NVIDIA’s NVQLink, an open and universal interconnect, is set for adoption by 17 of the world’s leading scientific supercomputing centers across Asia, Europe, and the Middle East, joining the 9 U.S. national laboratories that had previously announced support. This technology provides the critical low-latency, high-throughput link needed to tightly integrate quantum processors (QPUs) with state-of-the-art NVIDIA accelerated computing—ushering in the era of hybrid quantum-classical supercomputers.

Unifying Quantum and Classical Hardware

NVQLink’s open system architecture overcomes control and error-correction challenges that have historically plagued the development of hybrid quantum-classical applications. By utilizing standard Ethernet technology, it delivers a high-performance connection between QPUs and NVIDIA GPUs, enabling large-scale quantum-classical workflows powered by the NVIDIA CUDA-Q™ software platform.

The interconnect boasts impressive specifications tailored for real-time quantum control:

  • Latency: Less than four microseconds (<4 µs.)
  • GPU-QPU Throughput: 400Gb/s.
  • AI Performance: 40 petaflops at FP4 precision.

Key institutions around the globe are integrating the technology:

  • Asia/Pacific: Japan’s RIKEN and AIST (G-QuAT), Korea’s KISTI, Taiwan’s NCHC, Singapore’s National Quantum Computing Hub, and Australia’s Pawsey Supercomputing Research Centre.
  • Europe/Middle East: Italy’s CINECA, Denmark’s DCAI, France’s GENCI, Czech Republic’s IT4I, Germany’s JSC, the U.K.’s NQCC, Poland’s PCSS, the UAE’s TII, and Saudi Arabia’s KAUST.
  • United States: Brookhaven National Laboratory, Fermi National Accelerator Laboratory, Lawrence Berkeley National Laboratory, Los Alamos National Laboratory. MIT Lincoln Laboratory. National Energy Research Scientific Computing Center. Oak Ridge National Laboratory. Pacific Northwest National Laboratory, Sandia National Laboratories

Real-Time Quantum Error Correction Demonstrated

A core challenge for scalable quantum computing is quantum error correction (QEC), which requires lightning-fast classical computation to monitor and correct the fragile state of qubits.

Quantum computing leader Quantinuum has successfully demonstrated the power of the NVQLink and CUDA-Q synergy. Their latest Helios QPU is integrated with NVIDIA GPUs using NVQLink to orchestrate QEC. This partnership achieved the world’s first real-time use of a scalable decoder for qLDPC codes (quantum Low-Density Parity-Check). The system achieved an error-correction reaction time of 67 µs., exceeding Helios’ two-millisecond requirement by a factor of $32$. This breakthrough showcases NVQLink’s ability to provide the low-latency, massive parallelism required for practical fault-tolerant quantum computation.

NVQLink’s microsecond latencies and high throughput are exposed to developers via real-time APIs within the NVIDIA CUDA-Q platform, streamlining the development and testing of next-generation QEC and hybrid quantum-GPU applications within a single programming environment.

A press release from NVIDIA announcing this additional support as well as a description of the demonstration with Quantinuum can be found here. And a blog posting containing additional technical information about NVQLink is available here.

November 18, 2025

Print Friendly, PDF & Email