The Country That Holds the AI Supply Chain Hostage
In the small Dutch city of Veldhoven, in a facility that looks from the outside like an unremarkable industrial complex, a company called ASML manufactures machines that might be the most strategically important objects on Earth. Each machine takes roughly a year to build, costs around 200 million euros, requires more than 100,000 components from suppliers across three continents, and is shipped in forty freight containers when it is finally delivered to a customer. There are only a handful of companies in the world that can afford to buy one, and every single one of the world’s most advanced chips is made using them.
ASML has a global monopoly on extreme ultraviolet lithography — EUV — the process by which the microscopic circuitry of advanced semiconductors is printed onto silicon. No other company, in any country, has successfully built a competing EUV machine. And because advanced chips cannot be made without EUV, and advanced chips are required for cutting-edge AI, smartphones, high-performance computing, and an increasing number of military systems, the Netherlands — population 17 million, GDP roughly that of Sweden — has stumbled into a position of global geopolitical power that its foreign policy establishment was entirely unprepared for.
To understand how this happened, you need to understand what EUV lithography actually does. Making a chip requires taking a design that specifies the location of billions of individual features — transistors, interconnects, capacitors — and transferring that design onto a silicon wafer with extraordinary precision. The transfer process works like a photographic negative: light is shone through a patterned mask, and where the light strikes the wafer, it chemically modifies a light-sensitive coating, allowing the underlying silicon to be etched away in the desired pattern.
The challenge is that the features on advanced chips are now so small — just a few nanometers across — that conventional light cannot resolve them. The wavelength of the light is too long: it blurs the image before it reaches the wafer, the way a paintbrush is too crude to write fine text. EUV uses light with a wavelength of just 13.5 nanometers — far shorter than visible light, in the extreme ultraviolet range — which allows it to resolve the tiny features needed for advanced chips.
Generating light at 13.5 nanometers is not simple. It requires firing high-powered lasers at tiny droplets of molten tin fifty thousand times per second, generating a plasma that emits EUV light, then collecting and directing that light through a system of mirrors polished to a smoothness measured in fractions of an atom. The mirrors must be kept in a vacuum, because EUV light is absorbed by air. The entire machine is a marvel of precision engineering that sits at the boundary of what physics permits.
ASML did not set out to build the world’s most geopolitically sensitive piece of manufacturing equipment. The company emerged from a joint venture between Philips and ASMT in 1984, grew by doing the unglamorous work of making photolithography machines more and more precise, and invested heavily in EUV at a time when many in the industry believed the technology would never be commercially viable. When it finally worked — when ASML delivered the first production-capable EUV machines to chipmakers around 2018 — the company found itself in a position of extraordinary leverage that it had not fully anticipated.
The geopolitical significance became undeniable when the United States began pressuring the Dutch government to restrict ASML’s exports to China. ASML had been selling older lithography machines to Chinese chipmakers, and the US wanted those sales stopped — or at minimum, restricted so that Chinese chipmakers could not acquire the most advanced EUV equipment. The Dutch government, after years of deliberation and under sustained American pressure, ultimately complied, initially restricting the most advanced EUV machines and later extending restrictions to older deep ultraviolet machines that China was using as a workaround.
The Strait of Hormuz analogy is apt here, though it requires a slight adjustment. The Strait of Hormuz is a geographic chokepoint: thirty percent of the world’s seaborne oil passes through a waterway that is just twenty-one miles wide at its narrowest. Any power that controls that strait controls oil flows to a large portion of the global economy. ASML is a technological chokepoint: the world’s ability to produce advanced chips passes through a single company’s production floor in a single Dutch city. Any government that controls that company’s export policy controls chip flows to any target it chooses.
For the Netherlands, this has been a genuinely disorienting experience. Dutch foreign policy has historically been oriented around multilateralism, international institutions, and the careful balancing of relationships with larger powers. The Netherlands is a trading nation — its wealth was built on being a node through which goods and capital flow rather than on wielding coercive power. The discovery that it possesses, in ASML, a tool of coercive power more precisely targeted than anything in its military arsenal has forced a rapid rethinking of how Dutch foreign policy works.
The Dutch government has been, by its own admission, learning on the job. The initial ASML export restrictions were imposed somewhat reluctantly, under American pressure and against the preferences of some Dutch officials who worried about the precedent of weaponizing a commercial company’s product for geopolitical ends. But as the strategic importance of the technology became clearer, and as other countries began to recognize and acknowledge the Netherlands’ leverage, the Dutch foreign policy establishment has become more comfortable wielding it — though still with considerable unease about the long-term consequences.
China’s response has been predictable and, so far, only partially successful. The Chinese government has invested heavily in domestic alternatives to ASML, pouring resources into companies like SMEE, the Shanghai Micro Electronics Equipment company, which is developing its own lithography machines. But the gap between what SMEE can produce and what ASML produces is not a gap that can be closed quickly through investment. It represents decades of accumulated process knowledge, supplier relationships, and engineering refinements that do not exist anywhere else. Industry analysts generally estimate that China is a decade or more behind ASML’s current capabilities.
In the meantime, Chinese chipmakers are being forced to make do with older lithography technology. This does not prevent China from making chips — there are vast quantities of chips for which advanced EUV is not required, and China has significant capacity to produce those. But it does mean that Chinese chips for the applications that matter most for AI — advanced GPUs, high-bandwidth memory, leading-edge processors — will lag behind what TSMC and Samsung can produce using ASML’s most advanced machines.
The broader lesson of the ASML situation is about the distribution of geopolitical power in a world of hyper-specialized supply chains. The conventional metrics of national power — military spending, GDP, population size — poorly capture the leverage that comes from controlling a specific, irreplaceable node in a global supply chain. Qatar controls a significant fraction of the world’s liquefied natural gas exports and has wielded that leverage to punch far above its demographic weight in regional politics. Taiwan controls the production of advanced chips. The Netherlands controls the machines that make those chips possible.
These are new forms of power, and the international system is still figuring out how to incorporate them into the established frameworks of diplomacy, alliance, and deterrence. The traditional alliance structures — NATO, bilateral security agreements — were built around military power and territorial defense. They were not designed to manage a world in which a medium-sized European country’s export licensing decisions can meaningfully affect the military capabilities of a country of 1.4 billion people.
ASML, for its part, navigates this with the particular discomfort of a company that built something extraordinary in pursuit of engineering excellence and then found itself at the center of a geopolitical conflict it never sought. Its machines were designed to help humanity make better computers. They do that, with extraordinary effectiveness. That they also happen to be the chokepoint through which the AI race must pass is a consequence of physics and of the extraordinary difficulty of the engineering problem ASML solved — not of any strategic intention. The geopolitical power was always latent in the technology. It just took a semiconductor cold war to make everyone notice.