What was shown
Qubits, the basic units of a quantum computer, make mistakes constantly. Error correction spreads one piece of information across many qubits, so a slip in one can be caught by its neighbours. Picture a group of clerks who each hold a copy of a figure and keep comparing notes with the clerks beside them. A slip by one person gets spotted and fixed. This only works if each clerk is careful enough to begin with; otherwise a bigger group just produces more slips. The paper shows the careful case. On a real superconducting processor, the team built a protected qubit, called a logical qubit, from many physical ones using a scheme called the surface code. Each time the code grew in size, the protected qubit's error rate fell further. A second, smaller processor ran a code with a decoder working in real time: the classical software that reads the checks and decides the fixes.
“The logical error rate of our larger quantum memory is suppressed by a factor of Λ=2.14±0.02 when increasing the code distance by two, culminating in a 101-qubit distance-7 code with 0.143% ±0.003% error per cycle of error correction.”
— Abstract (result restated in Section II, Fig. 1c-d), in: Quantum error correction below the surface code threshold1
- How it was done: run on a physical quantum processor
- Qubits: A 105-qubit processor hosted the largest code, using 101 qubits; a 72-qubit processor hosted a smaller one.
- Error correction: Error correction actually ran on existing superconducting hardware, but only to store information, with no calculation steps on the protected qubit.
- Compared against classical computing: no comparison stated in the paper
Who it affects
This concerns the industries that expect to rely on large quantum computers later: chemistry and materials, and anyone judging how long today's encryption stays safe. Nothing changes in daily work today. The result is not a usable machine; it is hardware evidence that the underlying error-correction scheme behaves as hoped when it is made larger. Planning estimates for fault-tolerant machines, including those for attacks on cryptography and for quantum chemistry, assume this behaviour. They now have a measurement on a physical device behind them.
Quantum computing vendors and their customers: McKinsey & Company estimates $43B to $71B by 2035 — quantum computing market size (vendor side) (Quantum Technology Monitor, April 2026, p. 8).
Chemicals and materials: McKinsey & Company estimates $450B to $800B by 2035 — value at stake for the industry, not vendor revenue (Quantum Technology Monitor, April 2026, p. 22).
These figures are estimates, not measurements: McKinsey & Company is a consultancy with its own quantum advisory business. They cover whole use cases, not the result of this paper.
What is still missing
The paper shows a memory only: no logical gates, which are the calculation steps on protected qubits, and no algorithm. An error rate of 10⁻⁶ would need a protected qubit of code distance 27, the paper's measure of code size, using 1457 physical qubits. Many applications need error rates below 10⁻¹⁰. The paper's simpler test codes also hit a noise floor at 10⁻¹⁰, caused by correlated bursts of errors that remain unexplained. Real-time decoding was run on the 72-qubit processor and still has to work at larger code distances.
Who is working on it
The paper (August 2024) lists its authors' affiliations as: Google Research; Department of Electrical and Computer Engineering, University of Massachusetts, Amherst, MA; Google DeepMind; Department of Physics, University of California, Santa Barbara, CA; and 9 further institutions. That is who signed this work — it is not a ranking of who is ahead.
This strengthens superconducting circuits as a platform for error correction with the surface code, which has neighbouring qubits check one another. It also shows that real-time decoding, the fast classical processing that reads those checks, is feasible at this approach's fast cycle time.
How this brief was made
This brief was written by software, not by a person: a language model read the full text of the paper, and code then checked every quotation, number and name in this brief against that text before it was published. The rules are set out under Method. Responsible for the rules and for this page: Georg Singer. The paper is peer-reviewed, published in Nature 638 (2025) 920-926. Georg Singer holds no shares in any company or institution named in this brief (as of October 2026).