RIKEN Activates ROQUO Supercomputer Integrating Quantinuum’s Reimei System and NVIDIA Blackwell Cluster

RIKEN Activates ROQUO Supercomputer Integrating Quantinuum’s Reimei System and NVIDIA Blackwell Cluster Japan’s flagship research institute, RIKEN, has initiated operations for ROQUO, a new hybrid Quantum-HPC supercomputer located at the RIKEN Center for Computational Science (R-CCS) in Kobe. Constructed under the Japanese government’s JHPC-quantum project, ROQUO connects directly to on-premises quantum hardware—including Quantinuum’s trapped-ion Reimei system installed at RIKEN’s campus—and is linked to Japan’s flagship Fugaku supercomputer. The deployment represents a major milestone for the U.S.-Japan Genesis Mission partnership, marking the launch of a joint operational framework that combines GPU-accelerated classical computing, agentic AI, and physical quantum processing units (QPUs). [ ROQUO Hybrid System Architecture ] Location & Lead ──► RIKEN Center for Computational Science (R-CCS, Kobe, Japan). Classical Backends ──► 540 NVIDIA Blackwell GPUs across 135 GB200 NVL4 compute nodes. Measured HPC Speed ──► 19.80 PFLOPS FP64 double-precision (High Performance LINPACK). Quantum Backends ──► Quantinuum Reimei (Trapped-ion) & IBM Quantum System Two (ibm_kobe). Interconnect & OS ──► NVIDIA Quantum-X800 InfiniBand & SQC Interface (CUDA-Q).

Tightly Coupling Trapped-Ion QPUs with Evolutionary AI Frameworks ROQUO is among the world’s first full-scale operational platforms to deploy the NVIDIA GB200 NVL4 architecture. Unlike GPU configurations optimized purely for large-scale LLM training, the NVL4 platform balances FP64 floating-point performance with fast communication bandwidth, making it suited for scientific HPC matrix calculations and real-time quantum error correction workloads. During its initial operational phase, the multi-institutional research team is focusing on several key technical tracks: Evolutionary AI Circuit Synthesis: Researchers from RIKEN, Quantinuum, and NVIDIA are utilizing an evolutionary AI framework—built on top of the NVIDIA CUDA-Q platform—to automatically generate and optimize native quantum circuits for execution on the Reimei trapped-ion QPU. Low-Latency Interconnect Integration: Operating via RIKEN’s software-based SQC Interface, ROQUO manages data synchronization between classical GPU arrays, classical supercomputers (Fugaku), and quantum backends.

Quantum Error Correction & Calibration: NVIDIA is deploying its open Ising AI models to assist with automated QPU calibration, state preparation, and real-time error-decoding algorithms across attached quantum hardware.

Accelerating Industrial Quantum Chemistry Workloads In addition to system-level integration, corporate and academic partners—including Mitsubishi Chemical, Mizuho Bank, Keio University, AIST, and the University of Toronto—are leveraging the converged infrastructure to benchmark quantum chemistry applications. Initial demonstrations centered on molecular spectral analysis achieved a 13.4x speedup over CPU-only baselines. The accelerated molecular workflows target early industrial applications in semiconductor manufacturing—such as evaluating extreme ultraviolet (EUV) photoresist compounds—and materials design for next-generation energy storage. By establishing an operational link between Quantinuum’s trapped-ion processors, NVIDIA’s Blackwell hardware, and RIKEN’s supercomputing infrastructure, ROQUO serves as an active testbed for scaling “AI for Science” frameworks toward large-scale, fault-tolerant quantum computing. Review the official system activation announcement via the RIKEN Center for Computational Science Portal here, examine international partnership details on the NVIDIA Blog here, and track hardware integration profiles on Quantinuum’s post here. For our previous coverage on the supercomputer’s hardware architecture and cooling deployment, read the initial ROQUO System Launch Report on QCR here. July 22, 2026 Mohamed Abdel-Kareem2026-07-22T21:49:26-07:00 Leave A Comment Cancel replyComment Type in the text displayed above Δ This site uses Akismet to reduce spam. Learn how your comment data is processed.