’s not a science you can study in physics textbooks. It’s a unique tacit manufacturing know-how that only these companies have, which is why many of these companies have been in their market position for 20, 30, 40, or even 50 years. I think we should assume that there will be a whole lot of similarity in the way the market looks in 10 years or 20 years, as what does today.

Now the geography of where manufacturing is happening can shift a bit. I think Taiwan plays such a central role today that it can’t become more dominant in the market. So we already see TSMC beginning to open some plants in Japan and the US and Germany. And so I think we are going to see a bit less of a Taiwan-centric system, but I still think TSMC will play a fundamental role.

Q: I have two questions in the context of US-China tensions across the Taiwan Strait. One, do you see TSMC, as the Taiwanese government seems to see it, as a shield, or do you think it adds to Taiwan’s vulnerability? Two, you allude to three possible scenarios – a direct US-China conflict; China being able to take control of TSMC directly or indirectly, or TSMC operations getting knocked off. How are those scenarios going to play out for the world of chips?

A: I don’t think Taiwan’s chip industry is a shield. If you ask why has Taiwan retained its autonomy from China, the answer is not its chip industry. It’s the fact that the US military is protecting it. China has repeatedly, all the way back to 1949, tried to use military pressure, to take control of Taiwan. And the reason it hasn’t worked is because the US military has been there to push back against that. And that’s not about semiconductors. The US military has been defending Taiwan since before the first chips were invented. I think we must recognise if there is a war in the Taiwan Straits, the impact on the chip industry and thus on the global economy will be catastrophic. But that doesn’t mean that semiconductors are a likely cause of conflict, because the cause of conflict predates the industry. The cause of the conflict is that China thinks it has a right to control Taiwan, and Taiwan doesn’t want to be ruled by China.

The American ecosystem

Q: Last week saw the first anniversary of the CHIPS Act, and under the Joe Biden administration, there has been a concerted effort to bring back semiconductor manufacturing or elements of the semiconductor manufacturing ecosystem, where the US had lagged, back home. How would you evaluate this one year of the CHIPS Act and where the US is at the moment?

A: It’s hard to say definitively because the US government is still in the early stages of implementing the CHIPS Act. It is yet to distribute any funds to companies, for example. A lot depends on how this implementation happens. It’s already clear that there is a tremendous increase in investment in semiconductor manufacturing facilities that are happening, partly because private sector companies are doing some of their own volition, and partly because they anticipate receiving funds from the government going forward. But the impact on manufacturing investment is already clear.

The bigger question is can the CHIPS act revitalise the US chip industry in a way that leaves it in a better position even after government funds are no longer being spent? And that remains to be seen. You know, it’s not a surprise that when the government hands out money, companies respond by building more factories. But what’s harder to guarantee is when the government stops sending out money, whether they keep building factories.

Q: Isn’t the entire semiconductor space getting crowded? All governments want to build at least some element of the semiconductor manufacturing ecosystem at home. They are providing massive subsidies. And this is true even when you just look at the US and its allies and partners. Some of it appears complementary. I say this in the context of, say, India, where the US now has nudged a company like Micron to invest in an assembly, testing and packaging facility. So are there different elements of the ecosystem that the US and allies and partners can parcel among themselves, building on what they already have? Or do you see this as a competitive dynamic?

A: I think it’s mostly complementary, rather than competitive. Certain aspects are certainly competitive, but when you look at where most of the governments that we are discussing are putting their funds, it’s largely in different spheres.

So in the US, a lot of the funding is going to go towards leading-edge manufacturing facilities at TSMC, Samsung, Intel, and Micron on the memory front – not all, but a substantial proportion is my guess. In Europe, we have recently seen TSMC announce plans to build a facility there in close partnership with three European design firms. And so that’s again not competitive. That is building out the European ecosystem in a way that doesn’t compete with what the US is doing. In Japan, it is largely a similar story.

When it comes to the assembly, test and packaging part of the supply chain, there is some competition there between Southeast Asia, for example, Vietnam, Thailand, Indonesia, and India. But I think the key source competition is not between those countries. Key competition is capacity that’s currently in China and whether it will move out of China.

China is the biggest source of competition when it comes to assembly, test packaging. And I think that’s true at the fabrication stage. If you look at semiconductor subsidy programs around the world, China is spending as much as everyone else combined this year. There is a lot of discussion in Western media about the risks of competition between European and Japanese and US subsidies. And I think that is something that must be managed by the governments involved. But, you know, any objective look at the numbers suggests that the real competition is between China’s subsidy and everyone else’s.

Q: In that backdrop, do some crystal gazing for us. Five to 10 years from now, assume the contours of a bipolar tech ecosystem become clearer, a US-led one and a Chinese one where China is behind but catching up. How do you see this competitive dynamic playing out in different spheres, both consumer goods as well as military and intelligence applications?

A: Well, I think it’s going to be hard for China to compete in consumer goods that require cutting-edge semiconductors. And remember, that’s smartphones, that’s PCs, they all require cutting-edge chips. I think the reason it’s going to be hard is that today China is 20% of global GDP, the rest of the world is 80% global GDP. And so China’s task right now is to try to catch up while selling to a market 25% the size of industry leaders and doing so without the tools and materials and software. So I think we should expect that to present real difficulties for Chinese consumer device makers.

