PsiQuantum

the all-or-nothing photonic bet

3 min readQuantum Explained

Key facts

1Mfault-tolerant
Qubit target
2027-28target
Timeline
$1bnNVIDIA in round
Series E
$620mAUD, Brisbane
Gov backing
99.22%two-qubit
Fusion fidelity
99.98%single-qubit
Prep fidelity

The all-or-nothing photonic bet. Photonic qubits using silicon photonics manufacturing with GlobalFoundries, skipping the NISQ era entirely and going directly for a million-qubit fault-tolerant system, targeted 2027 to 2028.

The wager

PsiQuantum has made one of the boldest wagers in computing: that the shortest route to a useful quantum computer is to skip the machines everyone else is building and go straight for a million-qubit, fault-tolerant system. Most of the field is shipping small, noisy processors and hunting for problems they can already solve. The company is stepping over that phase, the so-called NISQ era, altogether. Its qubits are photonic, encoded in individual particles of light, and its manufacturing partner is GlobalFoundries, whose silicon photonics production lines already turn out optical chips at commercial volume. The target for a working fault-tolerant machine is 2027 to 2028.

Why photonics

The logic behind the photonic choice is industrial rather than exotic. Photonic qubits can be produced using silicon photonics manufacturing, the same processes that make the optical components inside data centres and telecoms networks. That lets PsiQuantum draw on an existing supply chain instead of inventing one, which is why it has bound itself so closely to GlobalFoundries. The argument is simple: if quantum computers are ever to reach millions of qubits, they will have to be fabricated the way ordinary chips are, and photonics is the modality best suited to that kind of scale.

Fusion and fidelities

Photons behave differently from the atoms and ions used by rival machines. They do not sit still in a trap; they travel, so a photonic computer generates them, routes them through optical circuits and combines them in measurements, an operation the field calls fusion. Progress is tracked through fidelities, the probability that a given operation does what it is meant to. The company has published milestones for its Omega chipset, with reported single-qubit preparation of 99.98% and two-qubit fusion of 99.22%. Numbers of that order are what a fault-tolerant design needs, because error correction only works once the underlying hardware is already good enough to be corrected. Alongside the chip work, it has begun assembling the physical estate a large machine demands, with compute centres in Brisbane and Chicago due to break ground in 2026.

The funding

None of this is cheap, and the funding reflects the scale of the ambition. PsiQuantum raised a $1bn Series E with NVIDIA among the participants, and secured a further $620m AUD from the Australian government, the public money tied to the Brisbane site. That mix of private capital and sovereign backing is becoming familiar in quantum computing, where the sums are now large enough that governments treat the technology as strategic infrastructure rather than a speculative punt.

An all-or-nothing bet

The honest way to describe PsiQuantum is as an all-or-nothing proposition. It is not selling access to a modest processor today and iterating towards something better, the path most of its rivals have taken. It is building for a single outcome, and it will either deliver the first genuinely fault-tolerant machine or exhaust its funding in the attempt. That makes the company easy to misread. Judged as a 2026 vendor, it has little to put in front of a paying customer. Judged as a 2028 bet, it is one of the most consequential firms in the sector.

Where PsiQuantum sits in the wider field is a question of temperament as much as technology. Most of the industry has concluded that the safe path is incremental: build what you can, demonstrate advantage on narrow problems, and let error correction improve year by year. It has taken the opposite view, that incremental progress on noisy hardware is a distraction from the only prize worth having. If the 2027 to 2028 timeline slips badly, or if those fidelities fail to hold as systems grow, the strategy will look reckless. If it holds, the company will have crossed into fault tolerance while others were still counting qubits. Readers can follow how the competing approaches are faring through our quantum explainer hub and the wider quantum coverage; for now, PsiQuantum remains the clearest test of whether the direct route to a fault-tolerant machine can be reached at all.