Researchers at RIKEN in Japan have controlled five silicon spin qubits at the same time through a single shared microwave connection, keeping accuracy at about 99.9 percent. Published in Nature Physics on 15 October 2026 and reported widely on 26 October 2026, the work does not build a working quantum computer. It removes a stubborn obstacle to building a much larger one: the wiring problem that grows with every qubit added.
That distinction matters. This is a control experiment on five qubits, not a finished machine. Its value lies in the method it proves — that qubits can share a control line without losing the accuracy error correction will one day need.
What is a silicon spin qubit?
A normal computer bit is either 0 or 1. A qubit can be 0, 1, or a combination of both at once, a property called superposition. A silicon spin qubit stores information in the spin of a single electron trapped in a tiny structure called a quantum dot.
Think of spin as a tiny internal arrow. Pointing one way means 0, the other way means 1, and superposition allows combinations in between. Physicists rotate that arrow with microwave pulses. Each controlled rotation is a "gate", the quantum equivalent of a basic operation in an ordinary processor.
Silicon spin qubits are attractive for two practical reasons. They are extremely small — each one is essentially a single trapped electron — and they are made from silicon, the material at the heart of ordinary computer chips. In principle, they could be produced using existing semiconductor manufacturing methods, though that does not mean a quantum chip can roll off a factory line tomorrow.
Qubits are also fragile. Their states are easily disturbed by heat, vibration, stray electrical noise, and even the signals used to control them. Keeping one qubit accurate is hard. Keeping many accurate at once, as any real computation requires, is much harder.
Why controlling qubits at the same time is so hard
A useful processor must operate many qubits in parallel, just as an ordinary processor performs many operations at once. The RIKEN team, led by first authors Yi-Hsien Wu and Leon Camenzind at the RIKEN Center for Emergent Matter Science, with collaborators in Taiwan, Switzerland, and the Netherlands, worked with five quantum dots driven through one shared microwave line.
Sharing a line is deliberate, and it is also the problem. Giving each qubit its own control wire works for a handful of qubits. It fails for thousands or millions: every extra cable adds heat, cost, assembly work, and crowding inside a refrigerator that must stay extremely cold.
Sharing has a price, however. A microwave pulse tuned to one qubit also reaches its neighbours, slightly off tune. It cannot flip a neighbour outright, but it can disturb it. This is the AC Stark shift, which the team identified as the main error source when several qubits are driven together. The field changes a neighbour's effective energy levels, shifting its resonance frequency and therefore the phase of its quantum state. Small phase errors accumulate, and eventually corrupt the result. In a two-qubit experiment reported in 2022, driving a second qubit cut accuracy sharply, with error rates rising roughly tenfold.
The pair-based fix: measure pairs, predict the whole group
The team's key insight is that these phase shifts add up predictably. Measure how much one qubit drifts when a second qubit is driven, and how much it drifts when a third is driven, and the sum predicts what happens when both neighbours are driven together.
That changes the calibration problem completely. Calibrating every possible combination of simultaneously driven qubits separately grows exponentially with qubit number, and quickly becomes impossible. The pair-based method needs only calibrations between pairs. For five qubits, that is 20 pairwise calibrations. As arrays grow, the workload grows roughly with the square of the qubit number, not exponentially — still work, but work that can actually be done.
"Rather than requiring an entirely separate calibration procedure for every possible combination of qubits, our method uses calibrations between pairs of qubits to compensate for the unwanted phase shifts," said Yi-Hsien Wu, the study's first author, who did the experiments as part of his PhD research. The team also used an improved device design and carefully shaped microwave pulses to reduce signal leakage.
The numbers: 99.99 percent alone, about 99.9 percent together
Fidelity measures how accurately a quantum operation is performed. A fidelity of 99.9 percent means roughly one operation in a thousand goes wrong; 99.99 percent means roughly one in ten thousand. That difference matters because error correction only works if the underlying error rate is already low enough.
Operated individually, every qubit in the array reached single-qubit gate fidelities well above 99.99 percent, with some approaching 99.999 percent. With up to three qubits driven at once, fidelities stayed above 99.99 percent. With all five driven together, fidelity held at around 99.9 percent, the level widely targeted for error-correction schemes such as the surface code. Some summaries put four of the five qubits above 99.9 percent in full parallel operation, with the fifth close to that mark — a reminder that "around 99.9 percent" describes the group result, not an identical result for every qubit.
The remaining loss in parallel operation was attributed mainly to decoherence from the stronger microwave power needed to drive several qubits at once. Unlike the phase shifts fixed by calibration, decoherence is the fragile quantum state itself fading. "This is a limitation, but further improvements to device design could help mitigate this source of error," said Leon Camenzind, a research scientist at RIKEN and co-first author. Note, too, that these are single-qubit gates. Two-qubit gates, where qubits interact and become entangled, are harder and were not the focus here. A useful machine needs both.
Why this matters for scalable quantum computing
Scaling is the central problem of quantum computing. The difficulty is not only making a good qubit. It is making thousands or millions of good qubits, controlling them all, reading their results, and correcting their errors, without the system collapsing under its own complexity.
Wiring is one of the least glamorous parts of that problem, and one of the most decisive. A machine needing a separate control line for every qubit cannot easily grow, because the refrigerator, electronics, and chip all run out of room. Accurate shared lines reduce the number of lines, the heat they carry, and the assembly burden, bringing the design closer to something a chip factory could plausibly build.
That is why five qubits can matter more than their number suggests. Five qubits cannot run a useful program, but five qubits sharing one line at error-correction-level accuracy test the old worry that sharing must cost accuracy. If pair-based calibration keeps working at larger scales — still to be proven — one main barrier to scaling silicon processors becomes an engineering problem rather than an open physics question. This kind of careful measurement also runs through other recent quantum work, from Bethe Strings Finally Seen: A 95-Year-Old Quantum Prediction Comes True to The Ghost of Superconductivity Finally Caught in Action.
What this does not mean — yet
It would be wrong to read this result as the arrival of practical quantum computers. First, scale: five qubits are a laboratory array, not a processor. Useful error-corrected machines are expected to need many thousands, and for some problems millions, of physical qubits, because each reliable "logical" qubit will be built from many physical qubits checking each other.
Second, completeness: a full machine also needs accurate two-qubit gates, fast reliable readout, and error correction running continuously. None of those are solved by this paper alone. Third, the remaining drive-power decoherence still limits parallel performance. Finally, silicon is one of several leading designs, alongside superconducting circuits and trapped ions, and this result does not settle which approach will reach practical machines first. The same measured view applies to other careful demonstrations, such as Maxwell's Demon Harnessed: Turning Information Into Energy, which clarified a fundamental idea without claiming a finished technology.
What to watch next
The next tests are easy to state, even if hard to perform. Can pair-based calibration hold its accuracy on ten qubits, then fifty, then hundreds? Can it combine with accurate two-qubit gates on the same shared lines? Can device redesign reduce drive-power decoherence, and can readout keep pace with control?
If the answers are yes, silicon processors gain a credible route to the large arrays error correction demands. If they are mixed, the method may still help smaller devices, or need combining with other control schemes. Either way, the question has changed: no longer "does sharing ruin accuracy?", but "how far can accurate sharing go?" For now, the honest summary stands. RIKEN controlled five silicon spin qubits through one microwave line, held accuracy above 99.99 percent for up to three at once, and at around 99.9 percent for all five — and practical quantum computers remain distant, with one of their obstacles now smaller.