RIKEN Activates 2 Supercomputers With 2,140 Blackwell GPUs for AI and Quantum Research
Updated
Updated · NVIDIA Blog · Jul 16
RIKEN Activates 2 Supercomputers With 2,140 Blackwell GPUs for AI and Quantum Research
2 articles · Updated · NVIDIA Blog · Jul 16
Summary
RIKEN has begun operating two new systems: RIKYU for AI-for-science work with 1,600 NVIDIA Blackwell GPUs, and ROQUO for quantum-HPC workloads with 540 Blackwell GPUs.
RIKYU will support development of open foundation models and AI adoption in life sciences, materials science and laboratory automation, while ROQUO links accelerated computing with on-premises quantum machines in Wako and Kobe.
ROQUO is already being used to test an evolutionary AI framework integrated with NVIDIA CUDA-Q to generate quantum circuits for Quantinuum’s trapped-ion Reimei system.
The launch sits within a broader Japan-U.S. push on AI and quantum science, with NVIDIA also working with AIST, Fujitsu and others on hybrid quantum-GPU systems and reporting a 13.4x speedup in a quantum chemistry workflow.
Is Japan's AI boom creating true independence or just deep reliance on NVIDIA?
As physical AI enters Japanese workplaces, what is the future for human labor?
Japan Unveils ROQUO and RIKYU: Next-Gen AI & Quantum Supercomputers Powering the Road to FugakuNEXT
Overview
In Spring 2026, RIKEN activated two powerful GPU-accelerated supercomputers, ROQUO and RIKYU, marking a major step for Japan’s advanced computing. These systems are built for high-performance parallel processing, which is essential for modern AI workloads and complex simulations. Their main purpose is to drive AI-driven scientific research and lay the foundation for future quantum-classical hybrid computing. ROQUO and RIKYU also serve as 'proxy machines' for the development of FugakuNEXT, Japan’s next flagship supercomputer, by providing platforms for codesigning and developing the hardware, software, and applications needed for this ambitious national project.