Supercomputing Centers to Integrate Quantum Processors Using Nvidia's NVQLink (thequantuminsider.com)

NVIDIA announced that more than a dozen leading supercomputing centers and national labs across Asia, Europe and the U.S. will adopt NVQLink — a low‑latency, high‑throughput interconnect that links quantum processors (QPU) with GPU-accelerated systems and the CUDA‑Q software stack. The move formalizes a push toward hybrid quantum–GPU supercomputers by providing a universal, Ethernet-based fabric and real‑time APIs that let researchers orchestrate quantum control, error correction and classical GPU workloads within a single programming environment. Technically, NVQLink delivers sub‑4 microsecond latency and up to 400 Gb/s GPU–QPU throughput (with system-level claims like 40 PFLOPS at FP4 precision), enabling microsecond‑scale feedback needed for live quantum error correction and low-latency hybrid algorithms. Quantinuum’s Helios QPU demo used NVQLink and CUDA‑Q to run a scalable qLDPC decoder in 67 μs — 32× faster than Helios’ 2 ms requirement — showing real‑time active error correction is achievable. By combining tight co‑design with controller builders and supercomputing centers, NVQLink aims to simplify scalable classical resources for growing QPU workloads, accelerate development of hybrid quantum‑classical applications, and make practical, large‑scale quantum error correction and simulation experiments broadly accessible.