Jülich Centre Deploys NVIDIA DGX Quantum, HPC Combines Quantum and Classical

Prof. Dr. Kristel Michielsen from the Jülich Supercomputing Centre announced the deployment of NVIDIA DGX Quantum, a hybrid platform that couples the Grace Hopper Superchip with the Quantum Machines OPX1000 controller, achieving round‑trip data‑transfer latency under 4 µs—a 1,000‑fold improvement over previous implementations. The system incorporates Arque Systems’ 5‑qubit processor, which uses electron shuttling to couple qubits, and delivers microsecond‑scale analog feedback that keeps classical processing within qubit coherence times. By embedding this quantum‑classical stack into JSC’s exascale JUPITER supercomputer, the centre becomes the world’s first high‑performance computing hub to host a fully integrated quantum accelerator, paving the way for large‑scale hybrid algorithms and real‑time quantum‑error‑correction benchmarking.

Jülich Supercomputing Centre Deploys First NVIDIA DGX Quantum System with Arque Systems and Quantum Machines

Prof. Dr Kristel Michielsen, Director of the Jülich Supercomputing Centre (JSC) in Jülich, Germany, announced on 18 September 2025 that JSC has become the first high‑performance computing centre worldwide to host an NVIDIA DGX Quantum system. The installation couples NVIDIA’s Grace Hopper Superchip, the Quantum Machines OPX1000 hybrid quantum‑classical controller, and an Arque Systems 5‑qubit processor that couples spin qubits via electron shuttling. Together, these components enable round‑trip data transfer with a latency of fewer than four microseconds, a thousand‑fold improvement over earlier laboratory‑grade implementations.

The DGX Quantum stack is engineered to keep classical processing within the coherence window of the spin qubits, employing microsecond‑scale analog feedback loops integrated into the OPX1000 controller. Classical GPUs execute neural‑network‑based adaptive calibration and decoder optimisation in real time while the quantum processor performs core operations. The system demonstrates microsecond‑scale interaction between quantum control hardware and classical compute resources, successful execution of hybrid algorithms that combine GPU‑accelerated machine‑learning models with quantum gates, and benchmarking of quantum error‑correction protocols on the 5‑qubit Arque processor.

Implications for HPC and Quantum Research

JSC’s existing infrastructure, which hosts Europe’s fastest supercomputer JUPITER – the continent’s first exascale system and currently ranked fourth on the global list – provides the high‑throughput, low‑latency environment required for such integration. The deployment shows that quantum resources can be embedded as first‑class computational elements alongside traditional supercomputing capabilities, offering a practical pathway for embedding quantum acceleration into established HPC workflows. According to Dr Markus Beckers, CEO of Arque Systems, and Dr Itamar Sivan, CEO of Quantum Machines, the installation marks a major milestone in making quantum computing accessible within existing HPC infrastructure and signals that quantum acceleration can become as ubiquitous as GPU acceleration. NVIDIA’s Sam Stanwyck emphasised that tight integration between AI supercomputing resources and quantum control hardware is essential for achieving practical quantum error correction, underscoring the potential for hybrid quantum‑classical systems to become a standard component of high‑performance computing environments and to accelerate research in fields ranging from materials science to cryptography.