Supply-Chain Chokepoints Turn Cheap Inputs Into Geopolitical Leverage

China’s most effective leverage in the 2025 U.S.-China trade confrontation, as Chad Bown described it, did not come from matching U.S. tariffs. It came from control over production inputs that manufacturers needed but could not quickly replace.

President Trump imposed tariffs first on China and then more broadly, with China-facing tariffs escalating above 100%. A simple tariff-retaliation model suggested China had limited room to respond because it imported relatively little from the United States. Bown said that view missed the real channel of pressure: chokepoints in specific upstream goods.

The first was rare earths and permanent magnets. Permanent magnets are not consumer-facing products; they are small industrial inputs embedded in ordinary manufacturing. Automakers need them to move seats, operate windshield wipers, raise and lower windows, and build cars that consumers will buy. China, Bown said, supplied about 90% of the world’s rare earths and about 90% of the world’s permanent magnets. When China stopped selling them to the United States in April and May 2025, the U.S. auto industry was “on the verge of having to shut down.”

The second case came later in 2025, after the U.S. administration expanded export controls. China cut off another input needed by automakers: a type of semiconductor. Bown emphasized that this was not a cutting-edge chip and that many semiconductor firms could make it. The vulnerability came from sourcing concentration. Auto-parts suppliers, including the German company Bosch, appeared to be sourcing from the same company, Nexperia. When the Chinese government cut off Nexperia’s chips to the world, automakers again risked being unable to make cars.

For Bown, these episodes help explain what brought the Trump administration back to the negotiating table after its tariff escalation. The pressure did not come mainly from Chinese tariffs on U.S. exports. It came from Chinese chokepoints in inputs that firms needed to keep production running.

Steven Davis treated that as the central policy problem: cross-border supply chains generate efficiency, but also leave points of coercive leverage when a geopolitical rival controls inputs that are hard to replace. The vulnerability is not merely that a country buys something from abroad. It is that a foreign supplier, or a foreign government with power over that supplier, can interrupt access to an input whose absence disables production elsewhere.

The analytical frame that followed was narrower than “globalization creates risk” and more useful than “make everything at home.” A chokepoint is not just import dependence. It is the interaction of concentration, substitutability, private incentives, and state coercion: who controls the input, how quickly users can switch, whether firms have reason to pay for resilience before a crisis, and whether a government is willing to exploit the dependency.

Soumaya Keynes located the modern wave of vulnerability mapping in the pandemic. The shock pushed governments to ask where they had become dependent on suppliers that might not be reliable. Australia, Britain, the European Commission, the OECD, and the United States all undertook exercises of this kind. Britain’s effort, Project Defend, was quite secretive, with little public disclosure of what it found.

Keynes warned that supply-chain shocks tend to vindicate, at least rhetorically, anyone already inclined to close borders and produce everything at home. Her account of the post-pandemic exercises was partly a defense of analytical discipline against that instinct. Politicians “freak out,” she said, and commission studies; economists then try to make the exercise rigorous, numbers-based, and resistant to overreading the data.

The simplest version of vulnerability mapping asks where a country imports a large share of some product from one country. Keynes said that can be a hint, but it is not the right question. The better question is where an economy could not recover quickly if it lost access to a particular input.

That shifts the analysis from import shares to adjustment capacity. It requires asking where substitutes can be found, whether domestic production exists, how quickly production could scale, and whether firms can reconfigure supply chains. Davis added another layer: demand-side substitution. If a critical input becomes scarce, can the economy stop using it in less important applications and redirect it toward uses that are essential for defense, national security, or core economic functioning?

Keynes illustrated the demand-side point with a historical example from the First World War. Britain looked more vulnerable than Germany because it imported much of its food and used more energy. Germany produced more food domestically and used less energy. But Britain’s higher energy use included discretionary or leisure uses, such as driving around, that could be cut back in an emergency. Germany, by contrast, was using more of what it really needed. The country that appeared less exposed could in some respects be more vulnerable because it had less demand it could shed.

The distinction changes what counts as a critical good. A country may import a product heavily from one source and still be resilient if substitutes, inventories, demand reductions, or domestic ramp-up capacity are available. Another country may import less directly yet be more exposed if the input is embedded in indispensable production chains and there is no quick substitute.

The post-pandemic exercises did not simply confirm the most alarmist view of globalization. Soumaya Keynes said a common headline from these studies was that countries were vulnerable in fewer products than many people thought. In some cases, the products were low-stakes goods such as umbrellas: inconvenient, but not systemically important.

Yet the exercises were also hard to evaluate from the outside because governments were often reluctant to say exactly which products had been identified. Keynes said that reluctance was understandable. If a government has worked out precisely where it can be coerced, it may not want to broadcast that information.

