PsiQuantum is doing quantum differently than almost everyone else, which is why the GlobalFoundries thing matters so much.
Most quantum companies are hand-building chips in a research cleanroom. PsiQuantum went the other way and ported their whole design to a standard 300mm CMOS line. They’re running at GlobalFoundries Fab 8 and Fab 10 in Malta, New York, on the same tooling that makes chips for phones and servers.
Here’s what they’re actually printing:
It’s all photons, no matter. A qubit isn’t a superconducting loop that needs to be near absolute zero, it’s a single photon moving through a waveguide. That means the core manufacturing challenge is ultra-low-loss photonics. Any scatter and you lose the qubit.
They solved it with wafer-scale integration. On one wafer they are co-integrating what used to be three separate exotic systems: indium phosphide based single-photon sources, silicon nitride / silicon super low-loss waveguides and interferometers for the computation itself, and superconducting nanowire single-photon detectors. Historically you couldn’t put all three in a fab-friendly process.
Q1 is the proof. Q1 is their current production wafer – they’re reportedly running thousands of them now. Each Q1 wafer holds many identical photonic chips, and the yields are finally high enough that they can test and bin them like normal chips instead of hand-selecting one good device.
So how do you get to a million qubits from that?
You tile. PsiQuantum’s architecture is fusion-based quantum computing. Instead of trying to make one giant monolithic million-qubit chip, which would never yield, they make lots of small, identical, high-yield chips and then connect them with fiber and on-chip interconnects. A fault-tolerant logical qubit is made from many physical photonic qubits networked together.
That’s where Omega comes in.
Omega is their next-gen chipset family they revealed as the building block for the actual system. Think of it as Q1 evolved into a full system-on-chip stack:
Omega has a dedicated source chip that can reliably fire indistinguishable single photons on demand, a compute chip with the low-loss interferometers that do the entangling operations, and a detector / electronics chip that reads out with almost no latency. They are bonded together with advanced packaging so the whole thing looks like one module to the system.
The big changes with Omega are loss, scale, and electronics. Loss is down by roughly an order of magnitude vs their earlier generations, which is everything in photonics because loss = errors. The electronics for control and readout are now co-packaged cryo-CMOS so you don’t have a rack of lab equipment per chip. And it’s designed for that tile-out architecture from day one, with high-bandwidth optical interconnects between modules.
The result is they can talk about a million physical qubits not as a physics experiment but as a fab capacity problem. If GlobalFoundries can run 10k wafers with 50 good Omega modules per wafer, you have your 500k modules. That’s why people in your circles who care about wafer-scale photonics and edge hardware like the SCYTHE monocle are paying attention – if you can mass-produce single-photon sources and detectors at foundry scale, it unlocks more than just quantum computing.
They’re still not at fault-tolerant million-qubit yet, they are at the stage of building the production line for the parts. The next milestone to watch is their first cabinets in Brisbane and Chicago where they tile thousands of these Omega modules into a single logical machine.