Atom Computing Demonstrates Repeated Error Correction in Neutral-Atom Quantum Computer Over 90 Rounds, With Some Error Accumulation

Atom Computing operated a neutral-atom quantum computer that repeatedly detected and corrected its own errors across as many as 90 consecutive rounds. The system spread information across groups of qubits used for error detection. Researchers increased those groups from 16 to 32 qubits without adding new errors, and the larger groups showed lower error rates than the smaller ones.

Ben Bloom at Atom Computing said the experiment marked steady progress rather than large leaps. “The differences between [experiments] we were doing before were big step changes, but now, it is just about building it better, faster, cheaper,” he said. The team kept the machine running and monitoring the alert qubits through the full sequence of checks.

In 2023, researchers at Google increased qubit counts while lowering error rates in a superconducting system. A University of Science and Technology of China team repeated the result in 2025, and a Harvard University group achieved it the same year with neutral atoms. Jeff Thompson at Princeton University said the Atom Computing work combined the necessary capabilities in one experiment.

“This study is the first to bring together all of the capabilities needed to build a real neutral-atom quantum computer in a single experiment,” he said. Mark Saffman at the University of Wisconsin-Madison noted that some additional errors still accumulated over the 90 rounds. Bloom said the group is already addressing those remaining errors.

Bloom added that physical mechanisms previously limiting neutral-atom systems are diminishing. “What this work is showcasing is that a lot of the physical mechanisms that stop neutral atoms from being as awesome as superconducting qubits are starting to disappear,” he said. Thompson said rapid progress is now expected across the industry.