It’s been only three short years since pandemic-era shortages brought production lines to a standstill for many goods that rely on semiconductor chips. In 2021 alone, chip shortages led to an estimated production loss of 10 million to 12 million cars globally, equivalent to more than US$300 billion in lost revenue. Also hit hard were manufacturers of consumer electronics, computers, and industrial products ranging from precision lab equipment to heavy machinery.

This widespread impact delivered a sharp reminder that silicon chips are increasingly integral to the health of the global economy, making them important to every company—whether its goods rely on them or not. Yet the complex global semiconductor supply chain is in some ways more vulnerable to disruption than ever, particularly when it comes to the advanced chips that power artificial intelligence (AI). AI has become a vector of heightened geopolitical rivalry between the United States and China. As a result, semiconductor companies—and their customers—must contend not only with physical supply chain risks, such as earthquakes, typhoons, and factory fires, but also with export controls and a shifting policy environment.

To discuss the issues, and how executives can navigate through them,   strategy+business sat down with Gregory C. Allen, director of the Wadhwani AI Center at the Center for Strategic and International Studies (CSIS) in Washington, DC. Prior to joining CSIS, Allen was director of strategy and policy at the Joint Artificial Intelligence Center of the US Department of Defense. What follows is an edited version of the conversation.

S+B: Why do executives outside the technology sector need to pay attention to the dynamics of the semiconductor industry?

ALLEN: Semiconductors underpin almost every other technology. There are semiconductors not only in your computer and smartphone but also in your car and washing machine. We’ve had one supply chain disruption in recent memory, related to the covid-19 pandemic. The [US] Department of Commerce estimated that the shortage of chips arising from this disruption shaved a full percentage point off US GDP growth. This tells you a lot about how critical semiconductors are to the economy.

Almost every CEO, especially in manufacturing and industrial sectors, already has a sense of how oil or energy prices affect their business. I would argue that they need to develop that same sort of muscle memory for how [whatever’s] going on in the semiconductor industry might affect their business.

The other reason is that the future of the global technology industry is pegged to artificial intelligence, and AI software runs on AI hardware. One computer chip running in an AI data center might cost more than US$40,000. Some of the data centers under construction will have tens or hundreds of thousands of chips.

S+B: Why are policymakers paying attention in ways they weren’t ten or 15 years ago?

ALLEN: While the semiconductor industry is global, there are some concentrated regional hubs where certain parts of the value chain are disproportionately weighted. And in the case of the most advanced semiconductors—specifically, advanced logic semiconductors—more than 90% are currently manufactured in Taiwan, which is at the heart of so many geopolitical tensions.

The risk that policymakers are wrestling with is what might happen to the global economy if something happens to Taiwan. By some estimates, the economic consequences of Taiwan’s semiconductor industry being cut off from global trade are in the tens of trillions of dollars. It would be a global economic catastrophe.

S+B: You’ve written that we entered a new world on 7 October, 2022. What’s the significance of that date?

ALLEN: I think two dates in 2022 are going to echo through geopolitical history. The first, February 24, 2022, is when Russia invaded Ukraine. The second, October 7, 2022, is less familiar to most people. It’s the date on which the United States enacted a new set of export controls that dramatically restricted the sale of artificial intelligence semiconductors, as well as advanced chip-making technologies, to China.

This reversed more than 20 years of US trade and technology policy toward China. Secretary of State Anthony Blinken later said that the post–Cold War era was over, and technology was at the heart of the competition to define what comes next. On October 7, 2022, the United States effectively said: “We recognize that the future is AI, we’re extremely concerned about what China might do in terms of military and intelligence uses of AI if it leads in that future, and we don’t want US companies providing technological support to China’s AI development efforts.”

S+B: To play devil’s advocate, I’ll say that most manufacturers aren’t buying advanced logic chips—they are buying chips that cost a few dollars or less. Why do they need to worry about geopolitics at the leading edge?

ALLEN: Everybody is usually buying chips that are made in fabs [factories] that not long ago [were] making the most advanced chips. The used equipment from the most advanced factories is a critical enabler of the entire downstream industry. Taiwan is the absolute juggernaut in advanced chips, but they’re incredibly important in legacy chips as well. They build the latest and greatest fabs, and those make the newest chips. But when they’re no longer the latest and greatest, they make chips for aviation or appliances or automotive—industries that typically require less than the latest and greatest.

