According to economic historian Dr. Chris Miller, “You can’t understand the modern world without putting chips at the center of the story.” That conviction led Miller, an associate professor at the Fletcher School at Tufts University and nonresident senior fellow at the American Enterprise Institute, to research and write the 2022 nonfiction best seller Chip War: The Fight for the World’s Most Critical Technology. In an interview with McKinsey alumnus Ondrej Burkacky, Miller asserts most people underestimate the “absolute centrality of semiconductors to every aspect of their daily lives and to the big shifts in international economics.” The two explore the global trends affecting the chips industry, including localization and China’s capacity expansion; demand for chips for AI infrastructure, systems, and devices; and longer-term prospects for the industry. An edited version of their conversation follows.
Ondrej Burkacky: You’ve interviewed a lot of semiconductor industry leaders. What have you learned that has surprised you?
Chris Miller: First is the pervasiveness of semiconductors in all types of devices. Next is the extraordinary level of technology in every step of the production process. The chips supply chain—whether it’s the chemicals producers, the tool manufacturers, or the materials producers—has an exceptional level of precision in every step. A series of technological miracles make modern semiconductor manufacturing possible. And the companies that have mastered these processes and can manufacture at the nanometer level have vast market opportunities because it’s very difficult.
Prevalent global trends affecting the semiconductor industry
Ondrej Burkacky: To what extent are geopolitics affecting the semiconductor industry?
Chris Miller: Government officials in countries with a semiconductor industry (including markets in Europe, Japan, and the United States) are much more focused on chips than in the past because they want to ensure against supply chain risks and succeed in a fiercely competitive sector.
In the meantime, companies across the chip supply chain have had to reckon with more (and more complex) regulatory and global dynamics than they did five or ten years ago, and in an environment of rising competition. The industry is balancing the need to retain highly efficient international supply chains with risks of internationalization, including cross-border tariffs and regulations. These issues have risen on boardroom and executive agendas.
Ondrej Burkacky: We think local-to-local supply chains likely wouldn’t work, at least not today, without significant investments—and even then without inducing inefficiency from overcapacity and demand–supply mismatch. Where do you think this trend is headed?
Chris Miller: You’ll often hear rhetoric from government leaders about localization and even sometimes self-sufficiency. But at the policy level, you’ll find broad recognition that self-sufficiency is not viable except for in some industry niches. Instead, policies are intended to boost resilience. If something goes wrong, countries want the capacity to produce supply chain steps themselves. That will cost some money and create some inefficiency. China may be an exception; it is pouring more resources into self-sufficiency, although there is still no clear pathway to achieve it.
Ondrej Burkacky: China is the biggest global market for semiconductors. How should semiconductor companies be thinking about it?
Chris Miller: The Chinese government has invested more than any other country in semiconductors and has vastly expanded not only its leading-edge production capacity but also capacity in foundational, mainstream processes because they are relatively straightforward to build out.
A key question for companies operating in the Chinese market is “How far ahead of local competitors are we?” Companies that are technologically on par with domestic producers are already facing real pressure from Chinese customers to localize. By comparison, companies that are far ahead of their closest domestic rivals can be much more confident in their ability to sell into the Chinese market. But companies that are further ahead also face the risk of being seen by the United States and other advanced economies as providing China with “chokepoint technologies.”
Ondrej Burkacky: What are other governments doing to support the semiconductor industry in their countries?
Chris Miller: Every major country is investing public money in semiconductors, with some interesting regional differences. In the United States, for example, the focus is on leading-edge manufacturing, with big grants being given to firms such as TSMC, Samsung, and Intel. Japan and Germany, by comparison, are shoring up the production of semiconductors for manufacturing supply chains to support industries such as automotive.
Although India has not historically been a major player in semiconductors, the government is now supplying up to 75 percent of the capital needed by investors and companies to build front-end and back-end manufacturing. Saudi Arabia and the United Arab Emirates are showing a lot of interest in semiconductors. They have the capital but face steep learning curves in terms of expertise and building the ecosystem necessary for semiconductor manufacturing.
Concerns about overcapacity
Ondrej Burkacky: Given all the capital being poured into the semiconductor industry, the risk of overcapacity is spreading, which could lead to price pressures. Do you see that as a potential development?
Chris Miller: On the one hand, a lot of government funds are flowing into the industry, including $40 billion in direct subsidies in the United States, a comparable amount in Europe, and only slightly less in Japan. On the other hand, in the context of industry capital expenditures, these are not transformative sums. The entire US CHIPS Act is only slightly more than TSMC’s capital expenditures last year, for example. But, yes, overcapacity, which is the flip side of resilience, is a concern.
