The ASML Chokepoint

Somewhere in Veldhoven, a small city in the southern Netherlands, machines the size of buses are being assembled in cleanrooms. Each machine costs €350 million. Each takes 18 months to build. Each requires three Boeing 747s to ship. And each is the only way to manufacture the most advanced semiconductors on Earth.

ASML Holding is not a household name. It has no consumer products, no retail presence, no brand recognition outside the semiconductor industry. But every chip in every iPhone, every AI accelerator in every data center, every advanced processor in every military system passes through ASML's machines before it exists.

The company has 100% market share in extreme ultraviolet lithography — the only technology capable of printing transistors at the densities required for leading-edge chips. There is no second supplier. There is no alternative. There is only ASML.

This is the story of how a single Dutch company became the gatekeeper of Moore's Law — and what that means for the geopolitics of technology.

A semiconductor is a photograph printed in silicon. Light passes through a mask containing the chip's circuit pattern, then through a lens that shrinks the image, and finally onto a silicon wafer coated with light-sensitive material. Where light hits, the coating changes. Acid washes away the changed parts. What remains is the circuit.

The challenge is the light.

For decades, chipmakers used deep ultraviolet (DUV) light with a wavelength of 193 nanometers. But physics imposes a limit: you cannot print features smaller than half the wavelength of the light you're using. By the 2010s, DUV was hitting its wall. Tricks like multiple patterning — printing the same layer multiple times with slight offsets — extended the technology, but at enormous cost and complexity.

EUV uses light with a wavelength of 13.5 nanometers — nearly fifteen times shorter than DUV. At this wavelength, features can be printed at scales approaching individual atoms. But 13.5nm light is not light in any conventional sense. It's absorbed by air. It's absorbed by glass. It's absorbed by almost everything.

To generate EUV light, ASML's machines fire a high-powered CO2 laser at tiny droplets of molten tin — 50,000 droplets per second, each one hit twice. The first pulse flattens the droplet into a pancake; the second vaporizes it into a plasma that emits EUV radiation. The light is then collected by the most precise mirrors ever manufactured — polished to atomic smoothness — and directed through a vacuum chamber onto the wafer.

The machine contains over 100,000 components, 3,000 cables, and several kilometers of hoses. It operates in a vacuum. Its light source generates temperatures hotter than the surface of the sun. Its mirrors must be positioned with picometer accuracy — trillionths of a meter.

No single company could build this. ASML doesn't try.

ASML is an integrator, not a manufacturer. The company's true competitive advantage is not its own engineering — though that is formidable — but its position at the center of a supply chain that took decades to assemble and cannot be replicated.

Carl Zeiss SMT manufactures the mirrors. These are not ordinary mirrors — they're multilayer coatings of molybdenum and silicon, 100+ alternating layers each just a few nanometers thick, polished to a surface roughness of less than an atom. If the mirror were scaled to the size of Germany, the largest bump would be one millimeter high. Zeiss has been developing this capability for over 20 years. ASML owns 25% of Zeiss SMT.

Cymer, acquired by ASML in 2013 for $2.5 billion, builds the light source. The tin-droplet laser system is the single most complex component — and the one that delayed EUV commercialization by nearly a decade. Getting sufficient power from the light source (250+ watts, sustained) was the critical bottleneck until the late 2010s.

These relationships are not supplier contracts. They are decades of co-investment, shared R&D, and mutual dependency. The knowledge required to build EUV components exists in a handful of facilities in the Netherlands, Germany, and San Diego. It is not documented in papers. It is not taught in universities. It lives in the hands and minds of a few thousand engineers.

ASML ships machines to three places: Taiwan (TSMC), South Korea (Samsung), and the United States (Intel). It used to ship to China. It no longer can.

The restriction did not come from ASML. It came from governments — first informally, then formally.

The trilateral coordination — US, Netherlands, Japan — is unprecedented. These are not sanctions against a country for human rights violations. They are technology controls designed to maintain a structural advantage in semiconductor manufacturing. The explicit goal is to keep China at least two technology generations behind the leading edge.

ASML, as a private company, has limited influence over these decisions. It has lost a major customer — China represented 15% of sales before controls. But the Dutch government controls the export licenses, and the Dutch government has aligned with American strategic priorities.

Cut off from EUV, China is pursuing three strategies. None of them closes the gap.

Strategy 1: Stretch DUV. SMIC has demonstrated 7nm-class chips using older DUV equipment and multiple patterning — the same technique Intel and TSMC used before EUV. The Huawei Mate 60 Pro, which surprised Western analysts in 2023, contained a SMIC-manufactured 7nm chip. But multiple patterning is expensive, slow, and yield-constrained. It cannot scale to 5nm or below without EUV.

Strategy 2: Domestic lithography. Shanghai Micro Electronics Equipment (SMEE) is China's national champion in lithography. Its most advanced tool produces 90nm chips — roughly 2004-era technology. The gap to EUV is not one generation; it is twenty years of accumulated learning and supply chain development. SMEE has announced plans for 28nm by 2025. Western analysts are skeptical.

