IonQ Decoder Runs 408 Simulated Logical Qubits on One CPU

The decoder handled simulated workloads of up to 408 logical qubits on one conventional CPU. IonQ's own paper also shows why this is not a substitute for better hardware: decoding delay rises sharply as the assumed two-qubit error rate worsens.

IonQ has demonstrated an end-to-end quantum error-correction decoder that runs in real time on a single conventional CPU, reducing the amount of classical infrastructure its preferred fault-tolerant architecture may need.

The headline scale is easy to misread. IonQ benchmarked the decoder on simulated workloads containing as many as 408 logical qubits, 88 memory blocks and magic-state factories, with more than 31.5 million quantum operations. It did not operate a physical machine containing 408 working logical qubits.

The more useful result is latency. IonQ's technical paper says the complete decoding pipeline can run on a single CPU and, assuming a trapped-ion cycle time of 1 to 5 milliseconds, adds less than 0.3% execution stretch across the tested workloads when the assumed CNOT error rate is 10^-4. At a CNOT error rate of 5 x 10^-4, the worst tested stretch rises to below 12%.

That sensitivity turns a promotional milestone into a system-design result. Classical decoding need not require a specialised compute cluster if the underlying physical hardware is already operating at sufficiently low error rates. The same decoder becomes less invisible as physical errors rise.

Hardware fidelity still sets the timetable

Fault-tolerant quantum computing has several coupled scaling problems. More physical qubits create more syndrome data; error-correcting codes add overhead; magic-state factories consume resources; the classical controller has to process corrections without stalling the quantum workload.

IonQ has now made the classical side of that chain more credible. It has not demonstrated the physical qubit count, two-qubit fidelity, logical error rates or manufacturing repeatability needed to run those simulated workloads on an actual machine.

That is particularly important against IonQ's roadmap. The company has published an aggressive path from small logical systems to hundreds of logical qubits. A decoder that keeps up at MegaQuOp scale removes one architecture objection to that plan. It does little to reduce the probability of delay if trapped-ion hardware cannot scale while preserving the physical fidelity assumed in the paper.

The paper removes one classical-scaling objection from IonQ's roadmap. Valuation still rests on the part the simulation cannot supply: physical logical qubits running useful algorithms at scale, with measured error rates low enough that the classical decoder remains close to invisible.

Editorial source record

• IonQ, real-time decoder announcement, 22 September 2026.
• Ye, Maksymov and Delfosse, “Real-time decoder for a MegaQuOp quantum computer using a single CPU,” arXiv:2608.25027.
• IonQ technical roadmap.