News

Quantum Machines Demonstrates Sub-Microsecond Surface-Code Decoding on OPX1000 Using NVIDIA CUDA-Q QEC and NVQLink

Quantum Machines (QM) has demonstrated mid-circuit quantum error-correction (QEC) decoding on its OPX1000 quantum control platform by integrating the NVIDIA CUDA-Q Realtime API and transport-agnostic CUDA-Q QEC decoding server over NVIDIA NVQLink. Executed on an NVIDIA GH200 Grace Hopper server using pre-sampled surface-code memory syndromes at a 0.1% physical error rate, the bridge maintained a 2.4 μs round-trip controller-to-decoder communication latency while achieving sub-microsecond per-round decoding for code distances 7 and 11.

OPX1000 & NVIDIA CUDA-Q QEC Real-Time Decoding Benchmarks (0.1% Physical Error Rate)
Surface Code DistanceTest Workload (10,000 Shots Each)Per-Round Decode Latency (PyMatching on GH200)
Distance 725, 50, and 75 QEC rounds0.19 to 0.22 μs per round (steady 2.4 μs round-trip link).
Distance 1125, 50, and 75 QEC rounds0.56 to 0.67 μs per round (below 1 μs cycle target).
Distance 1525, 50, and 75 QEC rounds1.20 to 1.50 μs per round (target for GPU dispatch optimization).

To execute real-time feedback within a single quantum program, QM implemented a CUDA-Q Realtime bridge that compiles CUDA-Q static memory circuits through Quake to QUA and streams syndrome measurements via enqueue_syndromes and get_corrections calls during circuit execution rather than in post-processing. Because the CUDA-Q QEC decoding server decouples decoder handlers from transport networking code, users can swap between algorithms such as PyMatching and NVIDIA’s fusion decoder via YAML configuration edits without modifying control firmware.

The initial release routes callbacks through host CPU dispatch threads on the GH200 accelerator server, with direct GPU dispatch planned for subsequent updates. Early access to the OPX1000 CUDA-Q QEC integration will open to customers in December 2026.

Read Quantum Machines’ press release here, view QM’s AQC26 classical control overview here, and read QM’s technical breakdown of Alice & Bob’s OPX1000-NVQLink LDPC-cat decoding test here. Access NVIDIA’s technical blog on the CUDA-Q QEC Decoding Server here and the prebuilt cudaq-decoding container on NVIDIA NGC here. Related QCR coverage includes Alice & Bob’s GPU decoding demonstration via NVQLink here and our SC26 report on NVIDIA’s CUDA-Q QEC 0.9.0 release here.

In GQI Portal

The players behind the news

The team that writes QCR tracks every company, deal and technology in the GQI Factory, GQI's verified database of the quantum industry. Next up: every story linked to its players, coming to QCR's paid plans.

Leave a comment

All fields are required. Your email address will not be published.