The Chip That Started the Second Cold War

In October 2022, the United States government did something that most of the world barely noticed and almost no one fully understood. It published a set of export control rules targeting advanced semiconductors and the equipment used to make them, barring their sale to Chinese companies and Chinese-affiliated entities without a license that would almost never be granted. The announcement was buried in regulatory language. There were no press conferences, no presidential speeches, no dramatic declarations of policy.

What happened next will define the next thirty years.

To understand why a set of export control rules constitutes a geopolitical rupture of historic significance, you have to understand what semiconductors actually are and what makes certain ones so hard to build. A modern advanced chip — the kind that powers AI training, the kind that goes into a smartphone flagship, the kind that the US government decided to restrict — contains billions of transistors etched onto a piece of silicon roughly the size of a fingernail. The individual features on these chips are measured in nanometers: a few nanometers, at the cutting edge. A human hair is roughly 80,000 nanometers wide.

Creating structures at this scale requires a manufacturing process of extraordinary precision, and that process depends on a tiny number of companies with capabilities that no one else on earth has been able to replicate. TSMC — the Taiwan Semiconductor Manufacturing Company — fabricates the most advanced chips. ASML, headquartered in the Netherlands, makes the only machines capable of printing the most advanced circuitry. And a handful of American companies, including Cadence and Synopsys, provide the software tools without which none of this is possible.

The US export ban effectively weaponized this concentration. By restricting the sale of advanced chips and the equipment to make them, Washington told Beijing: we have decided you should not have the industrial tools necessary to build competitive AI infrastructure. The logic was straightforward, if aggressive. Advanced AI requires advanced chips. Advanced chips require advanced manufacturing. Advanced manufacturing requires equipment and software that American-aligned companies control. Therefore, restricting that equipment and software restricts Chinese AI. At least in theory.

China’s response was predictable in its broad strokes if not in the details. The Chinese government accelerated its existing program to develop a domestic semiconductor industry, throwing hundreds of billions of renminbi at the problem through a combination of state funds, directed bank lending, and government procurement commitments. Companies like SMIC, the leading domestic chipmaker, were tasked with bridging a gap that the industry consensus said would take a decade to close. HUAWEI, which had been the most visible early victim of the chip restrictions, launched a domestic chip program that surprised many analysts when it produced a reasonably capable advanced processor in 2023.

But the surprise should not be overstated. What Huawei achieved was impressive given the constraints, but it was still generations behind the frontier. The gap in semiconductor manufacturing is not primarily about engineering talent or investment capital — China has both in abundance. It is about accumulated manufacturing knowledge, about the tens of thousands of incremental process improvements that TSMC and its equipment suppliers have made over forty years of doing almost nothing else. You cannot buy that knowledge. You can barely steal it, because much of it lives not in documents but in the tacit expertise of engineers who have spent their careers on one specific problem.

Taiwan’s position in all of this is genuinely precarious in a way that does not have a clean historical precedent. The island sits ninety miles off the coast of mainland China, which claims it as a province and has never renounced the use of force to bring it under control. It also hosts the facilities that produce essentially all of the world’s most advanced chips. This is not a coincidence — TSMC was founded in Taiwan with government support, and its presence on the island is a product of deliberate industrial policy over decades — but the strategic implications of this geography are now impossible to ignore.

There is a concept sometimes called the Silicon Shield: the idea that Taiwan’s indispensability to the global chip supply chain protects it from Chinese military action because any invasion would destroy the infrastructure that Beijing would need the island for, while simultaneously triggering a global economic catastrophe that would make the 2008 financial crisis look modest. This is partly reassuring and partly terrifying. It means that the thing protecting Taiwan from invasion is not primarily military deterrence but economic interdependence — and that interdependence, while real, depends on an equilibrium of rational actors that geopolitical crises have a history of disrupting.

The United States, the European Union, Japan, and South Korea have all launched major efforts to reduce their dependence on Taiwan by constructing domestic chip manufacturing capabilities. The CHIPS Act in the US, passed in 2022, committed over fifty billion dollars to this goal. TSMC has announced factories in Arizona, in Germany, and in Japan. Intel and Samsung are expanding advanced manufacturing in their home countries. The political consensus is that the concentration of advanced chip manufacturing in one small island is an unacceptable strategic vulnerability.

But building a semiconductor fab is not like building a car factory. The lead times are measured in years, the costs in tens of billions of dollars, the expertise requirements in decades of specialized experience. The Arizona TSMC factory, despite enormous subsidies and political support, has faced repeated delays because qualified engineers who know how to operate advanced fabs are a globally scarce resource concentrated in Taiwan. You cannot simply transplant a manufacturing ecosystem by building a building.

What the chip restrictions have most clearly demonstrated is the geography of the technology supply chain. Every product that depends on advanced semiconductors — which is to say, increasingly, every consequential product — depends on a chain of manufacture that runs through a small number of locations, most of them in East Asia, and through a small number of companies, most of them with American intellectual property at their core. The internet era created the illusion that geography had become irrelevant to economic competition. The semiconductor crisis has demolished that illusion.

The arms race framing is imperfect but illuminating. In the original Cold War, the US and USSR competed to build weapons of increasing destructive capability, with the arms race eventually consuming a significant fraction of both economies’ productive capacity. The semiconductor competition has a different character — it is primarily about economic and technological advantage rather than physical destruction — but it shares the essential feature of a competition that neither side can afford to lose and that both sides are devoting enormous resources to winning.

The difference, and it matters, is that the Cold War was essentially bilateral. It was two superpowers, two blocs, two sets of allies. The semiconductor competition involves many more players. South Korea, Japan, the Netherlands, Taiwan itself — all of these countries are now significant geopolitical actors in a drama that would previously have been determined almost entirely by Washington and Beijing. The Netherlands has imposed its own export restrictions on ASML’s most advanced equipment, discovering in the process that it possesses a form of geopolitical leverage it never anticipated and is not entirely sure how to use. South Korea has to balance its economic relationship with China — its largest trading partner — against its security relationship with the United States, which has been pressing Seoul to align with the chip restrictions.

This is the new normal of tech geopolitics: a world where the supply chain for a product no one can easily explain is the fault line along which the major powers are aligning and realigning. It is not a cold war in the original sense. There are too many actors, too many economic dependencies, too much genuine uncertainty about where the lines are drawn. But it shares the Cold War’s defining characteristic: the sense that a technological competition has become the primary arena in which the question of which civilization model prevails in the twenty-first century will be decided.

A piece of silicon smaller than a fingernail. Billions of transistors. And underneath it all, a question that is much older than semiconductors: who controls the most essential tools of the most powerful technology of an era controls more than the technology. They control the future.