The Impossible Machine That Runs the Modern World
There's a machine sitting in a handful of chip factories around the world that most people have never heard of. It costs north of $350 million. It weighs as much as a commercial aircraft. It gets shipped in 250 containers across 25 trucks and seven Boeing 747s. And without it, the phone in your pocket — and pretty much every other advanced piece of technology you own — simply wouldn't exist.
It's called an EUV lithography machine. And for about thirty years, the smartest people in the semiconductor industry thought it was impossible to build.
To understand why this machine matters, you need to know a little about what's inside a microchip. When you zoom into one — really zoom in — you find something that looks like a city. Skyscrapers of circuitry hundreds of layers tall, hundreds of kilometers of wiring connecting billions of tiny switches called transistors. These transistors are the ones and zeros of computing. They work by shuttling electrons back and forth, and the smaller you make them, the faster and more powerful the chip.
For decades, this worked out beautifully. The number of transistors you could fit on a chip doubled roughly every two years — a pattern so reliable it became known as Moore's Law, named for Intel co-founder Gordon Moore who noticed it back in 1965. It was the engine behind fifty years of technological progress. And then, around 2015, it started to stall.
The problem was light.
To etch transistors onto a chip, manufacturers use a process called photolithography — essentially, they shine light through a patterned mask to imprint a circuit design onto a silicon wafer. The finer the features you want to print, the shorter the wavelength of light you need. By the late 1990s, the industry had settled on 193-nanometer deep ultraviolet light, and for years it worked well. But eventually, they'd squeezed everything they could out of it. The features were getting so small that the physics of light itself was getting in the way.
Something radical had to change.
The seeds of the solution were planted back in the 1980s, in a lab in Japan, by a scientist named Hiroki Kinoshita. His idea was straightforward in theory and borderline insane in practice: why not use much shorter wavelengths — closer to x-ray territory, around 13 nanometers — to print far smaller features?
The problem was that light at these wavelengths doesn't behave like ordinary light. It gets absorbed by almost everything — glass, air, you name it. You can't use lenses to focus it. You'd have to work in a vacuum. And to reflect it off mirrors, those mirrors would have to be smoother than anything that had ever been made.
When Kinoshita presented his early results at a 1986 physics conference, the audience wasn't impressed. They were skeptical to the point of dismissive.
"People seemed unwilling to believe that we had actually made an image by bending x-rays,"
he later recalled, saying they regarded the whole thing as "a big fish story."
Across the Pacific, a researcher named Andy Havrilak was running into the same wall. Working at Lawrence Livermore National Lab — a place originally built for nuclear weapons research — he'd gotten excited about the same idea. When he presented his findings at a conference, the response was brutal.
"I was literally laughed off the stage,"
he said. Every person he respected in the field came up to the microphone to tell him why it wouldn't work. He flew home, and when his boss asked how it went, he told him: "I will never speak of it again."
Three days later, he got a call from Bell Labs. The idea wasn't dead yet.
By the early 2000s, a fragile coalition had formed around what was now being called Extreme Ultraviolet Lithography, or EUV. The US government had funded early research through its national labs. Intel, Motorola, AMD, and others eventually pooled $250 million to keep it alive. And one company in particular had decided to bet everything on it: ASML, a Dutch firm that had spun off from Philips in the 1980s with little more than a shed and a barely working wafer machine to its name.
ASML's task was to figure out how to actually build a commercial EUV machine. And to understand why that was so hard, consider just one of the problems they had to solve: the light source.
EUV light doesn't occur naturally on Earth. The closest thing is the sun. So ASML had to build an artificial one inside a factory. Their solution was to fire a 20,000-watt carbon dioxide laser — four times more powerful than lasers used to cut through steel — at tiny droplets of molten tin, each one roughly the size of a white blood cell, falling at around 250 kilometers per hour.
Each droplet gets hit not once, not twice, but three times. The first pulse flattens the droplet into a pancake shape. The second spreads it into a low-density gas. The third obliterates it entirely, heating it to over 220,000 degrees Celsius — roughly forty times hotter than the surface of the sun — and producing a burst of EUV light. This happens 50,000 times every second.
"We don't miss them,"
one of the engineers said, when asked how often the laser misses a droplet. Out of 150,000 shots per second, they miss none.
