There is a company in the Netherlands whose machines are so complex, so singular, and so indispensable that without them, the machinery of the modern world would come to a halt. ASML, the Veldhoven-based semiconductor equipment maker, produces the photolithography systems used to print microscopic circuitry onto silicon wafers — the chips inside nearly every smartphone, server, electric vehicle, and AI accelerator. As of 2025, it holds 83% of the worldwide market for lithography machines. In the most advanced category, extreme ultraviolet lithography, its market share approaches 100%.
This concentration of capability in a single company, in a single country, is one of the most consequential and least understood forces in the global economy — and it is becoming harder to manage with every passing year.
An Impossible Machine
Modern chipmaking is, at its core, a printing process. A machine projects light through a finely etched stencil and onto a slice of silicon, leaving behind a pattern of circuitry at scales far too small to see with the naked eye. This is done dozens of times per chip, building up layer after layer until the nanochip is capable of executing billions of operations per second. The smaller the features you can print, the more powerful the chip, and the shorter the wavelength of light needed to achieve it.
ASML’s most advanced machines, its EUV systems, generate light with a wavelength of just 13.5 nanometres — roughly ten thousand times shorter than visible light — by firing a laser at microscopic droplets of molten tin 50,000 times per second, vaporising them into plasma hotter than the surface of the sun. Because light at this wavelength is absorbed by almost every material, including air, the entire optical path must operate in a hard vacuum, with mirrors doing all the work. Those mirrors are made exclusively by Carl Zeiss SMT in Germany, each set taking months to produce, with no alternative supplier anywhere in the world. ASML’s current flagship costs up to $200 million, whilst its next-generation system, which enables the sub-2nm chips, runs to $370-400 million.
Why there is no Second Place
The ensuing question is why no one else builds these machines. Nikon and Canon both attempted to develop competing EUV systems and eventually walked away, as they fell too far behind to catch up. Each ASML machine is the cumulative output of two decades of co-development with over 5,000 subcontractors across multiple continents, many of whom have spent decades co-developing components to precision that cannot simply be replicated from scratch. The software alone, which coordinates optics, wafer positioning, and patterning logic simultaneously represents hundreds of years of accumulated engineering. Most critically of all, chip fabrication plants run continuously, and a day of lost output at Taiwan Semiconductor Manufacturing Company (TSMC), the company which buys those machines to manufacture the world’s most advanced computer chips, costs hundreds of millions of dollars. As a result, a self-enforcing loop is perpetuated: a new entrant would ship too few machines to accumulate operating data, and without data, reliability can’t be matched; without reliability, chipmakers won’t adopt; without adoption, there’s no data.
This is the loop that makes ASML’s position so formidable. It is also what makes it so dangerous. A supply chain this concentrated becomes a point of systemic fragility for the entire global economy. A single High Numerical Aperture EUV machine, which uses an optical system that gathers and focuses light at much wider angles, draws on optical components from Germany, light sources from San Diego, wafers from Japan, and final chip production in Taiwan, assembled in the Netherlands, serviceable only by ASML engineers. A Taiwan Strait incident, a natural disaster in Japan, a policy disagreement between The Hague and Washington: any of it reverberates through the entire system. The 2020-2023 chip shortage cost the automotive industry alone an estimated $200 billion, and that was without any of the major nodes actually failing. ASML’s monopoly, in other words, is not just ASML’s problem to manage.
The Espionage Problem
Predictably, the monopoly has been systematically targeted. In 2015, a former employee at ASML’s Silicon Valley subsidiary stole proprietary source code and trade secrets before moving to a Chinese competitor; a 2023 investigation by Dutch newspaper NRC Handelsblad reported they subsequently went to work for Huawei. That same year, ASML alleged a second former worker in China had stolen technology information. The pattern is consistent: where building a rival machine through legitimate means has proved impossible, the knowledge embedded in the people who built the original has become the target instead.
In December 2025, Reuters reported what may be the culmination of those threads: a prototype EUV machine, built in a high-security Shenzhen laboratory by a team largely comprising former ASML engineers, using components from second-hand machines acquired on the secondary market for used semiconductor equipment. The machine reportedly succeeds in generating EUV light, but has not yet demonstrated the ability to produce functional chips. The Dutch government has responded by ruling that ASML may reject job applicants from sanctioned countries: an awkward policy for an industry built on cross-border collaboration, and one that addresses symptoms rather than causes.
China’s Long Game
If ASML’s monopoly has a credible long-term challenger, it is not a company. It is a state that has decided production of equipment for nanochips is a priority.
Since 2018, the US has run an escalating export control campaign designed to keep ASML’s most advanced machines out of China, banning EUV exports in 2019 and progressively tightening restrictions on older technology in 2023, 2024, and again in early 2025, as part of an emerging technological cold war. The underlying logic is containment: without access to the machines that print the most advanced chips, China’s semiconductor industry would be frozen at a level of capability that posed no serious competitive threat.
