A Normal Chip Factory Just Made Working Qubits

Frontier Signal · Deep Dive · Quantum & Investing · 8 min read · Wednesday, 23 September 2026

Quantum computers are usually built like laboratory instruments. Researchers tune materials, pattern tiny devices in specialist cleanrooms and celebrate when one carefully made chip behaves well enough to run an experiment.

Classical computing became an industry when that stopped being good enough. The breakthrough was not making a transistor. It was learning to make millions of nearly identical ones on full wafers, measure the yield, discard the bad dies and repeat month after month.

French startup Quobly is trying to drag quantum computing across the same line. On 16 September it said one of its QSOI chips, fabricated inside STMicroelectronics' commercial 300 millimetre FD-SOI facility in Crolles, France, had demonstrated qubit readout, single-qubit gates and two-qubit gates on the same device.

That does not prove a factory can make useful quantum processors at scale. Quobly did not publish gate fidelities or wafer yield. But it does change the manufacturing question. A working device has now come through an industrial semiconductor process rather than a one-off academic flow.

300 MM COMMERCIAL FAB · READOUT PLUS ONE- AND TWO-QUBIT GATES · CLOUD SYSTEM TARGETED FOR END-2026
One working die proves transfer. Yield across many wafers proves manufacturing.

Quobly said the experiments used QSOI devices made at STMicroelectronics' 300 mm Crolles site on FD-SOI CMOS. QSOI combines silicon spin qubits with transistor structures and cryogenic control electronics on one silicon platform.

On one chip the company reports three operations: reading a qubit's state, controlling one qubit, and running a gate between two. Those are the basic verbs a processor needs before algorithms are possible. Quobly calls it an initial validation of technology transfer into an industrial semiconductor process.

The choice of words matters. Initial validation is not volume manufacturing. Quobly says further metrics will appear in a scientific publication. Until then the announcement establishes that the process can produce a device that works, not how often or how accurately.

Quobly uses silicon spin qubits. Instead of storing a quantum state in a superconducting circuit or a trapped ion, the system confines a single electron in a tiny electrostatic pocket called a quantum dot. The electron's spin, a property that behaves roughly like a microscopic magnetic orientation, carries the 0 and the 1.

The appeal is size and compatibility. Quantum dots can be patterned using techniques far closer to transistor manufacturing than rival qubit technologies use. Quobly's QSOI process builds on fully depleted silicon-on-insulator, or FD-SOI, which ST already runs at 300 mm scale.

The chip still has to run at cryogenic temperatures, so ordinary control electronics are not enough. Quobly is building cryogenic electronics and a cryogenic process design kit, letting engineers lay out quantum and control blocks together rather than treating the qubits as an isolated experiment.

In plain English: Quobly is betting that the way to build a very large quantum computer is to make the qubit look like something a semiconductor fab already knows how to reproduce. We set out why the ratio between physical and logical qubits decides the timeline in quantum's problem is too many qubits, not too few.

Quantum roadmaps love qubit numbers because they are easy to compare. Manufacturing does not. A fab asks how many devices on a wafer meet specification, how much they vary from die to die, and whether the next lot repeats the result rather than yielding one hero chip.

That distinction is the heart of this story. A hand-tuned chip can be excellent and still be commercially awkward if every device behaves differently. A less glamorous architecture can win if it inherits statistical process control, automated metrology, 300 mm tooling and decades of manufacturing discipline. We covered the error-correction side of the same problem in the year quantum computers learned to fix their own mistakes.

Quobly has signed agreements with TNO and Orange Quantum Systems to strengthen measurement and high-throughput testing around its 300 mm flow. Those are not side projects. If the company wants to claim industrialisation, wafer-scale characterisation is the loop that has to prove it.

Quobly is private. In June it raised EUR 115 million in a Series A led by Bpifrance, SEALSQ and STMicroelectronics, with the European Innovation Council Fund and Air Liquide's venture arm also taking part. The financing is aimed at industrialisation and bringing the first Alloy product to market.

STMicroelectronics (NYSE: STM; Euronext Paris: STMPA) is the clearest listed read-through: its 300 mm FD-SOI fab is the platform at the centre of the result, and ST also invested in Quobly. The signal is strategically meaningful and financially immaterial at ST's scale today. Treat it as optionality, not a near-term earnings catalyst.

SEALSQ (NASDAQ: LAES) led through its quantum fund, and Air Liquide (Euronext Paris: AI) brings the cryogenic expertise a spin-qubit machine depends on. Neither is pure-play exposure to this result.

Quobly's direct test is whether Alloy Pioneer reaches the cloud by the end of 2026 as planned. The company has also discussed integrating with high-performance-computing environments rather than building a stand-alone appliance that demands bespoke infrastructure.

The competitive field includes other silicon-spin approaches, among them teams using imec and Intel manufacturing. Quobly's milestone is important but not unique in thesis: several groups believe silicon wins because the industry already solved much of the scaling machinery. Quobly has to show its FD-SOI route delivers competitive performance as well as manufacturability.

The factory has made a working chip. It has not published a manufacturing yield.

Quobly did not disclose the fidelity of its single- and two-qubit gates. That omission prevents serious comparison with competing platforms and with error-correction requirements. A two-qubit gate can exist and still be far too noisy to scale.

The second missing number is wafer statistics. A 300 mm process becomes commercially interesting when hundreds of dies across multiple lots show repeatable distributions. One device proves the transfer did not destroy quantum behaviour. It does not prove the fab can make thousands.

The million-qubit target for 2032 belongs in the roadmap paragraph, not the headline. The nearer proof point is less dramatic: publish gate fidelities, then show across-wafer and lot-to-lot yield.

What percentage of the dies on a 300 mm wafer meet the same qubit and gate specification?

A quantum computer becomes a manufacturing story when that answer can be expressed as a yield distribution rather than a photograph of the best chip.

Qubit readout, a single-qubit gate and a two-qubit gate on one QSOI chip made through STMicroelectronics' commercial 300 mm FD-SOI process in Crolles, France.

It is the mainstream wafer size for advanced fabs. An industrial 300 mm line opens a path to mature tooling, statistical process control and much larger volumes.

No. Quobly has not published wafer-level yield or the gate-fidelity numbers needed to judge whether the process is ready for scale.

They can be extremely small and patterned with processes related to conventional transistor manufacturing, which creates a plausible density and manufacturing advantage.

Quobly targets cloud access to Alloy Pioneer by the end of 2026. The more important milestones are published fidelity and repeatable yield across full wafers.

Frontier Signal explains frontier technology in plain English. Company-reported performance claims should be independently verified. This is general information, not investment advice.