Strategy 3: Stockpiling. Before the 2023 restrictions took full effect, Chinese foundries accelerated purchases of ASML DUV systems. ASML's China revenue spiked as customers front-loaded orders. But equipment degrades. Spare parts require export licenses. And DUV cannot reach leading-edge densities regardless of how many machines you own.

The mathematics are unforgiving. Leading-edge AI chips — the GPUs training large language models — require billions of transistors packed into a few hundred square millimeters. This density is impossible without EUV. China can build chips. It cannot build the chips that matter most for AI, advanced computing, and military systems.

ASML's backlog is the most reliable leading indicator in the semiconductor industry.

Unlike chip demand, which fluctuates with consumer cycles, EUV demand reflects multi-year capital investment decisions by the world's largest foundries. When TSMC orders EUV machines, it is committing to build a fab that will operate for a decade. When Intel orders machines, it is betting on the success of its foundry comeback. The order book reveals what the industry's most informed players believe about the future.

As of Q4 2025, ASML's backlog stands at approximately €36 billion — down from the €39 billion peak in 2023, but still representing nearly two years of revenue. The composition tells a story:

TSMC remains the dominant customer, absorbing the majority of EUV shipments for its Arizona, Japan, and Taiwan expansion. The company's AI-driven demand — Nvidia, AMD, Apple — has created a structural shortage of advanced packaging and leading-edge capacity.

Samsung has pulled back. After yield struggles at 3nm, the Korean giant has reduced near-term EUV orders while focusing on improving existing lines. Samsung's market share in advanced foundry has slipped below 15%.

Intel is the wildcard. The company's "five nodes in four years" roadmap requires massive EUV investment. Intel 18A (1.8nm-class) is scheduled for 2025 production. If Intel executes, it will be the first American company with leading-edge EUV manufacturing capability. If it fails, the strategic implications are severe.

The order book also reveals what's not happening: no Chinese foundries, no Russian customers, no alternative buyers emerging. The chokepoint is holding.

ASML is already building the next generation.

High-NA EUV — where "NA" stands for numerical aperture, a measure of the optics' light-gathering ability — uses an 0.55 NA lens instead of the current 0.33 NA. This allows even finer resolution, enabling chip geometries below 2nm. The machines are larger, more complex, and more expensive: €350 million becomes €380 million or more.

Intel has received the first High-NA system. TSMC and Samsung have orders in queue. The technology is expected to enter volume production around 2026-2027.

Beyond High-NA, the roadmap gets speculative. Hyper-NA (0.75 NA) is theoretically possible but faces severe engineering challenges. Alternative lithography approaches — nanoimprint, directed self-assembly — remain laboratory curiosities. For the foreseeable future, ASML's technology trajectory is the industry's technology trajectory.

The ASML chokepoint inverts conventional assumptions about technology power.

The United States designs the most advanced chips (Nvidia, AMD, Apple). Taiwan manufactures them (TSMC). But the enabling technology — the equipment that makes manufacturing possible — comes from a small country of 17 million people that has no native chip industry and no military capable of projecting power beyond Europe.

This creates a strange dependency structure. America can design chips it cannot manufacture. Taiwan can manufacture chips whose equipment it cannot build. China can build equipment whose precision it cannot match. And the Netherlands — through a single company in a single city — holds the key that makes the entire system work.

The strategic implications are still being processed:

For the US: The export control regime works only as long as the Netherlands cooperates. Dutch strategic autonomy has a price — ASML has lost Chinese revenue, and the Dutch government faces pressure from a company that employs 42,000 people and generates €28 billion in annual revenue. American policymakers cannot take Dutch alignment for granted.

For China: The gap is not closing. Despite massive investment in domestic semiconductor equipment, Chinese lithography remains 15-20 years behind. The path to self-sufficiency in leading-edge chips does not exist on any reasonable timeline. China must either accept permanent disadvantage in AI and advanced computing — or find a way to change the structure of the competition.

For Taiwan: TSMC's dominance depends on continued access to ASML equipment. A Chinese blockade of Taiwan would not just disrupt chip manufacturing — it would disrupt chip equipment supply, since ASML service engineers and spare parts flow through Taiwanese logistics. The chokepoint creates vulnerability for everyone who depends on it.

In the end, the ASML chokepoint is not about politics or policy. It is about physics.

EUV light cannot be generated without tin-droplet lasers. Tin-droplet lasers cannot work without Cymer's engineering. The light cannot be focused without Zeiss's mirrors. The mirrors cannot be polished without twenty years of accumulated process knowledge. The knowledge cannot be transferred without the people who hold it.

This is not a supply chain that can be disrupted by tariffs, subsidized by industrial policy, or replicated by throwing money at the problem. It is a supply chain built on physics that no one else understands well enough to reproduce.

The United States discovered this in the 1980s when it lost semiconductor manufacturing to Asia. China is discovering it now. The lesson is the same: you cannot buy what can only be built. And some things can only be built in Veldhoven.

The future of computing — AI, quantum, whatever comes next — will run on chips made by machines that come from one company, in one country, using technology that exists nowhere else on Earth.

That is the chokepoint. It is not opening anytime soon.