The Invisible Trillion · Sand, Light, and the World’s Most Precise Factories

What This Chapter Is For

Here is why I wanted you to know all of this.

When you hear someone talk about “chips” — the chip shortage, AI chips, data center chips, the cost of chips — it is easy to picture a chip as a commodity, something that gets made the way t-shirts or toasters get made. Lots of factories around the world, competing on price, cranking out the product.

It’s not like that. The manufacture of modern chips is almost nothing like the manufacture of almost anything else. It happens in five or six buildings, in four or five countries, with machines built by two or three companies, using a process so delicate that a power outage can ruin a week’s production. It is more like a cathedral than a t-shirt factory. The whole modern world — every phone, every data center, every car with airbags, every hospital monitor, every AI model — sits on top of this strange, concentrated, fragile pyramid of manufacturing.

When people talk about the environmental cost of AI and ignore the fact that a fab uses more water than a city, that is missing something real. But when people talk about the environmental cost of AI and forget that the fab also makes the chips in the medical equipment that kept their grandmother alive, that is also missing something real. Everything is connected to everything. You can’t tell a sensible story about one part of it without at least knowing the other parts exist.

The next chapter is about what the chips in those packages actually do — how a CPU is different from a GPU is different from the chip in an airbag sensor, and why the differences matter. After that, we go to the data center, which is where most of the chips that run AI end up.

But for now, close your eyes for a second and picture it: somewhere on Earth, right now, a tin droplet is falling through a vacuum chamber. A laser is pulsing. A flash of ultraviolet light that doesn’t exist in nature is bouncing off a stack of molybdenum mirrors. It hits a silicon wafer in a room cleaner than surgery. An atom-thin layer of material is patterned by that light. And the pattern — one of the hundred that will be made on that wafer that day — becomes, eventually, the thing you use to text your friends about how your day is going.

It’s a weird world. You are living in it.

A note on sources: the ASML machine specifications and the details of EUV lithography are drawn from ASML’s own technical publications, from Chris Miller’s Chip War (which I recommend if this chapter was interesting to you), and from academic coverage of semiconductor manufacturing. The cleanroom particulate specifications are ISO 14644-1 standards. The 9N purity figure is standard industry practice for electronic-grade silicon. The percentages of global leading-edge chip production are from industry reports; they shift year to year but the concentration in Taiwan has been roughly stable at 85–92% for about a decade.

All 19 chapters of The Invisible Trillion · Library · All courses