NVIDIA has unveiled NVQLink, a new platform architecture designed to tightly integrate conventional GPU supercomputing with quantum processors (QPUs). The announcement, made on Monday, November 17, 2025, details a system that uses a high-speed, low-latency interconnect to enable real-time quantum error correction (QEC) and continuous calibration, critical steps toward useful fault-tolerant quantum computing.
NVQLink creates a machine model referred to as the “Logical QPU,” which combines physical qubits, their control electronics, and classical compute resources into a single, unified system. The architecture connects a real-time host—such as an NVIDIA GPU superchip—to a quantum system controller via a standard RDMA over Ethernet network. This connection delivers latency of less than four microseconds, allowing classical computers to process data and send corrections back to the quantum processor within the incredibly short timeframes required for stable quantum operations.
By using standard C++ or Python through the NVIDIA CUDA-Q platform, developers can now write single applications that execute code across both the QPU and the GPU, treating them as peers in a heterogeneous computing environment.
NVIDIA also announced that Quantinuum has adopted NVQLink for its latest “Helios” quantum processor. In a recent demonstration, the companies used an NVIDIA GH200 Grace Hopper Superchip connected via NVQLink to perform real-time quantum error correction on the Helios system.
The setup successfully decoded a quantum low-density parity check (qLDPC) code in just 67 microseconds, enabling feed-forward corrections in real time. The experiment resulted in a 5.4x reduction in error rates for an 8-logical-qubit memory, showcasing the potential of the hybrid architecture to stabilize quantum information against noise.
According to NVIDIA, more than a dozen scientific supercomputing centers worldwide have already committed to adopting the NVQLink architecture. This “open platform” approach is designed to be compatible with various quantum technologies, including superconducting, trapped-ion, and photonic qubits, allowing researchers to integrate their custom quantum hardware with standard high-performance computing infrastructure.