On the specific 2025 cases, Keynes drew a distinction. Rare earths were not a surprise. China had weaponized its dominance in rare earths in 2010, and governments had long known of the dependency. The semiconductor episode was less likely, in her view, to have shown up cleanly in mapping exercises, partly because the vulnerability was complex and fine-grained.

Indirect exposure is the harder problem. Chad Bown pointed out that standard import statistics may show that the United States imports permanent magnets from Japan, not from China. That can create a false sense of insulation. If Japan depends on Chinese rare earths to make the magnets, then a Chinese cutoff to Japan also becomes a U.S. vulnerability.

The same logic applies to Nexperia chips. Even if the United States does not directly import the chips, it may import auto parts from Bosch in Germany; if Bosch depends on Nexperia, the U.S. auto sector is still exposed. Bown’s point was that governments need data not only on their own imports, but also on the vulnerabilities of key trading partners.

That is a much more difficult information problem than scanning customs codes for high import concentration. It requires visibility into multi-tier supply chains across countries, including the inputs used by suppliers’ suppliers. It also requires understanding whether the input is technically substitutable, commercially substitutable, and available at the necessary scale under stress.

Davis summarized the implication: the exercises can be done and did teach governments something, but they are imperfect. The data are hard to assemble, the vulnerabilities may be indirect, and the question of how much to communicate publicly is itself strategic.

The rare-earth case is uncomfortable because it was not unknown. Governments had already seen China use rare earths as a tool of economic statecraft before 2025. The harder question is why the United States and others remained exposed anyway.

Price was the first answer from Chad Bown. It is hard to give up low-cost supply, especially when private firms do not internalize the national-security cost of concentration. The cheapest supplier remains attractive until the geopolitical risk becomes immediate.

Japan’s response to China’s 2010 rare-earth restrictions was the counterexample. Bown said Japan subsidized an Australian firm, Lynas, which mines rare earths in Australia and processes them in Malaysia. The structure itself showed the policy complexity: one government subsidizing another country’s company operating in a third country. But the effort reduced Japan’s direct dependence on China from roughly 90% to roughly 60%.

Davis emphasized the reported scale of that effort: about $250 million in subsidies. He described that as “chump change” relative to the damage a rare-earth chokepoint can cause, especially because rare earths are often used in small quantities but are critical to many production processes. If that is what it took Japan to reduce direct dependence so sharply, he asked why the United States had not done more.

The 2010 response reflected a different view of the trading system. Soumaya Keynes noted that China’s rare-earth restrictions then applied through a licensing regime, and that Japan, the United States, and Europe all responded in different ways. The United States and Europe sued China at the World Trade Organization. At the time, Keynes said, the rules-based system was supposed to work: if China was shown to be breaking the rules, the expectation was that it would stop.

That expectation now looks dated, but Keynes cautioned against treating it as irrational in its own time. The United States was operating from a position of confidence. It had the dollar network, major technology strengths, and its own means of coercive leverage. In such a world, policymakers could believe that mutual vulnerability and rules would restrain abuse.

Bown said the Biden administration had been trying to address the issue, while acknowledging Davis’s point that the sum involved in Japan’s case was not large relative to the stakes. His emphasis was that money alone is not enough. It takes concerted government effort. The private sector has little incentive to diversify away from the cheapest source, and even less incentive when China can respond to new entrants by flooding the market and driving down prices. Bown said China’s state-owned enterprises, including two that dominate rare earths, can discourage entry by increasing supply and making alternative production unprofitable.

Japan’s reduction from 90% to 60% also has to be interpreted carefully. Keynes said that figure likely referred to direct dependence, not rare earths embedded in other products that Japan imports. That distinction became more important because China’s export-control scheme had broadened. The issue was no longer only raw ingredients, but anything containing a Chinese rare earth. If dependency is measured across all products containing Chinese rare earths, Keynes said, Japan’s exposure likely fell by much less.

The defense sector alone may also be too small to solve the scale problem. The defense industry uses rare earths, but in some cases it does not create enough demand to sustain ex-China production. Automobiles and other private-sector uses may provide the scale required for a commercially viable supply chain. That means a national-security intervention cannot simply subsidize a narrow defense supply chain. It may need incentives that change private-sector behavior as well.

Some supply-chain vulnerabilities are consequences of earlier policy choices. Steven Davis cited U.S. gasoline price controls in the 1970s as having worsened the economic impact of the OPEC oil shock. He then turned to rare earths: the United States has minerals in the ground, but, as he put it, regulation has made mining and refining cost-prohibitive. He also pointed to Europe’s natural-gas dependence on Russia and asked why Europe had not developed a fracking industry, suggesting that regulation and property-rights arrangements had discouraged it.