And I look at companies like Oppo or Vivo in the smartphone market, and I do wonder what their competitive edge is going to be as we have more bifurcation. Today, they are the world’s low-cost producers of smartphones. But if they at some point lose access to cutting-edge semiconductors, or if governments, like India’s government, want to squeeze them out and say, why don’t we just have an Indian firm assembling since anyway, the value add of Oppo or Vivo is quite small. It’s not like they are undertaking complex manufacturing that other companies can’t undertake. I think Chinese firms will face real challenges at this in the global consumer market. Certainly, they will have the Chinese digital market to turn back on, but the global market is where most of the dollars are.

In defence intelligence, it’s much trickier to judge because product cycles in defence and intelligence are much longer. And so, militaries will buy a new fighter jet hoping to have it in operation for 30 years, which means that new technology filters in some ways more slowly than you might expect in defence intelligence uses. That said, if you look at the Russia-Ukraine war, you see lots of evidence that cutting-edge technology is being deployed. When people think of cutting-edge technology in a military context as being like missiles, but the cutting-edge technology is in the satellite that’s taking photos of the battlefield, that’s applying a computer vision algorithm to identify what’s a tank and what’s a truck, and then delivering that in real-time through an electromagnetic spectrum where there’s constant jamming to a missile system. Firing and hitting the target is the easy part.

Where India fits in

Q: What do you make of what India is doing?

We have a semiconductor policy, and an Indian semiconductor mission housed within a key ministry. There is a production-linked incentive, which is this massive government subsidy and support for the sector. There was one big investment announcement by Foxconn and Vedanta to set up a fab but that has collapsed. But we now have a significant Micron investment. How would you judge how India is devising its objectives and its effort to meet them?

A: I would say a couple of things. I think first, if you look at historical examples of countries that started playing a really small role in the semiconductor ecosystem and went on to play much larger roles, Korea or Taiwan, for example, for both of those countries, it was a four decades-long process from their initial investments to having large, profitable high-tech players. And so I think that should put in context the journey that India’s electronics industry is on. Anyone who is expecting to get from India’s current position in the industry to where Taiwan or Korea are in a matter of years, that’s just not realistic. And so I think that’s the first thing to put in context. And I occasionally hear suggestions of disappointment from people in India that there’s not more rapid progress, but you know, I say this is the hardest industry that’s ever been invented. This is the most complex technology that’s ever been invented. So there’s no rapid progress. It takes hard work, and it takes long-term investment.

I think what India has going in its favour is a couple of things. One is that it’s not just the chip industry. It’s the entire electronics ecosystem, so the assembly of smartphones and PCs as well, that is looking very carefully at India as a location for manufacturing and assembly. And I think in the long run, that’s going to be a very strong tailwind to support not just electronics assembly, but also semiconductor packaging and fabrication.

Second, I think India has a very, very large base of workers with expertise in semiconductor design. Although there aren’t a large number of design firms that are headquartered in India, nearly every semiconductor design in the world has Indian staff members. And so the expertise in the workforce and chip design in India is second to none.

And then third, I think the Indian government has been quite generous with the various incentives schemes, both at the central level and also at the level of some states. Some risks must be managed because the incentives are generous. But it’s also the case that if you want to build a chip industry, it’s not something that happens cheaply.

Q: What elements of the ecosystem should India be focusing on? There is an obsession in some parts of the political circle to focus on the fab. But key experts seem to suggest that other elements need to be to built up first and policymaking has made space for that. Isn’t it a wiser course?

A: I think I would generally say that’s right. I don’t think it’s necessarily an either-or. I think there’s some space for both. But if you look at the way that Korea, Taiwan and Singapore entered the chip industry, they entered with assembly, testing and packaging before moving to fabrication. So, I think there’s a lot of scope for India to win investments in that sphere, particularly because it’s right adjacent to the device assembly, smartphone assembly, and PC assembly, where India is also in the early stages of winning a lot of market share.

That doesn’t mean that there’s no scope or fabrication. For compound semiconductors like the types of semiconductors that are in electric vehicles, there’s probably scope. For certain types of logic chips, there’s scope. I think a lot will depend on to what extent the government has local requirements and how they are structured. There are pros and cons of local production requirements, but in so far as the Indian government posts more of them, that will increase demand for locally made semiconductors.

Q: A final question. How has this journey been personally for you? When you started working on the subject, I’m sure people thought of this as some esoteric subject, and now you are among the most in-demand academics across governments and the business world.

A: You know, when I started this book, I told my mom, I was going to write a book on some semiconductors, and she said, Chris, that doesn’t sound very interesting. I am happy to say that I have convinced her now that they are extraordinarily interesting. But yes, I think this was a niche topic. It was something that people outside of the industry didn’t understand, and people in the industry didn’t take the time to explain its importance to outsiders. And I think the extent to which it’s become politicised, not just with China and the US but with really every major economy thinking seriously about semiconductors, pouring money into semiconductors, trying to reshape the supply chain for semiconductors, it’s put chips and the chip war front and centre on everyone’s agenda.