Quantum Motion and academic partners used a commercial CMOS process to shrink a quantum-dot charge sensor. Its sensitivity is striking, but the result is not yet high-fidelity spin-qubit readout.
A strip of titanium nitride measuring 0.36 by 50 micrometres sits at the centre of Quantum Motion’s latest result. It occupies 18 square micrometres, about one micrometre from a transistor used to trap a quantum dot. At deep-cryogenic temperatures, that unassuming film supplies the kinetic inductance needed for a compact radio-frequency sensor.
The work matters because readout hardware can dwarf the spin qubits it is meant to measure. Conventional spiral inductors consume valuable chip area and bring parasitic capacitance. Thomas H. Swift, Alberto Gomez-Saiz and colleagues instead used titanium-nitride films already present in GlobalFoundries’ commercial 22-nm fully depleted silicon-on-insulator process, known as 22FDX.
Their superinductor reached roughly four orders of magnitude greater inductance density than conventional CMOS spirals. The team integrated it with a silicon quantum dot to make a radio-frequency single-electron transistor, or rfSET. This measures changes in electrical charge. With separate spin-to-charge conversion, such sensors can be used to infer a qubit’s spin state.
A sensitivity claim with boundaries
In Nature Electronics on 4 November, the researchers reported a power signal-to-noise ratio of 4,500 at the apparatus’s shortest integration time of 4 nanoseconds. From measurements at longer integration times, they extrapolated a minimum integration time of 1 plus or minus 0.3 picoseconds to reach a signal-to-noise ratio of one at an incident power of minus 77dBm. That is an inferred sensitivity benchmark, not a demonstrated one-picosecond qubit measurement.
The charge sensor did not demonstrate high-fidelity spin readout. The next experiment requires multigate transistors and actual spin-to-charge conversion. Nor does a commercially fabricated chip establish a production-ready quantum computer: the devices operate at millikelvin temperatures, their resonator quality factor was about 50, and readout power can disturb neighbouring qubits.
The authors also model why sensitivity rises so steeply with incident power. A nonlinear response shifts the resonator’s kinetic inductance as the sensor dissipates energy, supplementing the usual change in reflected signal. Above minus 82dBm, their inferred minimum integration time scaled approximately with the inverse cube of incident power, rather than the inverse relationship expected in the linear regime. That agreement between model and measurement supports the mechanism, but the same nonlinearity narrows the usable operating range.
There are other revealing constraints. Static power dissipation held the characterised chip’s base temperature near 350 millikelvin, while the two titanium-nitride variants became superconducting below approximately 0.75K and 1.1K. Above the useful nonlinear regime, about 4dB beyond the best operating point, the film returned to its normal state and the resonance disappeared. The supplementary information matters here because the sensitivity advantage depends on staying within that power window.
Industrial logic and intellectual property
Quantum Motion’s strategy is to borrow semiconductor manufacturing rather than construct every component as a bespoke quantum device. GlobalFoundries announced a partnership with the London company in January 2025 around its 300mm 22FDX platform. The new paper supports the technical premise that cryogenic quantum components can be co-integrated, but it does not prove high-volume yield for this design.
Quantum Motion disclosed a $160 million Series C round in May 2026, co-led by DCVC and Kembara, with new and returning investors including the British Business Bank, Bosch Ventures and Porsche Automobil Holding. That money backs the company’s wider silicon-quantum programme, not this experiment alone. The paper also lists public support from UK research councils, the European Research Council, the Royal Commission for the Exhibition of 1851 and a Gates Cambridge fellowship.
M. Fernando Gonzalez-Zalba, who supervised the work with John J. L. Morton, is named as inventor on relevant patent EP3958188B8. The authors are based at Quantum Motion, University College London, Imperial College London, Cambridge and CIC nanoGUNE in Spain.
For GlobalFoundries, a successful silicon-quantum sector would create another specialised use for an existing process, although quantum volumes would initially be tiny beside conventional chips. Quantum Motion could gain from owning designs and readout intellectual property. Established dilution-refrigerator and cryogenic amplifier suppliers would remain essential, while bespoke superinductor approaches could face pressure if integrated alternatives reproduce across wafers.
Those are exposure scenarios, not forecasts of orders or losses. Replication, high-fidelity spin measurements and arrays of sensors are the tests that will decide whether this is a valuable component business or an elegant laboratory demonstration. Other proposed uses, from parametric amplifiers to astronomy detectors, are further options rather than validated products.