The Most Important Machine You’ve Never Heard Of

Few people know the name ASML, and fewer still could explain what the company makes. This anonymity is surprising, because the modern world rests on a machine that only ASML can build. Every smartphone, aircraft navigation computer, data-center server and artificial intelligence model depends, directly or indirectly, on a process that occurs inside a sealed steel chamber in a business park in the southern Netherlands.

The machine in question is an extreme ultraviolet (EUV) lithography system. Its purpose is straightforward to describe: it projects patterns onto silicon wafers so that transistors can be manufactured at ever smaller scales. Its execution is anything but simple. EUV lithography relies on a wavelength of light—13.5 nanometers—that does not exist naturally on Earth, cannot travel through air, cannot pass through glass and will only reflect from mirrors whose surfaces are polished to the scale of atoms. It is not merely an industrial tool; it is a negotiation with the laws of physics.

The survival of Moore’s Law, the doubling of transistor density roughly every two years, now depends entirely on ASML’s ability to manufacture and ship these machines. Were they to stop, the shrinkage of transistors would stop with them. Software could still evolve, but computational progress would slow to the pace permitted by fixed hardware. Behind every celebrated leap in artificial intelligence therefore lies an uncelebrated one in optics, plasma physics and precision engineering.

What the Machine Does (in human terms)

For decades, the semiconductor industry shrank transistors by shrinking the wavelength of light used to pattern them. First visible light, then ultraviolet, then deep ultraviolet were pressed into service. Eventually the industry reached a limit: even the shortest conventional ultraviolet wavelengths could not produce features much smaller than 20 nanometers. At that point, further improvement required a smaller wavelength. EUV, at 13.5 nanometers, was the only candidate.

At such dimensions, the scale becomes difficult to visualize. A transistor made with EUV lithography can be smaller than a single virus particle. Tens of billions of them can be printed onto a piece of silicon smaller than a postage stamp. The energy efficiency of modern smartphones, the training speed of neural networks and the battery life of electric vehicles all derive not only from cleverness in software, but from the quiet fact that transistors can now be printed at absurdly fine resolutions.

EUV lithography is the enabling mechanism: it uses extreme ultraviolet light to etch these patterns. It is not a clever refinement of existing methods; it is a change in kind. Without EUV, chip scaling had reached its physical limit. With EUV, new generations of microprocessors became possible. The relationship is brutally direct: no EUV => no 7 nm chips; no 7 nm chips => no modern AI.

The Three “Impossible” Hurdles

If the objective—printing smaller transistors—was conceptually simple, the obstacles to achieving it were not. ASML and its partners spent nearly two decades solving three problems that had previously defeated every attempt at extreme ultraviolet lithography. Each problem was regarded by much of the industry as not merely technically difficult, but physically prohibitive. They concerned the generation of EUV light, its transmission, and its application to silicon with tolerances approaching the atomic.

The first hurdle was generating EUV light itself. No conventional light source can produce a wavelength of 13.5 nanometers. Lasers cannot reach it; lamps cannot create it; LEDs cannot be engineered to emit it. EUV does not occur naturally in Earth’s atmosphere. Producing it requires the controlled creation of a plasma—a cloud of ionized atoms—hotter than the surface of the sun, sustained and repeated tens of thousands of times per second, with unwavering precision.

The second hurdle was transporting that light once it existed. EUV photons cannot travel through air: they are absorbed almost immediately by nitrogen and oxygen molecules. Glass is equally hopeless, as it absorbs EUV entirely. These two facts eliminate lenses, fibre optics and any form of open path. The light must therefore travel through a vacuum, guided only by mirrors, which themselves must be manufactured to tolerances so strict that the usual vocabulary of precision becomes inadequate.

The third hurdle was applying the EUV light to a silicon wafer with absolute stability. When features measured in single-digit nanometers are being printed, any vibration—whether originating from motors, cooling systems, or even machinery in adjacent rooms—is large enough to blur the pattern. Motion systems must not only position a wafer to nanometer accuracy, but do so repeatedly, at high speed, without mechanical contact and without transmitting vibration.

Solving any one of these hurdles would have justified a doctoral thesis. Solving all three simultaneously required a sustained collective commitment from an ecosystem rather than a single firm. ASML supplied the integration; Zeiss the mirrors; Trumpf and Cymer the lasers; specialized suppliers in Japan the motion systems. The EUV machine is less a product than an achievement: a collection of dependent miracles operating in sequence.