The tin droplets presented another problem: where does all that vaporized metal go? The collector mirror — one of the most precisely polished objects ever made — sits just 30 centimeters away. A single nanometer of tin contamination on it and the whole thing is out of commission. ASML's solution was to fill the chamber with low-pressure hydrogen, which slows the tin debris and reacts with it to form a gas that can be flushed out. But the hydrogen itself gets heated by all those tiny explosions. The team eventually realized they needed to flush it through at around 360 kilometers per hour — faster than a Category 5 hurricane.
To understand how the energy was distributing through the chamber, one of the engineers borrowed a formula typically used to model nuclear blasts and supernovae.
"It's just fantastic that we're seeing these little tiny supernovae happen in our vessel 50,000 times a second,"
he said.
The mirrors are their own story entirely. Because EUV light gets absorbed by almost any material, you can't use conventional glass optics. Instead, ASML worked with their German partner Zeiss to develop mirrors built from hundreds of alternating layers of silicon and molybdenum, each layer thinner than a single nanometer. The result can reflect about 70% of EUV light — close to the theoretical maximum.
But the reflectivity only holds if the mirror surface is almost perfectly flat. We're talking about a smoothness that's genuinely hard to picture. If you scaled one of the low-NA machine's mirrors up to the size of Germany, the tallest surface imperfection would be about a millimeter high. The newer high-NA mirrors are smoother still — scaled up to the size of the Earth, the biggest bump would be no thicker than a playing card.
And then there's the alignment. When you're printing a chip with dozens of layers, each layer has to land almost exactly on top of the previous one. The maximum allowable misalignment — what engineers call the overlay — is one nanometer. That's five silicon atoms of precision, in a machine where parts are whipping back and forth at accelerations twenty times that of gravity.
"You get a nanometer in total,"
one systems engineer explained. "So you have to fight for your part of the nanometer."
None of this came easily, or quickly. ASML spent years promising customers breakthroughs that were perpetually around the corner. They were, as one of their executives put it, "crucified at every conference." Customers who'd been waiting for a working machine were running out of patience, and the goalposts kept moving — just as ASML hit 100 watts of source power, the industry realized they'd need 200. Just as they hit 200, something else needed fixing.
There were moments of genuine desperation. In 2012 or 2013 — with the power numbers still stubbornly low — one of the executives took Kinoshita to dinner at a restaurant near ASML's headquarters. Across the street was a small chapel.
"We have come to the limits of science,"
he recalled thinking. "Let's go for divine intervention." They went in. Kinoshita lit three candles — one for each company still working on EUV technology.
"Lo and behold, and I have the data to prove it, there is a very strong correlation between us lighting the candle and power going up. It's not a causal effect, but there is a strong correlation."
The real breakthrough came from a deceptively simple insight: instead of hitting each tin droplet once with the main laser, hit it first with a weaker pre-pulse to flatten it into a pancake shape, then hit the pancake. The larger surface area meant more of the tin got vaporized all at once, producing far more EUV light with far less debris. By 2014, they finally hit the 100-watt mark they'd been chasing for years.
And one day in 2015, with customers threatening to walk if they didn't see 200 watts immediately, the ASML board flew to Korea for a make-or-break demonstration. When they boarded the plane in the Netherlands, the experiment was still running. When they landed in Seoul, they had their answer.
By 2016, orders were coming in. Today, every advanced chip in the world — the processors in your laptop, the neural engines in AI hardware, the chips in your phone — requires an ASML EUV machine somewhere in its production process. It's hard to overstate how much of the modern world runs through this one company in a small Dutch city.
Their newest high-NA machines, at $350 million-plus each, can print features just 8 nanometers wide — smaller than most viruses. They ship disassembled, packed into 250 containers across twenty-five trucks and seven 747s, then rebuilt in cleanrooms so pure that a single cubic meter can contain no more than ten particles of dust. Hospital operating rooms, by comparison, allow up to 10,000.
It took thirty years. It took people getting laughed off stages, funding getting cut, candles getting lit. It took engineers modeling their equipment on supernova physics and fighting each other over fractions of a nanometer. It took a company in the Netherlands deciding to go all-in on a technology that most of the industry had given up on.
George Bernard Shaw once wrote that "the reasonable man adapts himself to the world; the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man."
Kinoshita was unreasonable. Havrilak was unreasonable. The people at ASML who kept pouring money into a machine that didn't work yet were unreasonable. And because of that, you're reading this on a device that shouldn't exist.
Maybe being a little unreasonable isn't such a bad thing after all.
The article is based on the video “The World's Most Important Machine”
Transcribed into text using Speech2Text 📱 https://speech2text.pro/
