the millikelvin · companion to the Deep · July 17, 2026
Two ways to build a
quantum computer
Google now bets on both. One is a chip in a giant refrigerator in Santa Barbara. The other is a grid of single atoms, held in light, in Boulder. Here is how the Boulder machine actually works, and why the giant wanted it.
The strategy
Google's two-lane roadmap
For a decade Google chased one design. In March 2026 it added a second, and built it in Boulder. The two approaches solve the same problem in almost opposite ways.
Superconducting
Neutral atoms
The machine
How a neutral-atom computer works
No chip, no fridge. Just atoms, laser light, and exquisite control. Four moves. Click through them.
Try it
A tweezer array, in your browser
This is a toy version of what Kaufman's team does for real: a grid of atoms held in light. Fire a Rydberg pulse to entangle neighbors. Rearrange to see why "the atoms can move" is the whole point.
In a real machine the atoms are ytterbium, cooled to a millionth of a degree and held by tightly focused lasers. Excite two neighbors into big, puffy Rydberg states and they feel each other strongly enough to entangle. That is the two-qubit gate. And because the traps are just light, you can pick atoms up and set them down somewhere else, rewiring the computer on the fly. Ytterbium has a bonus: the same atom that stores a qubit is also one of the best atomic clocks ever built.
why here
This is the physics Boulder perfected
Trapping single atoms in light, arranging them, and entangling them through Rydberg states was pioneered on the Front Range. Google did not invent this lane. It hired the people who did, and left them plugged into the labs that trained them. That is the story of the Deep.
Read the Deep: When the Giants Move In →