How ASML Solved the Impossible Hurdles

Hurdle 1: Generating EUV light

Producing extreme ultraviolet light is an exercise in coercing matter to behave in a manner it does not naturally prefer. Because no conventional light source emits photons at 13.5 nanometers, ASML had to invent one. The solution involves firing a high-power laser—thirty kilowatts, comparable to the electrical load of a small industrial workshop—at droplets of molten tin. Each droplet is about the width of a human hair and is expelled in a continuous stream through a vacuum chamber.

The laser strikes each droplet twice. The first pulse deforms the droplet into a thin disk, which increases the surface area. The second pulse vaporizes the metal into a plasma. In the few billionths of a second that follow, electrons are stripped from the tin atoms and then recombine. As they shed energy, they emit photons in the extreme ultraviolet range. The machine repeats this procedure fifty thousand times per second. It is, in effect, a controllable micro-explosion factory.

It is difficult to overstate the coordination involved. The droplets are moving at roughly 200 miles per hour. The laser must hit each one within a window of a few tens of nanoseconds. If the laser fires too early, the droplet is merely dented; too late, and it has already fallen out of position. The entire system resembles a highly choreographed industrial ballet whose dancers are molten metal flying through vacuum.

Hurdle 2: Controlling and transporting EUV

Generating EUV is only the beginning. Once created, the light must reach the silicon wafer without being absorbed. Unfortunately, EUV interacts enthusiastically with the universe. A millimeter of air is enough to extinguish it. Glass, quartz and every transparent material used in conventional optics swallow it completely. Lenses are therefore useless. Any optical path containing air, glass or dust is equally useless.

ASML’s answer was to create an internal vacuum environment in which the EUV beam travels only by reflection. Mirrors—rather than lenses—guide the beam. But even mirrors are barely adequate. At this wavelength, a mirror does not behave like a reflective surface in the everyday sense. Instead, its reflectivity depends on constructive interference within a stack of precisely deposited thin films. Zeiss, ASML’s optical partner, manufactures each mirror by layering alternating sheets of molybdenum and silicon, each layer deposited with atomic-scale control.

Even then, the mirror reflects only about seventy percent of the light that reaches it. After multiple reflections, the beam would vanish were the mirrors anything less than nearly perfect. To achieve the required performance, the surface of each mirror must deviate from flatness by no more than roughly fifty picometers. For a sense of perspective, if one of these mirrors were scaled to the size of a football field, its tallest imperfection would be thinner than a sheet of paper.

Hurdle 3: Printing on silicon with picometer stability

The final challenge is to apply the EUV light to a silicon wafer with absolute positional accuracy. A transistor at the five-nanometer scale is twenty atoms wide. At that resolution, a vibration from a cooling pump, a truck passing on the street outside, or even a minor thermal expansion within the machine would be large enough to create a defective pattern.

ASML’s solution is an act of mechanical restraint. The wafer sits on a magnetically levitated stage that moves without friction. Positioning systems measure its location in real time with interferometers—the same kind of laser-based measurement instruments used in gravitational wave detectors. Every motion occurs in increments of nanometers. Every motion must begin and end without transmitting vibration.

The effect is that the most violent part of the system—the high-power laser creating miniature bursts of plasma—coexists with one of the most delicate: the placement of a wafer with sub-nanometer accuracy. The two are separated by design, control theory and the elimination of mechanical disturbances. The machine is simultaneously explosive and serene.

Why Only ASML Can Do This

Technological lock-in

The first reason ASML has no rival is that the company continued pursuing EUV long after the rest of the industry abandoned the effort. In the early 2000s, Canon and Nikon—then leaders in conventional lithography—invested heavily in EUV research. After several years and billions in spending, they concluded that the physics was intractable. Light that could not travel through air, could not be shaped by lenses, and could only be reflected by mirrors manufactured to atomic tolerances seemed, quite reasonably, a dead end.

ASML reached the same impasses, but made a different decision: it refused to quit. For more than a decade the company spent money on a machine that did not work, based on a bet that it eventually would. When EUV finally became viable, ASML held not only the intellectual property, but the world’s only accumulated experience in integrating EUV subsystems into a functioning product. Its competitors had intact technical capabilities, but none had the institutional memory necessary to assemble the pieces into a whole.

In technology, persistence is often as decisive as brilliance.