It’s also worth mentioning that industries that have historically not been especially interested in the most advanced logic chips are demanding more [today]. As cars become increasingly autonomous [and] have sophisticated infotainment systems, they’re demanding better and better chips. The economics of the leading-edge chip industry are increasingly interwoven into many, many other industries.

S+B: Against this background, what’s your advice to CEOs? What actions should they be taking?

ALLEN: If I could give one piece of advice to corporate executives, it would be to know your semiconductor supply chain—not only your direct supply chain but also your indirect supply chain, the companies you might not deal with face-to-face but on whose goods and services you are critically dependent.

An automaker knows who makes their infotainment system, but historically they might not have known who makes the chips that go inside the infotainment system. Now we’re in a world where automotive leaders recognize that they need to have this knowledge. They need to know who to call if there’s some kind of disruption. They need to be able to forecast whether the market is going to be able to meet their needs.

Once you know where your dependencies are, you can start thinking about how to mitigate the risks. I think the absolute minimum step is diversification of your supply chain. The first thing that you want to understand is, Do we have a second supplier? Geographically, do we have an alternative to Taiwan? This could mean diversifying indirect suppliers so that if a shipping lane is blocked or whatever, you have someone you can call in a different part of the world.

S+B: In terms of future regulation, what do companies need to be prepared for?

ALLEN: Companies that have world-spanning operations are likely to face conflicting geopolitical and national security priorities. To take just one example, there are foreign companies operating semiconductor manufacturing facilities in China. Now, the United States might say you cannot provide new manufacturing equipment to those facilities. But at the same time, China might say you cannot get your existing equipment out—the capital stock needs to remain in China. That’s an example of where the United States regulations and Chinese regulations are directly in opposition in terms of the goals that they’re trying to achieve strategically.

In the case of China, it’s very difficult because Chinese regulators do not necessarily spell out in law what is allowed and what is not allowed. The bureaucracy might decide on the fly that something considered legal in the past is now illegal, and you’re going to be punished for your activities in the past. It’s a very difficult regulatory landscape.

Some companies are exiting the Chinese market. Others are taking tighter security precautions, thinking about the core intellectual property and process knowledge that they don’t want to transfer to local Chinese workers. Some companies are flying in workforces to perform critical process steps and then flying them back to their home countries. They don’t want to train the local workforce on certain types of operation.

As for the US, the policy emphasis right now is around semiconductors related to AI. Chips for dishwashers or other applications are treated quite differently in the regulations. This may or may not continue to be the case. The previous US administration applied tariffs to all different types of semiconductors produced in China. Who can say what’s going to happen in two, five, or ten years? Policy priorities change.

S+B: As well as regulating trade, governments are investing heavily in expanding national semiconductor manufacturing capacity. How do you see this playing out?

ALLEN: Historically, the industry has been optimized for a just-in-time supply chain. Now there’s a movement to a just-in-case supply chain—a willingness on the part of government and industry to anticipate disruptive scenarios, to make investments, to take action in advance.

In the United States, we have the CHIPS Act, the largest piece of industrial policy that the US government has enacted in quite some time—around US$52 billion spread over five years, so US$10 billion per year. The European Union also has a [European] Chips Act. Japan is making major investments. The advanced democratic economies are generally working together on this—how can we make sure that our investments are synergistic and build a more globally resistant supply chain among allied nations?

The other part of this story is China’s investment in its domestic industry. Some estimates of Chinese semiconductor subsidies are as much as US$65 billion a year, even before the October 2022 US export control policy. That is an astonishing investment. China has a very ambitious goal, which is to have not only superior technology but also incredible market share. China’s stated official goal is that 70% of all chips used in the Chinese economy will be produced in China.

S+B: Does this imply that the US and other governments outside China will need to do more?

ALLEN: The scale of China’s investment is one relevant data point. Another is the size of the overall semiconductors market—around US$600 billion a year. A third is the degree to which the US invests in something like energy security. There are entire government bureaucracies that live and breathe energy security every single day. Compare this to what we are currently doing in semiconductors. I think there’s going to have to be a second act to the CHIPS Act.