Quantum Machines Runs NVIDIA Decoders on OPX1000
Quantum Machines reports first results running NVIDIA CUDA-Q QEC decoders on its OPX1000 quantum control system, with measured communication and decoding latencies.
Updated on Nov. 16, 2026 in Quantum Computing

Quantum Machines announced first results running NVIDIA CUDA-Q QEC decoders on its OPX1000 quantum control system. In a surface-code memory experiment, the OPX1000 sent syndromes to an NVIDIA GH200 over NVQLink and received decoded results in real time.
Why it matters
Real-time decoding is a key requirement for error-corrected quantum computers, since decoding must keep pace with each error-correction cycle. The results show how quantum control hardware and GPU-based decoders can be integrated for U.S. researchers and developers building fault-tolerant systems.
Across nine configurations of 10,000 shots each, decoding latency per round rose from 0.19-0.22 μs at code distance 7 to 1.2-1.5 μs at distance 15, against a 1 μs threshold for a fast QEC cycle. The early implementation routes data through the host CPU, and GPU dispatch has not yet been reported.
The players
Quantum Machines
Quantum Machines is an Israeli company that builds quantum control systems used to operate and read out quantum processors.
NVIDIA
NVIDIA is a U.S. technology company whose GPUs and CUDA software platform are widely used for accelerated computing, including quantum error-correction decoding.
The details
The experiment ran a single CUDA-Q program on the OPX1000 performing a surface-code memory experiment, compiled once and executed across code distances 7, 11 and 15 at 25, 50 and 75 rounds of error correction. Syndromes at a 0.1% physical error rate were sent to an NVIDIA GH200 over NVQLink, where the PyMatching decoder processed them. The CUDA-Q QEC decoding server registers decoders as handlers and is transport-agnostic, and OPX1000 customers will be able to run these decoders through it.
Timeline
November 16, 2026: Quantum Machines announced first results running CUDA-Q QEC decoders on the OPX1000.
The Tech Race
Real-time quantum error-correction decoding has been a central hurdle for fault-tolerant quantum computing, and NVIDIA's CUDA-Q QEC decoding server is one attempt to supply GPU-based decoders for it. This demonstration extends that platform to Quantum Machines' OPX1000 control hardware as competitors pursue their own control and decoding stacks.
For now this work matters mainly to researchers and developers building quantum systems, not to everyday users. Its practical effect depends on whether GPU dispatch and further optimization reduce decoding latency at larger code distances.
The takeaway
Real-time decoding speed is becoming as important as qubit count for building useful quantum computers. Watching whether latency stays under the 1 μs threshold at larger code distances will show how close these systems are to practical fault tolerance.
Further reading
Explore more on Quantum Computing for background on error-correction hardware and decoding.