A monopoly built from suppliers

The second reason ASML is alone at the summit is that no single company controls the technologies required to build an EUV system. Each machine is a network of thousands of components sourced from more than 5,000 suppliers. The optical mirrors—arguably the most demanding element of the entire system—are designed and manufactured exclusively by Zeiss in Germany. The high-power laser system originates from Cymer in the United States, using industrial lasers from Trumpf, a German company. The ultra-precise positioning systems are produced by specialized firms in Japan.

ASML does not “own” these technologies, but it is the only firm that knows how to make them behave together. The moat is not a patent portfolio; it is an ecosystem. To reproduce EUV capability, a competitor would have to replicate not only the technologies, but the collaborative infrastructure behind them. Supply chains can be copied in theory; trust and accumulated coordination cannot.

The result is a monopoly that is not legally enforced, but practically inevitable.

Tacit knowledge

The third reason lies in a category that does not appear on balance sheets: accumulated tacit knowledge. EUV lithography works not only because the right components exist, but because tens of thousands of small decisions about how those components interact were made correctly. There is no manual that explains how to align mirrors that reflect only seventy percent of the photons that strike them, or how to damp the vibrations of a magnetically levitated wafer stage while plasma events occur a meter away. Those practices live in the experience of engineers who have solved those problems before.

This type of knowledge cannot be purchased and cannot be reverse-engineered. It is earned—expensively—by making mistakes and remembering the remedies. ASML is the custodian of that memory. Canon and Nikon would now need not years, but decades, to catch up. China is investing enormous sums in pursuit of an indigenous EUV capability, but money cannot accelerate time.

Where technical capability meets institutional memory, monopoly follows.

Geopolitics and Fragility

The fact that a single company produces the world’s only EUV lithography systems would be notable in any industry. In semiconductors, it is extraordinary. The machines are not sold in volume; they are allocated. Each shipment reshapes competitive dynamics among chipmakers. TSMC in Taiwan, Samsung in South Korea and Intel in the United States receive them. China does not. The United States government has made certain of that.

EUV machines now fall under export-control rules normally reserved for advanced weapons. Washington has pressured the Dutch government to block sales of ASML’s most advanced systems to Chinese chipmakers. The logic is brutally simple: the country that controls the ability to manufacture cutting-edge chips controls artificial intelligence, advanced weaponry and the future productivity of its industrial base. In earlier eras, nations competed for access to oil reserves or strategic waterways. Today, the strategic bottleneck is a factory in Veldhoven that produces machines capable of vaporizing tin in a vacuum.

This fragility is geographic as well as political. Most of the world’s advanced chips are manufactured in Taiwan, a location that sits uncomfortably close to geopolitical fault lines. Were access to ASML’s machines disrupted—by export restrictions, conflict or internal failure—progress in computing would slow globally within a year. The digital world appears immaterial, but its foundations are profoundly physical. It depends on cargo aircraft transporting boxes weighing 180 metric tons, on mirrors flatter than atoms, and on the diplomatic alignment of a handful of governments.

The idea that the trajectory of computing power hinges on a supply chain involving German optical specialists, Japanese precision-motion firms, American laser makers and a Dutch systems integrator sounds improbable. Yet here we are. The prosperity of the information age rests on a machine that requires the cooperation of countries that often struggle to agree on anything else.

Conclusion

It is easy to treat digital progress as an abstraction: lines of code, cloud platforms, or algorithms behaving intelligently. But the arc of computing advancement is constrained not by cleverness alone, but by physics. Software eats the world only because the world first manufactures ever-smaller transistors, and those transistors exist only because a machine can etch features onto silicon with extreme ultraviolet light.

That capability did not emerge from a single breakthrough. It emerged because a company persisted for two decades in pursuit of a goal that appeared physically absurd. ASML did not out-innovate its rivals so much as out-endure them. In doing so, it built not merely a product, but a lever that moves the global economy. Every improvement in artificial intelligence, every reduction in battery consumption, every gain in computational efficiency traces back to a moment when a droplet of molten tin met a precisely timed laser pulse inside a vacuum chamber.

Modern technology depends on many abstractions. EUV lithography is not one of them. It is a reminder that even the most ethereal software relies on the most concrete hardware, and that hardware, in turn, relies on photons behaving as quantum mechanics dictates. Somewhere in the Netherlands, mirrors and lasers and droplets of tin determine how far the digital world can advance.

Everything else is downstream.

This work is licensed under a Creative Commons Attribution 4.0 International License. CC BY 4.0
Feel free to share, adapt, and build upon it — just credit appropriately.