The $400M Machine That Saved Computing
Moore's Law hit a wall around 2015 when transistors couldn't get smaller using traditional lithography. ASML's extreme ultraviolet (EUV) lithography machine—the most complex commercial product ever built—solved this by using a laser to hit 50,000 tin droplets per second, creating plasma hotter than the Sun's surface. The machine required 30 years of development, mirrors smoother than anything in the universe, and precision down to five silicon atoms, making it the linchpin of modern chip manufacturing.
The Problem: Moore's Law Hits a Wall
Transistor Scaling Breakthrough
For over 50 years, transistors got smaller and smaller, doubling in number on a chip every two years—a pattern called Moore's Law. Smaller transistors mean faster signals and more computing power per unit area. Around 2015, this progress stopped because traditional photolithography using 193-nanometer ultraviolet light couldn't print features any smaller.
The Physics Barrier
Photolithography uses light to etch patterns onto silicon. As features get smaller, they approach the wavelength of the light itself, causing diffraction—light bends and spreads instead of focusing sharply. The Rayleigh Equation shows that smaller features require either shorter wavelengths or larger lenses, but both have hard physical limits.
The Solution: Extreme Ultraviolet Lithography
The 30-Year Journey Begins
In the 1980s, Japanese scientist Hiroo Kinoshita proposed using extreme ultraviolet (EUV) light at 13 nanometers instead of 193 nanometers. This shorter wavelength could print much smaller features, but X-rays at this wavelength are absorbed by air and most materials, requiring a vacuum chamber and special multilayer mirrors instead of conventional lenses.
Multilayer Mirrors: The Breakthrough
Underwood and Barbee discovered that alternating ultra-thin layers of tungsten and carbon (76 layers total, each less than 1 nanometer thick) could reflect X-rays through constructive interference. When X-rays pass through one layer, some reflect; the phase-shifted reflected wave from the next layer adds constructively, amplifying reflection. This principle became the foundation for EUV optics.
The Impossible Machine: ASML's EUV Lithography System
Specifications That Defy Belief
ASML's EUV machine costs over $400 million and performs feats of precision that seemed impossible for decades. It hits 50,000 tin droplets per second with a laser, heating each to 220,000 Kelvin (40 times hotter than the Sun's surface). The machine never misses a shot and maintains overlay accuracy of one nanometer—five silicon atoms—between chip layers.
The Smoothest Mirrors in the Universe
The collector mirror in ASML's machine is so smooth that if scaled to the size of Earth, the largest bump would be thinner than a playing card. The high NA machine's projection mirrors are even smoother: if scaled to Earth size, the tallest bump would be about the thickness of a playing card. This extreme smoothness prevents light scattering and ensures all photons reach the wafer.
The Laser-Produced Plasma Source
ASML generates EUV light by firing a 20,000-watt laser at tin droplets, creating a plasma hotter than the Sun. Tin was chosen over xenon because it has a higher emission peak at 13.5 nanometers, yielding 5–10 times better conversion efficiency. The machine uses three laser pulses per droplet: a pre-pulse flattens it into a pancake, a second pre-pulse rarefies it into low-density gas, and the main pulse ionizes it all at once.
Tin Droplet Chaos and Control
Tin is melted and pushed through a microscopic nozzle vibrating at high frequency, initially creating irregular droplets of varying size, shape, and velocity. Before reaching the laser, these chaotic droplets mysteriously self-organize into perfectly spaced, identical droplets traveling at the same velocity. Engineers modulate nozzle pressure and vibration frequency to force this behavior—described as 'magic sauce' by the team.
Hydrogen Gas Cleaning System
Tin plasma debris threatens to coat the expensive Zeiss mirrors. ASML fills the chamber with low-pressure hydrogen gas that slows and cools tin particles. If debris reaches the mirror, hydrogen pulls it off as stannane gas, self-cleaning the optics. However, too little hydrogen allows mirror contamination; too much absorbs EUV light and causes overheating. The team discovered that shockwaves from each plasma event follow the Taylor-von Neumann-Sedov formula (used to model supernovas), allowing them to calculate the exact hydrogen flow rate needed: 360 kilometers per hour—faster than a Category 5 hurricane.
Mirror Alignment: Pico-Radian Precision
Heat from the plasma slightly shifts mirrors, misaligning the light. Zeiss built a 'nervous system' into the optics: robot-guided sensors constantly measure each mirror's position and angle down to the nanometer and pico-radian (angular precision so extreme that a laser pointed from Earth to the Moon could distinguish which side of a dime it hits).
