The Semiconductor Shortage Was a Dress Rehearsal. The Real Crisis Is Coming.

In the summer of 2021, American consumers discovered that buying a new car had become surprisingly difficult. Not because of price — though prices rose — but because the cars simply didn’t exist in the quantities people wanted. Ford was parking F-150 pickups in enormous fields in Kentucky and Michigan, waiting for semiconductor chips before the vehicles could be completed. General Motors idled factories. The average wait time for a new vehicle stretched past six months at some dealerships. The culprit was not a war, not a natural disaster, not a financial collapse. It was a shortage of chips, primarily manufactured in Taiwan and South Korea, caused by a surge in pandemic-driven demand for consumer electronics colliding with manufacturing lead times measured in months, not days.

The shortage was genuinely disruptive. Auto industry revenue fell by hundreds of billions globally. Consumer electronics companies allocated chips through internal rationing processes. Industrial equipment manufacturers pushed back delivery schedules by a year or more. The economic damage, while significant, was bounded — the supply chain eventually cleared, manufacturing capacity gradually expanded, and by 2023 the acute phase was over.

The shortage also revealed something that most governments and CEOs had preferred not to think about: essentially the entire world’s supply of advanced semiconductors was being produced in a small geographic area by a small number of companies, with one company — TSMC — holding a position of extraordinary and concentrated power over global technology. This concentration had been building for decades. Almost nobody had noticed because supply had never meaningfully failed.

The 2021 shortage was caused by demand disruption. Factories were operating; the problem was that the mix of products demanded changed faster than manufacturing could adapt. Consumer electronics demand surged. Automotive demand initially fell (manufacturers cut orders) and then surged back (manufacturers couldn’t get chips back on the queue fast enough). The underlying manufacturing infrastructure remained intact throughout. TSMC’s fabs in Hsinchu and Tainan never stopped running. What the world experienced was a scheduling problem, not a production problem.

A Taiwan Strait conflict would be categorically different in kind, not just degree. If the People’s Republic of China were to impose a naval blockade of Taiwan — even a partial one that did not involve direct military strikes on manufacturing facilities — the effect on global chip supply would not be a scheduling problem. It would be a destruction problem. TSMC’s most advanced fabs, producing chips at the 3nm and 2nm nodes that power the latest AI systems, smartphones, and server processors, are in Taiwan. So are the facilities of several other significant chipmakers. A blockade lasting six months would not produce a chip shortage of the 2021 variety. It would produce something closer to an industrial shutdown.

To understand why, consider what semiconductor manufacturing requires beyond the physical facilities. The fabs need a continuous supply of specialty gases — nitrogen, hydrogen, helium, neon — several of which have highly concentrated supply chains of their own. They require photolithography equipment, primarily from ASML in the Netherlands, that takes years to manufacture and cannot be air-freighted in. They require chemicals that require global supply chains and cannot be substituted quickly. They require a workforce of specialized engineers whose skills took years to develop and who cannot be replaced by recruiting generalists. A blockade disrupts all of these inputs simultaneously.

What “reshoring” actually means is worth examining carefully, because the political discussion around it in the United States and Europe has been notably imprecise. When politicians invoke “bringing chip manufacturing home,” they typically gesture toward initiatives like the CHIPS and Science Act, which allocated approximately $52 billion to support domestic semiconductor manufacturing in the United States. This was a real commitment and a genuine policy shift.

But what does it actually buy? The CHIPS Act is primarily designed to support manufacturing of trailing-edge and mid-tier chips — the kinds of processors used in automotive, defense, and industrial applications that were hit hardest by the 2021 shortage. It has supported expansion of existing Texas Instruments and GlobalFoundries facilities. More ambitiously, it has underwritten the construction of TSMC’s Arizona fabs, which represent the first significant onshore production of advanced semiconductors in the United States in decades.

The TSMC Arizona story is instructive about the genuine difficulty of what reshoring involves. The project has run into serious difficulties. Construction of the first fab — targeting 4nm chips — was delayed by over a year, in part because of a shortage of the specialized technicians needed to install and calibrate the equipment. TSMC imported thousands of engineers from Taiwan to compensate, which created its own political controversy. The second fab, planned for 3nm production, has faced further delays. The third fab, planned for 2nm, remains aspirational. Even in the optimistic scenario where all three fabs operate as planned, they represent a small fraction of TSMC’s total advanced capacity in Taiwan.