Chad Bown agreed that policy choices matter, especially for rare earths. Rare-earth processing is dirty, polluting, and can involve toxic chemicals and, in some cases, nuclear-level risks. If a trusted trading partner is willing and able to do it, he said, it is attractive to let that partner bear the environmental burden. The United States used to be the dominant supplier before the 1990s, but China offered to do the work, and the United States in effect helped transfer the technology.

But concentration is not always evidence of bad policy. Some concentration reflects genuine efficiency. Economists praise agglomeration economies: smart people locating near one another, knowledge spillovers, and large fixed capital costs spread over high output. Bown used leading-edge semiconductor manufacturing as the example. TSMC’s concentration in Taiwan was not mainly because U.S. regulation pushed the industry away. It was because the concentration was efficient under free trade.

In a world of trust and limited shocks, such concentration can make economic sense. In a world of geopolitical rivalry, export controls, coercion, earthquakes, and climate shocks, the same concentration becomes a vulnerability. The policy challenge is that yesterday’s efficiency can become today’s exposure without having been foolish at the time.

Soumaya Keynes largely agreed with Bown’s formulation, and Bown observed that governments are recognizing these risks at different speeds. The United States had become more focused on China as a national-security threat earlier than Europe, while Europe had been more focused on Russia. But Europe, too, was affected by the rare-earth, permanent-magnet, and Nexperia episodes, and is now being forced to confront China-related economic security risks more directly.

Davis placed this in a broader geopolitical shift. For most of the postwar period outside the Soviet and communist sphere, the system had a leading actor that largely set the rules. He described that world as one in which a relatively enlightened hegemon supported the trading system. That world no longer exists in the same way. It is unclear whether a new leading actor will emerge or whether countries will operate for a long time in a system where no one is fully in charge. In such a system, scale economies and locational efficiency cannot be the only considerations, because supply-chain dependencies can be used deliberately as tools of geopolitical leverage.

A less interventionist response would start with information. Governments could identify vulnerabilities and publicize them, either broadly or privately to firms. Statistical agencies already provide information that private actors use to make better decisions. Supply-chain vulnerability mapping could, in principle, work similarly: the government identifies that a company’s product is critical and that its supply chain contains a dangerous exposure, then urges the company to act.

Chad Bown was skeptical that information alone could go far enough. Governments should survey firms and improve data, but companies often cannot or will not provide the necessary information. Large multinationals told him they could not get tier-two suppliers to disclose their own suppliers. The reason was commercial: those suppliers feared being cut out of the chain or facing tougher contract negotiations if they revealed too much.

An emerging private sector is trying to fill the supply-chain data gap. These firms combine datasets that government agencies sometimes cannot share with one another because of regulatory limits. They make supply-chain connections, encourage private companies to participate by offering data access, and use artificial intelligence where hard data are missing.

Bown saw value in that work but also discomfort. AI systems, he said, are good at producing an answer to a prompt and poor at communicating confidence in that answer. In this setting, either type of error can be costly: a model may indicate a vulnerability that does not exist and prompt a policy response where none is needed, or miss a real vulnerability and create false reassurance.

The deeper problem is incentives. Soumaya Keynes framed the issue this way: even if the information problem were solved, would the private sector reduce vulnerabilities on its own? Companies do complain that they only learn about vulnerabilities when a seventh-tier supplier reports a problem, so they do want more information. But in many major cases, lack of information is not the binding constraint. Firms already know that high-end chips are heavily dependent on Taiwan, or that China is a major supplier of reagents used to make active pharmaceutical ingredients. The problem is not ignorance; it is that the incumbent supply chain is efficient and cheap, switching is costly, and no single firm wants to bear the cost.

Davis favored better information partly because the alternatives—subsidies, tariffs, equity stakes, price floors, and similar tools—invite rent-seeking. Once government begins subsidizing rare-earth mining, it may be difficult to stop. Constituencies form around subsidies. Even if a subsidy makes sense at a point in time, the political system may preserve it after the reason has disappeared. Davis cited corn ethanol as an example of a subsidy that keeps returning despite contested national-security and environmental rationales.

Keynes agreed that subsidies are unattractive tools. But she argued that the economist’s traditional posture—subsidies are bad, governments use them badly, therefore do not use them—no longer works. In the trade-war environment described in the discussion, none of the tools are good. Tariffs, subsidies, export controls, and other interventions are blunt and prone to unintended consequences. The question, in her view, has to shift from whether to use them at all to how to use them better.

If governments decide that information is not enough, the design of intervention becomes the main issue. For rare earths, the possible tools include allied supply chains, government equity stakes, price floors, tariffs on alternative sources, and exemptions or faster treatment under environmental regulations and permits. The question is not how to make those tools clean. It is how to use them less badly.