In practice, the strategy has worked less cleanly than its architects had hoped. Denied access to EUV, Chinese chipmakers turned to what they had: older DUV machines and a technique called multi-patterning, stacking multiple slightly offset exposures to achieve feature sizes no single process cycle could reach. Semiconductor Manufacturing International Corporation (SMIC), China’s largest and most advanced contract chipmaker, produced 7nm-class chips in 2022 this way. Each additional exposure adds cost and reduces the proportion of chips that come out working correctly, making it roughly 40-50% more expensive per wafer and significantly less reliable. But when the state is absorbing the difference as a matter of national security, the normal economics of chip production stop applying.
Meanwhile, China’s chipmakers responded to the US export controls by stockpiling aggressively ahead of each successive round of restrictions, reasoning correctly that machines already delivered would be harder to caw back than export licences yet to be issued, driving China’s share of ASML’s revenue from 26% in 2023 to 36% in 2024. By 2024, roughly 70% of ASML’s global shipments of its older deep ultraviolet machines — the generation below EUV — were going to Chinese customers, before restrictions began to bite more seriously. That existing installed base is now large enough, by most credible estimates, for Chinese fabrication plants to hit near-term domestic targets for AI chip production without any further Western equipment at all.
The performance gap between China and ASML remains large and structural. Yet, what matters is the rate at which it is closing, and whether China can compress a decade of development into five years as it has already done, to general astonishment, in 5G infrastructure, solar panels, and electric vehicle batteries.
The Challengers
ASML also faces a different kind of pressure from companies that believe the entire physical approach underpinning EUV is mature enough to be disrupted.
Substrate, a San Francisco company founded in 2022 and backed by Peter Thiel’s Founders Fund at a valuation above $1 billion, is developing X-ray lithography using particle accelerators as a light source, rather than extreme ultraviolet light. As X-ray wavelengths are shorter than EUV’s 13.5nm, the technology could in principle print even finer features without the plasma generation and vacuum mirror complexity that makes EUV so hard to replicate. Substrate claims to have demonstrated 2nm-equivalent resolution at US national laboratories and aims to cut leading-edge wafer costs from a projected $100,000 to around $10,000 by 2028. Whether that ambition reflects the kind of naivety that ends in failure or the kind that occasionally breaks through is genuinely unclear; the semiconductor industry has seen both in abundance.
xLight, led by former Intel CEO Pat Gelsinger, is narrower and more pragmatic: not a full scanner replacement, but a free-electron laser light source that could be integrated into existing ASML machines as an upgrade, targeting wavelengths as short as 2nm with better power efficiency than ASML’s plasma approach. The Trump administration committed up to $150 million in CHIPS Act function to xLight in 2025, making the US government its largest shareholder. If xLight’s lasers integrate into ASML’s architecture, it becomes an upgrade path that ASML may eventually have to adopt on someone else’s terms. If it operates independently, it would begin to erode ASML’s pricing power at the most technically constrained point in the system.
Neither company will threaten ASML’s position within this decade. But both represent something structurally important: the global technology establishment has decided that ASML’s monopoly is a problem serious enough to solve through physics, and in xLight’s case the US government is funding the attempt directly. This is a policy signal about how Washington reads the strategic risk of a single Dutch company sitting at the centre of the world’s most consequential supply chain, and a sign that America’s relationship with that monopoly is considerably more complicated than straightforward allied solidarity.
Forever is the Wrong Question
The near-term case for ASML borders on unassailable. The AI boom is generating structural demand for EUV machines that is not going away, and the High-NA roadmap that underpins the next generation of chips has no commercial competitor anywhere in the world. The moat — 16,000 patents, a 5,000-node supply chain, and decades of irreplaceable field data accumulated across thousands of machines running continuously in the world’s most advanced fabs — is as formidable as any in global industry.
But the medium and long term require sitting with a harder truth: no industrial monopoly sustained purely by technical complexity has survived indefinitely when a major state has decided that breaking it is a national priority and committed the resources to do it. The Soviet Union built nuclear weapons from a position of total initial disadvantage and achieved parity in four years. The Shenzhen prototype is not a commercial threat to ASML this decade, but it is evidence that China has cleared the hardest technical hurdle, generating EUV light, from a standing start, using a programme that has near-unlimited political backing and a consistent track record of being underestimated by the West.
What makes ASML’s position genuinely precarious is not that any single one of these threats is decisive on its own. It is the simultaneity: a state adversary with effectively unlimited resources and a clear strategic mandate running a decades-long effort; a geopolitical ally in the United States that is simultaneously funding alternative physical approaches and pushing export controls that may be accelerating the very self-sufficiency they are designed to prevent; and a technology horizon on which the constraints that make EUV so hard to replicate may eventually become irrelevant because the underlying physics has changed around them.
Each new generation of chips that ASML enables buys more time for the existing order. Each one also builds more of the infrastructure — the fabs, the engineers, the supply chains — that would eventually allow a credible challenger to step in. ASML will be indispensable for this decade, almost certainly. Whether it remains the world’s only viable maker of the machines that define the technological frontier in 2040 depends on variables that its engineers, its CEO, and its 16,000 patents cannot fully control.
The most dangerous chokepoints in the global economy are not the ones under obvious threat. They are the ones that feel permanent — until, with very little warning, they don’t.
Leave a comment