The Machine Nobody Else Can Build

It's Tuesday, October twenty-seventh. The chip in your phone was drawn with light. A machine shines light through a stencil onto a silicon wafer, and wherever the light lands, the circuit gets built. Shorter light, smaller circuit, faster chip.

Every advanced chip on earth is drawn by one of those machines, and every one of them is made by a single company, in a single Dutch town called Veldhoven. Not most of them. All of them.

That's the story. The most important manufacturing tool in the world is a monopoly, and it now sits in the middle of a fight between Washington, The Hague, and Beijing.

The smaller the wavelength of the light, the smaller the features you can print. So the entire history of chips is a history of finding shorter light.

The industry got down to a hundred and ninety-three nanometers and stopped. That was the wall. The next step, extreme ultraviolet, is thirteen point five nanometers — about fourteen times shorter.

Which should be easy, except that extreme ultraviolet light is absorbed by everything. Glass absorbs it. Air absorbs it. So you can't use a lens, because there's no material transparent enough to make one. You have to use mirrors, in a vacuum, and the mirrors have to be the flattest objects humans have ever manufactured.

And you can't buy a light bulb for it. To make the light, you fire a laser at a droplet of molten tin fifty thousand times a second, hitting it twice — once to flatten it, once to turn it into plasma — and the plasma gives off the wavelength you need.

That's one machine. It weighs about a hundred and eighty tonnes and it's roughly the size of a school bus.

The reason nobody else builds one isn't the idea. It's that the machine is a stack of components that each have exactly one supplier.

The mirrors come from Zeiss, a German company that's been making optics since the eighteen hundreds. The light source came from a company ASML bought. The machine has thousands of parts from hundreds of suppliers, and the tolerances are absurd — the mirrors are polished to a smoothness where, if you scaled one up to the size of Germany, the tallest bump would be a millimeter.

ASML spent about two decades and a lot of near-bankruptcies getting this to work. The machine costs well over a hundred million dollars, and it ships in pieces and gets assembled in the customer's factory over months, because it doesn't fit through a door.

So when people talk about the chip industry as a global effort — design in California, fabrication in Taiwan, memory in Korea — there's a chokepoint underneath all of it. One company, one product line, one town.

Which is why the machine became a diplomatic problem.

The United States decided China shouldn't have extreme ultraviolet tools, and since twenty nineteen, ASML hasn't sold one there. Then the Dutch government added its own export controls in twenty twenty-three. Then Washington started pressing the Netherlands to stop servicing the older machines China already has.

ASML is caught in the middle. It's the Netherlands' most valuable company, and its biggest market for the older equipment is China — roughly a third of its sales in recent years, all of it the less advanced machines, because the advanced ones are banned.

And here's the thing about export controls: they only work when there's a chokepoint. If five companies made this machine, a ban would be a speed bump. Because one does, the ban is a wall. That's an accident of engineering history, not a strategy anyone designed.

Here's the part that makes the monopoly feel less like a strategy and more like luck.

ASML started in 1984 as a spin-off from Philips, with a few dozen employees and a shed for an office. It lost money for years. The early machines were late and unreliable, and at various points the parent company wanted out.

The extreme ultraviolet program nearly finished it. The physics was considered impossible, the schedule kept slipping, and the development cost ran into the billions before a single machine was sold. What kept it alive was that the customers decided to pay for it: in 2012, Intel, Samsung, and TSMC bought stakes in ASML specifically to fund the next generation of machines.

Which is a strange arrangement when you think about it. The three biggest chipmakers in the world had to invest in their own supplier, because there was no second supplier to fall back on.

The next version, High-NA, costs roughly twice as much again. So the chokepoint is getting narrower and more expensive at the same time.

So the phone in your hand has a lineage. It goes back through a fab in Taiwan, to a machine assembled in a cleanroom, to mirrors polished in Germany, to a laser hitting a droplet of tin fifty thousand times a second.

All of it behind one address in a Dutch town.

Stay informed, stay curious, and we'll see you tomorrow.