Bytebench SearchCtrl K

Blog / Tech

Silicon Qubits Hit 99.9% Accuracy While Working Together

Silicon Qubits Hit 99.9% Accuracy While Working Together
In this post

Quantum computers built from silicon have one huge advantage: they could be made in the same kind of factories that produce today's computer chips. Their weakness has been accuracy when many qubits work at once. In October 2026, a team led by Japan's RIKEN showed a way around that problem, running five silicon spin qubits simultaneously with about 99.9% accuracy through a single shared control line.

The study, published in Nature Physics on 15 October, involved researchers from RIKEN, National Taiwan University, ETH Zurich and TU Delft.

First, what is a silicon spin qubit?

A normal bit is 0 or 1. A qubit can be in a blend of both, which lets quantum computers attack some problems in ways ordinary computers cannot.

In a silicon spin qubit, the information is stored in the spin of a single electron trapped in a tiny structure called a quantum dot, etched into silicon. Microwave pulses rotate the spin to perform operations, called gates.

The problem: qubits that must take turns

A single silicon qubit can already be controlled with better than 99.9% accuracy, known as fidelity. The trouble starts when you drive several at once. Microwave signals meant for one qubit leak into its neighbors and shift their timing, a kind of crosstalk.

The obvious workaround, letting qubits take turns, has its own cost. An idle silicon qubit loses its quantum state roughly ten times faster than one being actively driven. Waiting in line makes errors pile up. And quantum error correction, the method that will make large quantum computers reliable, needs many operations happening at the same time.

What the RIKEN team did

The researchers drove all five qubits through one shared microwave line, sending a combined signal containing each qubit's own frequency.

They traced the main source of error to an effect called the AC Stark shift, where a microwave tone aimed at one qubit nudges the frequency of the others. Then came the clever part. Instead of calibrating every possible combination of qubits running together, they measured the shift between each pair and added the effects up. For five qubits, that cuts the work from 26 combinations to 20 pairwise checks, and the savings grow fast with more qubits.

The results

Condition Gate fidelity
Each qubit alone above 99.99%, best near 99.999%
Up to 3 qubits at once above 99.99%
All 5 qubits at once about 99.9%

The remaining drop with all five running was linked to the extra microwave power needed, which slightly disturbs the qubits' quantum state.

What it does and doesn't mean

This is a meaningful step toward fault-tolerant quantum processors in silicon. Using one shared control line also reduces wiring and heat inside the ultra-cold fridge that houses the chip, a major headache when scaling up.

It is not a finished quantum computer. The results cover single-qubit operations, not the two-qubit gates that entangle qubits or a full multi-step computation. Those are the next hurdles.

Sources

Frequently asked questions

What did RIKEN achieve with silicon qubits?

They ran five silicon spin qubits at once with about 99.9% gate fidelity, and three at once above 99.99%, using one shared microwave line.

Why are silicon qubits important?

They can potentially be manufactured with existing semiconductor technology, making large quantum chips easier to build.

What is gate fidelity?

A measure of how closely a quantum operation matches its intended result. 99.9% means about one error in a thousand operations.

Is this a working quantum computer?

No. It is a key building block; two-qubit gates and full computations still need to be demonstrated at this scale.

Where was the study published?

In Nature Physics on 15 October 2026.

Free tools that help with this

Keep reading

The First Nuclear Clocks Are Ticking, and One of Them Steers Itself

The first thorium-229 nuclear clocks are running, and Vienna's version keeps its own laser on track for over a day without help.

Tech

YouTube Music Now Labels AI-Made Songs: What Listeners and Creators Need to Know

YouTube Music now tags songs made mainly with generative AI, and here is what that label does and doesn't tell you.

Tech

MIT's Paper-Thin Robot Swims on Living Muscle and Follows the Light

MIT engineers built a gum-stick-sized robot that swims on a single layer of living, light-controlled muscle cells.

Tech

WhatsApp

↑ ↓ to move Enter to open / or Ctrl K to search anywhere