PsiQuantum is trying to build a commercially useful quantum computer from photons, the particles of light, and the company is getting close to the point where its claims have to survive contact with hardware at scale.
The Palo Alto-era quantum pitch is familiar: use qubits, which can occupy multiple states, to run calculations that overwhelm ordinary computers. The useful version remains missing. Existing quantum prototypes are generally too small, too noisy, or both. PsiQuantum’s answer is a machine that, according to the company, would eventually link roughly 100 stainless-steel cabinets, each cooled with liquid helium and packed with chips that route and measure individual photons.
The company was founded in 2016 by four physicists from UK universities: Terry Rudolph, Mark Thompson, Pete Shadbolt and Jeremy O’Brien. O’Brien served as chief executive until February, when Victor Peng, a semiconductor industry veteran, replaced him. PsiQuantum has raised serious money for a company still chasing its first useful machine. It raised $1 billion last year, broke ground on a Chicago-area site with local government partners, and is also building a site in Australia that it says will be operational, meaning ready for hardware installation, in 2027.
Why photons, and why this is hard
Most quantum hardware efforts use fragile systems such as superconducting circuits, ions, atoms or electrons. Google and IBM have backed superconducting qubits. Intel has worked with electrons. PsiQuantum’s bet is light.
Photons can preserve quantum states for long periods, Terry Rudolph has argued. They also create a nasty engineering problem: photons tend to pass through one another rather than interact. Quantum computing needs qubits to influence one another, so PsiQuantum relies on a scheme first described in 2001 by researchers at Los Alamos National Laboratory and the University of Queensland. The idea is to use networks of beam splitters and detectors to mimic interactions between photons.
In PsiQuantum’s design, lasers generate photons, the system entangles them so their states are linked, optical circuits route them through gates, and detectors measure them at the end. Error correction has to keep the computation from falling apart. Doing that once is a lab result. Doing it millions of times across a machine the size of a small data center is the part where the marketing slides meet plumbing.
The company says its approach can use existing semiconductor manufacturing infrastructure. Its chips are made by GlobalFoundries in Malta, New York. PsiQuantum has also invested in producing barium titanate, a material it says can route photons quickly while using little electrical input. The company says it now produces several wafer discs of the material per day.
The cold part is smaller, not cheap
PsiQuantum’s machines still need cryogenics. Its cabinets currently run around 2 kelvin, or about minus 456 degrees Fahrenheit. The company says photonic machines require only the single-photon detectors to be that cold, rather than the full computing system used in many other approaches. PsiQuantum said in May that some of a planned $100 million CHIPS Act award would go toward those detectors.
The company has connected three cabinets at its Milpitas, California, site, with 250 chips in each, according to its account. The next test is whether its error-correction methods hold up as it scales. PsiQuantum says the Australian facility’s cooling system is expected late next year, after which it aims to connect about 100 cabinets.
DARPA is one of the few outsiders with a structured view into the company’s progress. The Defense Advanced Research Projects Agency has been evaluating quantum firms, and PsiQuantum is one of two companies, along with Microsoft, to reach the third stage of that program. Joe Altepeter, who previously ran the effort and described himself as a quantum skeptic, said in March 2025 that he was more optimistic than at any point in the previous decade. His successor, Micah Stoutimore, said earlier this year that it now appears likely someone will build a utility-scale quantum computer by 2033.
Other observers remain cautious. Scott Aaronson, a theoretical computer scientist at the University of Texas at Austin, said PsiQuantum is difficult for outsiders to judge. That is the correct posture. The company has a plausible manufacturing story, a large checkbook, and an unusually ambitious target. It does not yet have the useful quantum computer it says photons can deliver.
This story draws on original reporting from MIT Technology Review.