I worry less about overcapacity on the leading edge because companies need to build new leading-edge facilities every few years anyway. Overcapacity could be an issue with mature-node, foundational semiconductors processes, although demand is also rising every year as we put more of these chips in cars and consumer devices.
Ondrej Burkacky: Automotive is one of the fastest-growing semiconductor segments, especially demand for mature-node chips. What chips trends do you see in that sector?
Chris Miller: Europe, Japan, and the United States, among others, have most of the existing capacity to produce mature-node chips used in auto manufacturing, although most of the new capacity is coming online in China. This could lead to pricing pressure, which, in turn, could deter new investment by incumbent countries. As a result, manufacturers in the West could become more reliant on suppliers in China for these chips.
Ondrej Burkacky: Might governments impose local requirements on suppliers to mitigate the risk?
Chris Miller: Local content requirements need to be balanced against the recognition that supply chains between advanced economies are already very complex. Government leaders are looking to find ways to align their policies to address the trade concerns coming from the expansion of capacity in China without fragmenting into separate markets any more than is absolutely necessary. Rather than broad implementation of local content requirements, we’re likely to see steps specifically targeted to chips made in China. For example, the United States recently announced it would double the tariff rate on semiconductors directly imported from China.
The next wave of demand for chips
Ondrej Burkacky: How do you see AI trends affecting the semiconductor industry?
Chris Miller: There are two aspects of the AI market. First are the data centers that train AI systems, which to date is where we’ve seen most of the investment, excitement, and focus. Second are the edge devices, all of which will need semiconductors for AI processing. Autonomous vehicles are a great example. All the sensors needed for autonomy increase semiconductor demand.
Governments are placing big bets on AI systems getting bigger and needing more (and a greater number of high-powered) chips every year. This is what AI leaders call the scaling laws. The more you scale up your training data, the better it gets. If that’s true, then access to the leading chips will be very important. And that’s why you see the United States restricting the sale of these chips to China. It’s also why you see governments in the Middle East, Central Asia, and India trying to gain their own supply of these high-end chips and all of the data center infrastructure around them, so that they can train and deploy their own AI systems on infrastructure they control. This is often referred to as “sovereign AI,” and it has become a major trend for many governments.
Ondrej Burkacky: According to McKinsey estimates, 7 to 10 percent of global electricity demand will come from data centers in 2030. How will the massive investments in infrastructure, including electricity required by AI systems, affect global industry competition?
Chris Miller: The competition is already under way. For example, Middle Eastern governments are trying to attract investments by US cloud computing firms to build massive AI data centers in the region. It takes a lot of capital. It takes a lot of power. And it also takes agreement between the countries to make it possible.
The longer-term outlook for the semiconductors industry
Ondrej Burkacky: Chris, you have a newborn baby at home. Fast-forward to the baby’s 18th birthday. What are your pessimistic and optimistic long-term outlooks for the semiconductor industry?
Chris Miller: The gloomy scenario is that the rate of technological progress slows. For almost 75 years we’ve benefited from Moore’s law, which says the number of transistors in an integrated circuit doubles about every two years. But it’s getting harder to continue at that rate. We’re already manufacturing at nanometer scale. How much more can you shrink transistors at an acceptable cost? This is a real risk hanging over the industry. Progress can be achieved, for example, by designing and packaging chips differently, but we can’t take for granted the rate of technological improvement. And it’s not just the chips industry; the entire economy has depended on this extraordinary free lunch of computing power without thinking about it, and it will be very difficult to continue.
Under a positive scenario, if AI delivers even a fraction of the capabilities people are promising, it would be extraordinary for society and also for the semiconductor industry. It would mean not only more data centers—with high-end GPUs, advanced memory chips, and everything else needed to produce high-end AI processing capacity—but also AI in all sorts of devices that will require semiconductors. Twenty years ago, it was possible to build a car with almost no chips inside; today, it’s hard to find a new car that doesn’t have a thousand semiconductors. We’ll need AI processing on the edge of the network, communications capabilities, sensors, data transmission, and so forth. That’s good news not just for the companies producing AI processors and GPUs but also for anyone who’s producing chips.
“Our firm is designed to operate as one—a single global partnership united by a strong set of values. We are equally committed to both sides of our mission: attracting and developing a talented and diverse group of colleagues and helping our clients create meaningful and lasting change.
From the C-suite to the front line, we partner with clients to help them innovate more sustainably, achieve lasting gains in performance, and build workforces that will thrive for this generation and the next.”