November 18, 2025 — Leads & Copy — NVIDIA and RIKEN, Japan’s leading national research institute, are collaborating to integrate NVIDIA GB200 NVL4 systems into two new supercomputers in Japan.
One supercomputer will be dedicated to AI for science, and the other will focus on quantum computing. The RIKEN AI and quantum systems will incorporate a total of 2,140 NVIDIA Blackwell GPUs, supporting Japan’s sovereign AI strategy and expanding secure domestic infrastructure to bolster leadership in science, industry, and technology.
The first system will feature 1,600 NVIDIA Blackwell GPUs on the GB200 NVL4 platform, interconnected by NVIDIA Quantum-X800 InfiniBand networking, as part of RIKEN’s AI for science initiative. This system aims to advance research in life sciences, materials science, climate and weather forecasting, manufacturing, and laboratory automation.
The quantum computing-focused system will utilize 540 NVIDIA Blackwell GPUs, also based on the GB200 NVL4 platform and NVIDIA Quantum-X800 InfiniBand networking. This system will accelerate research in quantum algorithms, hybrid simulation, and quantum-classical computing methods.
Ian Buck, vice president of hyperscale and high-performance computing (HPC) at NVIDIA, stated that the collaboration will help Japan build a foundation for sovereign innovation, driving breakthroughs to address complex scientific and industrial challenges.
Satoshi Matsuoka, director of the RIKEN Center for Computational Science, noted that integrating the NVIDIA GB200 NVL4 accelerated computing platform will significantly advance Japan’s science infrastructure. He anticipates the partnership will create a leading unified platform for AI, quantum, and high-performance computing, facilitating accelerated discoveries across various fields.
These new RIKEN systems build on an August announcement detailing a collaboration between Fujitsu and NVIDIA to codesign a flagship supercomputer, codenamed FugakuNEXT, which is the successor to the Fugaku supercomputer. The new GPU-accelerated supercomputers will also function as proxy machines, enabling the codesigning and development of hardware, software, and applications for FugakuNEXT.
The FugakuNEXT system is projected to include FUJITSU-MONAKA-X CPUs, paired with NVIDIA technologies using NVIDIA NVLink™ Fusion. This new silicon is designed to facilitate high-bandwidth connections between Fujitsu’s CPUs and NVIDIA’s architecture. FugakuNEXT is expected to deliver a 100x performance increase compared to CPU-based supercomputers and will integrate production-level quantum computers in the future.
By combining MONAKA-X and NVIDIA’s GPUs, FugakuNEXT is expected to foster scientific discovery through innovations in HPC, AI, quantum computing, and their integration.
NVIDIA is also collaborating with RIKEN to develop floating-point emulation software that leverages NVIDIA Tensor Core GPU performance to accelerate scientific computing. This technology will enable applications to utilize the power of GPUs for AI and HPC at RIKEN and other supercomputing centers worldwide.
RIKEN also plans to use NVIDIA CUDA-X™ to enhance its HPC applications with GPU platforms, advancing AI and quantum computing initiatives in Japan. NVIDIA CUDA-X provides over 400 optimized GPU-accelerated libraries, microservices, and tools.
The new RIKEN supercomputers are slated to be operational in spring 2026, with FugakuNEXT targeted for operation by 2030.
Alex Shapiro
NVIDIA Public Relations
1-415-608-5044
ashapiro@nvidia.com
Source: NVIDIA