The fundamental problem is that advanced semiconductor manufacturing is not simply a matter of having factories. It requires an ecosystem: materials suppliers, equipment manufacturers, specialized training institutions, a talent pipeline, and decades of accumulated process knowledge. Taiwan has built this ecosystem over fifty years, with enormous government support and strategic national commitment. The United States decided, sometime in the 1990s and 2000s, that manufacturing was not where its comparative advantage lay and allowed that ecosystem to atrophy. Rebuilding it takes time measured in decades, not years or legislative cycles.

There is a cruel irony in the CHIPS Act’s timing. The legislation was motivated substantially by concern about the geopolitical concentration of advanced chip production in Taiwan. But the most advanced node of TSMC Arizona — when it eventually operates — will still be behind TSMC Taiwan’s leading edge by two to three years. The facilities that matter most for cutting-edge AI and military applications remain in Hsinchu. The gap is not closing fast.

What would a twelve-month disruption to Taiwan’s chip supply actually look like for the global economy? This is a question that has been modeled by defense analysts, insurance companies, and supply chain researchers with results that are sobering to read carefully. The semiconductor industry directly generates revenue on the order of $600 billion annually, but its downstream economic significance is far larger. Chips are inputs to virtually every modern manufactured good: vehicles, appliances, industrial equipment, telecommunications infrastructure, data centers, medical devices, weapons systems. A study by researchers at the Atlantic Council estimated that a one-year Taiwan Strait crisis would produce global GDP losses of several trillion dollars — substantially larger than the 2008 financial crisis.

The concentration extends beyond manufacturing. The equipment required to make advanced chips is itself extraordinarily concentrated. ASML, a Dutch company, is the sole manufacturer of extreme ultraviolet lithography machines — the tools required to produce chips at the most advanced nodes. One company, in one country, makes the equipment without which cutting-edge chip manufacturing is impossible. This creates a second chokepoint that China cannot circumvent regardless of how much it spends on its domestic chip industry (and it is spending enormous amounts). A blockade of Taiwan doesn’t even need to target fabs directly; interdicting the supply of ASML machines and specialty materials achieves similar effects more gradually.

China’s own semiconductor industry has been accelerating rapidly in response to American export controls, which have restricted its access to advanced chips and manufacturing equipment since 2022. Chinese firms have made notable progress at trailing-edge nodes and have produced chips at 7nm without access to EUV lithography — an impressive feat of engineering improvisation. But leading-edge production at 2nm and below remains out of reach for the foreseeable future, not because of a lack of investment or will, but because of the accumulated ecosystem gap.

This matters because it affects the calculus of a potential conflict. China would suffer enormously from disruption to Taiwan’s manufacturing — Chinese consumer electronics, automotive, and industrial sectors are deeply dependent on Taiwanese chips. But China’s leadership has demonstrated tolerance for economic pain in service of strategic objectives. The more relevant question is whether the deterrent effect of mutual economic destruction is sufficient to prevent conflict, or whether, as with some historical parallels, political and nationalistic pressures eventually override rational economic calculation.

The real lesson of the 2021 shortage is not that supply chains were fragile in ways nobody expected. Supply chain analysts had been writing about semiconductor concentration risk for years. The lesson is that knowledge of a risk and preparation for a risk are different things, and that market actors rarely invest in resilience against tail risks when building that resilience is expensive. The shortage forced a global reassessment. The CHIPS Act, Europe’s Chips Act, Japan’s domestic semiconductor subsidies — all of these are products of that reassessment.

They are not sufficient. The timelines for building out diversified advanced semiconductor capacity are measured in decades. The geopolitical pressure points around Taiwan are measured in years or less. There is a window of acute vulnerability between now and when any meaningful supply diversification actually exists — a window whose duration is uncertain but whose existence is not.

The 2021 shortage was a dress rehearsal with a forgiving script: the factories kept running, the disruption was temporary, the economy recovered. Dress rehearsals are valuable precisely because the consequences of failure are limited. The real performance, if it comes, will not offer that mercy.