99.99% one at a time, 99.9% all five together: the gap researchers just closed
What happened
Researchers in Japan have solved a stubborn problem that has been holding back one promising route to a working quantum computer. The route is called silicon spin qubits. A qubit is the quantum version of a bit — the unit of information a quantum computer works with. Silicon spin qubits are attractive because they can be made using much the same equipment that already produces ordinary computer chips, by the million. The problem was odd. Individually, these qubits performed beautifully. The moment you tried to control several at the same time, accuracy dropped. A team at the RIKEN Center for Emergent Matter Science, with collaborators, traced much of that loss to something called the AC Stark shift. In plain terms: the microwave signals used to control a qubit also nudge its energy levels slightly, which throws its timing out. Controlling several qubits at once means more microwave signals, and more nudging. The obvious fix — calibrate every possible combination of qubits — does not scale. The number of combinations explodes as you add qubits. So the team found a shortcut. They calibrated in pairs, and used those pair measurements to work out the corrections for the whole group. Combined with an improved device design and adjusted microwave pulses, they tested an array of five qubits sharing a single microwave control line. The work was published in Nature Physics, and RIKEN described it in a release reported on 25 October 2026.
Why this matters
Quantum computers only become useful if they are both big and accurate, and those two goals have pulled against each other. It is not especially hard to build one very good qubit. It is hard to build many good qubits that stay good while working together. Every demonstration that only works one qubit at a time is, in a sense, a demonstration of a calculator. This result attacks that directly. Yi-Hsien Wu, the paper’s first author, summed up the method: “our method uses calibrations between pairs of qubits to compensate for the unwanted phase shifts”. Leon Camenzind, the RIKEN research scientist who supervised the experiments and is a co-first author, said the work “shows that high-fidelity control can be maintained even when several qubits are operated at the same time”. The silicon angle matters too. The semiconductor industry has spent sixty years learning to make silicon chips reliably and cheaply. A quantum technology that can borrow that manufacturing base has an advantage over one that needs an entirely new industry built around it. Seigo Tarucha, project manager for Japan’s JST Moonshot Goal 6 silicon quantum computing project, set out the destination: “Our long-term goal is to build a fault-tolerant quantum computer based on silicon spin qubits.” Fault-tolerant means a machine that can detect and correct its own errors. Nobody has built one yet.
What the evidence actually says
The numbers are the story here, and they should be read in order. Tested one at a time, each qubit achieved a single-qubit gate fidelity above 99.99%. The best approached 99.999%. Fidelity is roughly “how often the operation does what you asked”. With up to three qubits running at the same time, fidelity stayed above 99.99%. That is the headline achievement: the drop that used to appear did not. With all five running at once, fidelity was about 99.9%. That is still good, but it is ten times worse than the three-qubit figure. The problem was reduced, not eliminated. The team is open about where the remaining errors come from. They are linked to drive-induced decoherence — the qubits degrade because running five of them needs more microwave power, and that power itself causes damage. Camenzind called it plainly: “This is a limitation, but further improvements to device design could help mitigate this source of error.” It is also worth keeping the scale in view. Five qubits is a laboratory demonstration. Useful error-corrected quantum computing is generally expected to need thousands or more. This is a method that should scale better than what came before, which is different from having scaled. The journal reference is Wu, Y.-H., Camenzind, L. C., Bütler, P., et al., “Simultaneous high-fidelity single-qubit gates in a spin qubit array”, Nature Physics, 2026.
Practical advice
Quantum computing generates more noise than almost any other area of technology, so a few habits help when reading about it. Check whether a reported fidelity was measured with the qubits running alone or together. As this study shows, those are different numbers, and the difference is exactly what the field has been struggling with. Check the qubit count, and treat it as a scale rather than a score. A result at five qubits that is designed to scale may matter more than a bigger number achieved by brute force. Be cautious about claims that today’s encryption is about to fall. Breaking widely used encryption is expected to need error-corrected machines far larger than anything demonstrated. That said, organisations handling data which must stay secret for decades are already being advised to plan for post-quantum encryption, and that planning is sensible now. If you want to follow the field without a physics degree, follow fidelity at scale rather than qubit counts. It is the less exciting number, and the more honest one.
What to know
Silicon spin qubits work well individually but lose accuracy when several are controlled at once. RIKEN researchers traced much of that loss to the AC Stark shift, where the microwave control signals themselves shift a qubit’s energy levels and upset its timing. Rather than calibrating every combination, which does not scale, the team calibrated qubits in pairs and used that to correct the whole group. With improved device design and adjusted pulses, they tested five qubits on a shared control line. Results: above 99.99% fidelity for each qubit alone, above 99.99% with up to three running together, and about 99.9% with all five. The remaining errors come from the extra microwave power needed to drive five qubits. This is a laboratory demonstration at five qubits, not a working machine. Its value is in offering a calibration method that should scale. Sources: RIKEN, “Scientists overcome a hidden obstacle to more powerful quantum computers”, ScienceDaily, 25 October 2026, https://www.sciencedaily.com/releases/2026/10/261024222024.htm; Wu, Y.-H., Camenzind, L. C., Bütler, P., et al., “Simultaneous high-fidelity single-qubit gates in a spin qubit array”, Nature Physics, 2026, https://doi.org/10.1038/s41567-026-03466-2 This article is for general information and does not replace advice from a doctor, pharmacist or other qualified healthcare professional.
Related articles
620 light years apart, and due to collide: three black holes in one galaxy
Astronomers have found three supermassive black holes feeding at the same time inside a single galaxy in the very early universe. The galaxy is called J0148-42…
Jupiter-sized, orbiting something that is not a star, and nobody can agree what to call it
Astronomers have reported strong evidence of a moon-like object orbiting something outside our Solar System. If it holds up, it would be the first natural satel…