NVIDIA has announced that leading scientific supercomputing centers around the world are adopting NVIDIA NVQLink, a groundbreaking interconnect designed to seamlessly integrate quantum processing units (QPUs) with high-performance GPU-accelerated computing systems. The technology is being deployed at more than a dozen national research institutions and supercomputing centers across Asia, Europe, and the United States, marking a major step toward the next generation of hybrid quantum-classical computing. At the SC25 conference, NVIDIA emphasized that the future of supercomputing lies in systems that combine the unique capabilities of quantum processors—particularly their ability to simulate complex quantum systems—with the massive parallelism and programmability of GPUs. “In the future, supercomputers will be quantum-GPU systems,” said Jensen Huang, founder and CEO of NVIDIA. “NVQLink with CUDA-Q is the gateway to that future—uniting quantum and GPU computing into a single, coherent system to push the boundaries of what’s computable and unlock new scientific breakthroughs.” NVQLink’s open architecture enables low-latency, high-throughput communication between quantum processors and GPUs, delivering 40 petaflops of AI performance at FP4 precision, 400 Gb/s of GPU-QPU throughput, and sub-four-microsecond latency. This performance is critical for running complex hybrid applications that require real-time coordination between classical and quantum components. The interconnect is tightly integrated with quantum control systems and GPU supercomputing infrastructure through NVIDIA’s CUDA-Q software platform, which provides a unified development environment for building and testing quantum-GPU applications. Designed in close collaboration with quantum hardware developers, controller manufacturers, and research institutions, NVQLink supports flexible, scalable, and high-performance integration. A key milestone in the technology’s real-world application came from Quantinuum, which announced that its Helios QPU and future quantum processors will be connected to NVIDIA GPUs via NVQLink. Using CUDA-Q, the team successfully implemented real-time quantum error correction for a class of codes known as qLDPC (quantum low-density parity-check). This marked the first real-time use of a scalable decoder for qLDPC codes, with a reaction time of just 67 microseconds—32 times faster than the two-millisecond requirement of the Helios system. The achievement was made possible by NVQLink’s ability to deliver the necessary speed and parallelism for effective error correction. The microsecond-level latencies and high throughput of NVQLink are accessible through real-time APIs in CUDA-Q, enabling scientists and developers to experiment with quantum error correction and hybrid applications in a single, streamlined environment. Additionally, NVQLink leverages standard Ethernet infrastructure, allowing researchers to easily scale classical computing resources as quantum systems grow in complexity. Supercomputing centers in the U.S., Europe, the Middle East, and Asia are now adopting NVQLink to advance research in quantum simulation, materials science, drug discovery, and optimization. The technology is already being used at major national laboratories and research institutions to explore the full potential of hybrid quantum-classical systems. Quantum hardware developers and research centers interested in accessing NVQLink can sign up for early access through NVIDIA’s official webpage. Further technical details and use cases are available on the NVIDIA technical blog.