Lithography

Moore's Law hit a wall in 2015.

For 50 years, process nodes had shrunk, speeds had risen and the cost of computation had fallen exponentially. By 2015, conventional lithography was running into the wavelength of light and could not keep cutting features in half by following the old route.

ASML spent decades engineering another route.

The Engineering

They invented EUV, extreme ultraviolet lithography.

The machine fires a laser at a droplet of tin traveling at 250 kilometers per hour. You have to hit it three times in a 20-microsecond window. When you do, the tin heats to 220,000 Kelvin, 40 times hotter than the surface of the sun.

At that temperature, the tin emits extreme ultraviolet light at exactly the wavelength you need. That light reflects off mirrors to etch circuit patterns onto silicon.

They do this 50,000 times per second. Every single hit has to be precise.

The mirrors are polished to such smoothness that if you scaled them to the size of Germany, the biggest bump would be one millimeter. One millimeter across a continent.

The entire machine costs $400 million. It weighs 180 tons. It ships in 250 containers. That's 25 trucks and 7 Boeing 747s.

One nanometer precision. Seven hundred billion dollars of semiconductor value flowing out of it annually.

Why persistence mattered

EUV wasn't inevitable. For 50 years, people said "we can't shrink anymore, it's physics." And they were right about the physics. But they were wrong about whether you could engineer around it.

When EUV was being developed in the 1980s, the pioneer literally got laughed off stage at conferences. "This will never work." "It's 50 years away." "You're chasing fantasy."

Twenty years of patient capital. Billions in R&D. Dead ends and breakthroughs. Breakthroughs and dead ends again.

And then it worked. And it kept working. And we got another decade of Moore's Law.

That $400 million machine is sitting in fabs in the Netherlands, California, and South Korea. Every advanced AI chip, from systems used for neural-net training to military hardware, depends on patterns etched by technology that a pioneer was ridiculed for pursuing in 1980.

The Real Insight

This kind of progress is patient and unglamorous: engineers grinding on physics problems for decades while everyone else says they are impossible.

It's refusing to accept the wall.

Many technical problems yield only after someone spends 20 years and a billion dollars on them. The people who crack them are unreasonable. They're the ones who stare at consensus impossibility and say "not in my timeline."

The 10-Year Window

The West may have a 10-year window in which its semiconductor advantage matters.

China cannot buy EUV machines because of export controls. It cannot yet build them; ASML is the only manufacturer and cooperates with Western policy.

That gives the U.S. a genuine structural advantage in advanced computing. It won't last. China's investing in their own lithography. EUV will eventually be everywhere.

But right now? Right now the West has the machine. And that machine determines who wins the AI race.

The Unreasonable People

The reason we have EUV is because one person refused to accept that Moore's Law had to die.

The reason we have AI is because Gary Marcus, Yann LeCun, and others refused to accept that deep learning was the end of the story, even when everyone else was calling them Luddites.

The reason we have nuclear energy, space flight, and the internet is because unreasonable people kept going when reasonable people said it was impossible.

The EUV lesson reaches beyond "buy semiconductor stocks." Pay attention to the unreasonable people doing decade-long work on unsexy problems that everyone says are impossible.

If you want to understand where technology is going, follow the foundational breakthroughs, the expensive machines and the teams that keep building after the conference laughter has stopped.