Reticle Motion and Overlay Accuracy
To print ~185 wafers per hour, the reticle (mask) whips back and forth at accelerations exceeding 20 Gs—over five times a Formula 1 car's acceleration. Despite this violent motion, the machine maintains overlay accuracy of one nanometer between layers, meaning any two chip layers can be misaligned by no more than five silicon atoms. Engineers divide this one-nanometer budget among all subsystems; each component 'fights for its fraction of the nanometer.'
The Path to Production: Overcoming Skepticism and Setbacks
Early Rejection and Ridicule
When Kinoshita presented his EUV findings in 1986, the audience refused to believe him, calling it a 'big fish story.' Similarly, when Andy Hawryluk proposed applying multilayer mirrors to lithography in 1987, he was laughed off the stage at a conference; every expert told him the idea was stupid. Both scientists faced career-defining rejection but persisted.
Government and Industry Funding
After the US government cut EUV funding in 1996, Intel, Motorola, AMD, and others invested $250 million to keep development alive—the largest private investment ever in a Department of Energy research project. Later, Intel invested $4.1 billion and Samsung and TSMC combined invested $1.3 billion to fund ASML's next-generation high NA machine, even though no working product existed.
The Power Scaling Challenge
The Engineering Test Stand prototype (2000) produced only 9.8 watts and printed just 10 wafers per hour. To be economically viable, the machine needed 100+ watts and 100+ wafers per hour. Scaling laser-produced plasma proved far harder than expected. By 2013, ASML reached only 50 watts. The industry kept moving the goalposts: as soon as 100 watts was achieved in 2014, the target jumped to 200 watts.
The Pancake Breakthrough
A major turning point came when engineers realized that hitting a droplet once created too much debris and reabsorption. They switched to a two-pulse strategy: the first pulse flattens the droplet into a pancake, increasing surface area; the second pulse vaporizes it all at once. This reduced debris while maintaining efficiency, enabling the jump from 50 to 100 watts.
The Oxygen Fix
Mirrors required cleaning every 10 hours due to coating deterioration from high-energy photons and hydrogen ions. An engineer noticed that opening the machine introduced oxygen, which suddenly cleaned the mirrors. ASML experimented with adding controlled amounts of oxygen to the vacuum, extending mirror life and making the machine commercially viable.
Customer Patience Wearing Thin
By 2012–2013, ASML's customers were furious. The company repeatedly promised milestones it couldn't meet. Executives were 'crucified at every conference' for broken promises. Customers demanded proof: 'This is what you showed two years ago, last year, and today. Why would I believe you?' Desperation set in until the 200-watt milestone was finally achieved.
The Final Machine: Low NA and High NA
Low NA: First Commercial Success
ASML's first commercial EUV machines (low NA) have a numerical aperture of 0.33 and print 13-nanometer lines. By 2016, orders poured in and these machines became essential for advanced chip manufacturing. ASML still produces low NA machines today.
High NA: The Next Generation
ASML began developing the high NA machine in 2012—before the low NA machine even worked. High NA has a numerical aperture of 0.55 and can print 8-nanometer lines. It shrinks patterns 8 times vertically and 4 times horizontally. The mirrors are even smoother than low NA: if scaled to Earth size, the tallest bump would be the thickness of a playing card.
Machine Complexity and Shipping
ASML's high NA machine is assembled in a cleanroom so strict that no more than 10 particles (0.1 microns or smaller) can exist per cubic meter—hospital operating rooms allow 10,000 particles per cubic meter. After assembly and testing, the machine is disassembled for shipping: 5,000 companies supply 100,000 parts, 3,000 cables, 40,000 bolts, and 2 kilometers of hosing. It ships in 250 containers across 25 trucks and seven Boeing 747s.
Why It Matters: The Unreasonable Bet
The Most Important Machine in the World
Every advanced microchip made today—in smartphones, computers, data centers—requires ASML's EUV machine. ASML is the only company in the world that can make it, making them arguably the most critical tech company on Earth. Without EUV, Moore's Law would have ended around 2015, halting the exponential growth of computing power.
Unreasonable Persistence Drives Progress
For 30 years, almost everyone thought building an EUV machine was impossible. Kinoshita and Hawryluk were mocked and rejected. Yet they persisted. The video concludes with a quote: '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.' Without unreasonable people, most modern technology wouldn't exist.
Notable quotes
We don't miss them. You do 150,000 laser shots a second, and you don't miss one? Exactly. — ASML engineer and Derek (Veritasium)
People seemed unwilling to believe that we had actually made an image by bending x-rays. — Hiroo Kinoshita
All progress depends on the unreasonable man. — Derek (Veritasium, quoting George Bernard Shaw)