Chad Bown treated the CHIPS Act as the more instructive case. The CHIPS Act was a bipartisan effort to move some high-end semiconductor production capacity back to the United States because of concern about concentration in Taiwan. It used different forms of support. Grants required applications, screening by government officials, paperwork, and time. One company executive told Bown and Keynes it was “the most expensive free money” the company had ever received. Tax credits were simpler: firms got a reduction in tax liability if they undertook qualifying activity.

Bown said the United States was effective at encouraging additional high-end semiconductor production capacity. TSMC, Intel, and Samsung all developed U.S. capacity. But the policy revealed a further problem: supply-side capacity does not guarantee demand-side use. Chip design companies such as Nvidia and AMD could look at the new diversification options and still prefer to have TSMC make their chips.

That is why Bown argued that supply-side policies may not be enough. Governments may also need to address the cost of switching. He cited an idea from Dan Kim, formerly the chief economist on the CHIPS Act implementation team and now at TechInsights. For companies like Nvidia or AMD, porting chip designs from TSMC to Intel or Samsung can cost a couple hundred million dollars. A government subsidy could defray that switching cost and induce firms to try alternative manufacturers, increasing diversification away from TSMC.

Steven Davis liked that form of intervention because it has a more natural endpoint. If the subsidy is about switching costs, then once the switch has been made, the rationale for the subsidy declines. He compared that favorably to Operation Warp Speed during the pandemic, which he described as self-liquidating: once vaccines had been demonstrated to work with reasonable success, the intervention did not need to continue in the same way. Conventional subsidy programs, by contrast, often lack natural endpoints and become entrenched.

Soumaya Keynes emphasized humility. She and Bown found, while researching their book, that subsidies had not been a fashionable topic of economic study for a very long time, so the evidence base on which instruments work best is incomplete. A newer wave of studies is trying to examine the question, but even that literature often asks whether subsidies produce profitable, self-sustaining businesses. That may not match the policy objective in the current setting. If the goal were simply profit maximization, firms would probably stay with the incumbent supplier. The security goal may instead be production volume, redundancy, or even emergency production capacity.

Davis sharpened that point: the relevant objective may be capacity that can be drawn upon in an emergency, not current production volume. Keynes agreed that, in a country as large and innovative as the United States, enough money and time can build capacity. But the hard question is the tradeoff: how much cost, how much waste, which instrument, and what measure of success.

China’s industrial policy entered the discussion as both proof of capacity building and a cautionary case. Soumaya Keynes said that if a government throws enough money at capacity, eventually it will get capacity. Steven Davis responded that this is not a model the United States should emulate wholesale. He said U.S. GDP per capita is about six times China’s and described the U.S. economic model as much more successful. The challenge, for him, is to address economic and national-security vulnerabilities cost-effectively while preserving the virtues of a more market-oriented, decentralized system.

Keynes argued that the right lesson from China is not simply “massive state intervention works.” In particular sectors, she said, China has intense competition. That competition helps prevent a single subsidized, complacent industry from sitting inefficiently on government support. It also contributes to innovation and industrial success, even if China has sometimes gone too far.

Davis agreed that competition is powerful, but emphasized that competition works partly because failures exit. If subsidies are large enough, governments can sustain many loss-making enterprises. In China, he said, this has produced waste and excess often described as involution. Keynes agreed with that characterization.

Davis also pointed out that when China overbuilds subsidized manufacturing capacity, much of the surplus can flow to consumers abroad through lower prices. Those consumers are often in the rest of the world. In discussions of security vulnerabilities, he said, that benefit should not be forgotten. China’s model creates risks, but it has also subsidized foreign consumers. The West should not respond by copying the mistakes of that model, especially if the result is to subsidize consumers elsewhere without clear security gains.

Chad Bown drew out the implication: if the United States, Europe, Japan, and others respond by moving manufacturing activity outside China for national-security, economic-security, or political reasons, the era in which China subsidized their consumers will end. Prices will rise for some goods that consumers have taken for granted. If that cost is necessary, Bown said, policymakers should at least try to make it as small as possible.

That requires scale. In Bown’s view, the least-cost approach would involve cooperation among the United States, Europe, Japan, and other countries that largely see the security problem similarly. The obstacle is that the United States has not always appeared willing to work with them; instead, from their perspective, it has also imposed tariffs on allies.

Davis agreed strongly. The final policy tension, then, is not whether vulnerabilities exist or whether governments have a role. It is whether democratic, market-oriented countries can build enough redundancy and resilience without replicating the waste of a command-and-subsidy model—and whether they can do so together rather than fragmenting the very scale that would make resilience less costly.