One Island and One Dutch Machine

Two facts, and almost all of modern technology policy follows from them.

The most advanced logic chips in the world are manufactured overwhelmingly in Taiwan, principally by TSMC. The machines required to manufacture them at that level come from one company: ASML, in the Netherlands, which is the only firm on earth producing extreme ultraviolet lithography systems.

That is two chokepoints in series, and the phrase matters. In series means either one is sufficient. A country that controls the machine can stop the chips without touching the island; a country that controls the island holds the manufacturing whatever the Netherlands does. There is no redundancy anywhere in the arrangement.

This is why export controls since 2019 have aimed at tools rather than at chips. Restricting chips creates a demand that a competitor can eventually meet. Restricting the machine that makes the chips attacks the ability to compete at all, and there is only one place to apply the restriction.

The obvious response, build your own, is harder than it sounds in a specific way that is worth understanding.

A leading fabrication plant takes years and tens of billions of dollars, and those are the easy parts. The difficult part is process knowledge: thousands of small, undocumented, empirically discovered adjustments that make a given tool produce acceptable yields in a given building with given materials and given people. It exists in the heads and habits of a workforce that has been doing it continuously for decades. Money buys the building. It does not buy the twenty years.

The strategic argument about Taiwan turns on what this concentration does.

The silicon shield thesis says it protects the island: nobody can afford the factories to stop, so nobody will risk them. The counter-thesis says the same indispensability is exactly what makes the island worth controlling, and that a shield made of something valuable is also a prize. Both are reasonable and the question is empirical in the worst possible way, which is that it can only be settled by an event nobody wants.

TSMC is building in Arizona under a much enlarged commitment, and it is worth being precise about what that moves. It moves capacity. It does not move the tool suppliers, the materials chemistry, the specialised component makers, or the engineering base behind them, all of which remain concentrated elsewhere, much of it in Taiwan, Japan and the Netherlands. A fab in Arizona that depends on a Dutch machine and Japanese chemicals has relocated one layer of a stack several layers deep.

The last point is the one the whole book has been building towards.

Nobody planned this. There was no conspiracy and no strategic decision to concentrate the world's most critical manufacturing on one island. It emerged from decades of specialisation in which every firm did what it did best and bought the rest from whoever did that best, which is precisely what economic theory recommends and what shareholders required. Efficiency was traded against resilience continuously, by thousands of rational actors, none of whom was responsible for the total.

That is the shape of modern vulnerability, and you will see it twice more before this book ends.