Inside ASML’s $350M Machine That Makes AI Chips

David (0:00)

Right now, inside your phone, there is basically a sprawling metropolis.

Sophia (0:04)

Oh, absolutely.

David (0:05)

Like, billions of microscopic skyscrapers, all working perfectly. And to build that city, human beings literally had to figure out how to create a hurricane inside an artificial sun.

Sophia (0:17)

A sun that explodes, what, 50,000 times a second?

David (0:21)

Exactly, it's completely wild. Welcome, everyone, to techdaily.ai. I'm your host, David, and joining me today is our expert, Sophia.

Sophia (0:28)

Hi, everyone. It's so great to be here.

David (0:30)

Before we really get into this, you can sponsor this podcast for just $25.

Your message will be featured across major platforms like Apple Podcasts, Amazon Music, Spotify, and more. If you're interested, visit techdaily.ai to get started today.

Sophia (0:43)

So we're talking about what might be the single most complex commercial product humanity has ever built.

David (0:49)

Yeah, a machine that basically single-handedly saved Moore's Law, which I think most people know, right? Moore's Law.

Sophia (0:56)

Right, Gordon Moore, the Intel co-founder back in 1965

David (0:59)

Right, he observed that the number of transistors you could cram onto a microchip doubles roughly every two years.

Sophia (1:04)

And that relentless drumbeat is basically the foundation of our entire digital economy.

David (1:09)

Yeah, we just expect our devices to keep getting exponentially faster and smaller and cheaper. But what a lot of people probably don't realize is that around 2015, that progress came to a screeching halt.

Sophia (1:23)

It really did. The tech industry hit a physical brick wall.

David (1:27)

Like, they literally could not figure out how to shrink these transistors any further, using the laws of physics as they knew them.

Sophia (1:33)

Because to understand that wall, you have to look at how a chip is actually drawn. You have to look at photolithography.

David (1:39)

Right, so let's break that down for everyone listening. Basically, you take a pure, perfectly polished slice of a silicon crystal, a wafer.

Sophia (1:46)

Yep, and you coat it in this light-sensitive chemical, which is called a photoresist.

David (1:51)

Then you have a pattern mask, a reticle, and that holds the microscopic blueprint for chips, wires and transistors.

Sophia (1:57)

Exactly, and then you shine a light through that mask. The light hits the photoresist, and the exposed areas chemically weaken.

David (2:05)

So you can just wash them away.

Sophia (2:07)

Right, and then you etch the pattern into the silicon, and you just repeat that over and over to build those city layers you're talking about.

David (2:14)

But the whole process, the whole limitation, comes down to the light itself.

Sophia (2:19)

Specifically, the wavelength of the light.

David (2:21)

Which brings us to the Rayleigh equation. Because physics has hard boundaries.

Sophia (2:26)

Right, it does. And as engineers try to print these smaller and smaller features, the gaps in that mask start getting really close to the actual wavelength of the light they're shining through it.

David (2:37)

And when the gap is smaller than the wave, the light doesn't just travel in a straight line anymore.

Sophia (2:42)

No, it diffracts, it bends, it spreads out.

David (2:45)

So those wave fronts of light passing through a microscopic gap, they start interfering with each other.

Sophia (2:50)

Exactly. The waves from one gap overlap with the waves from the next gap. Sometimes they amplify each other, sometimes they just cancel each other out entirely.

David (2:58)

So instead of a razor sharp blueprint on your silicon, you just get this blurred useless mess. I always think of it like, imagine you're trying to draw an ultra fine, super detailed architectural blueprint, but you are forced to use a thick, leaky permanent marker.

Sophia (3:18)

That's a perfect way to put it.

David (3:19)

The ink just bleeds everywhere. All your fine details turn into a giant blob.

Sophia (3:23)

And for decades, the industry just solved this by finding thinner markers. They engineered machines to use shorter and shorter wavelengths of light.

David (3:31)

Right, until the late 90s, when they settled on 193 nanometer deep ultraviolet light.

Sophia (3:37)

And that 193 nanometer light, I mean, that was the workhorse for the entire industry, right up until that barrier in 2015

David (3:44)

They tried everything to keep it going, didn't they?

Sophia (3:46)

Oh yeah, they used every optical trick imaginable. They even started immersing the lenses in pure water to bend the light just a tiny bit more.

David (3:54)

Wow, just to squeeze out a few more nanometers of capability.

Sophia (3:56)

Exactly, but eventually, they just couldn't shrink the patterns any further. The marker was too thick.

David (4:02)