RIKEN, Japan’s national research institute, is deploying two new supercomputers powered by NVIDIA GB200 NVL4 systems to advance both AI for science and quantum computing initiatives. The facilities will integrate a total of 2,140 NVIDIA Blackwell GPUs, interconnected via NVIDIA Quantum-X800 InfiniBand networking, with the AI system featuring 1,600 GPUs and the quantum system utilizing 540. This partnership, announced November 17, 2025, establishes a unified platform for accelerated research across life sciences, materials science, climate modeling, and quantum algorithm development, bolstering Japan’s sovereign AI strategy and domestic high-performance computing infrastructure.
RIKEN and NVIDIA Advance Japan’s Supercomputing Infrastructure
RIKEN, Japan’s premier research institute, is significantly upgrading its supercomputing infrastructure through a partnership with NVIDIA. Two new systems, slated for operation in spring 2026, will integrate 2,140 NVIDIA Blackwell GPUs—1,600 for AI-driven scientific discovery and 540 dedicated to quantum computing. These systems leverage the GB200 NVL4 platform and NVIDIA Quantum-X800 InfiniBand networking to accelerate research across vital fields like life sciences, materials science, and climate modeling.
This investment isn’t isolated; it’s part of a broader strategy including the development of FugakuNEXT, a next-generation supercomputer aiming for 100x performance gains. FugakuNEXT will combine Fujitsu’s MONAKA-X CPUs with NVIDIA GPUs via NVLink™ Fusion, and even incorporate future quantum computing capabilities. Importantly, these new machines will also serve as “proxy” platforms for codesigning hardware and software for FugakuNEXT, streamlining the development process.
Beyond the hardware, NVIDIA and RIKEN are collaborating on software advancements. This includes floating point emulation to maximize GPU performance for traditional scientific workloads, and utilizing the CUDA-X suite—over 400 optimized GPU-accelerated libraries—to enhance existing HPC applications. This holistic approach underscores Japan’s commitment to leadership in AI, quantum computing, and high-performance computing, securing its position at the forefront of scientific innovation.
New Supercomputers for AI, Quantum, and HPC Research
RIKEN, Japan’s leading research institute, is deploying two new supercomputers powered by NVIDIA’s cutting-edge GB200 NVL4 systems. These machines—one for AI-driven science and another for quantum computing—will boast a combined total of 2,140 NVIDIA Blackwell GPUs. Interconnection utilizes NVIDIA Quantum-X800 InfiniBand networking, promising substantial boosts to research in fields like life sciences, materials science, and climate modeling. This investment solidifies Japan’s position in the global landscape of advanced scientific computing.
Beyond immediate research gains, these supercomputers serve as crucial “proxy machines” for codesigning and developing applications for FugakuNEXT – the successor to Japan’s renowned Fugaku supercomputer. FugakuNEXT will combine FUJITSU-MONAKA-X CPUs with NVIDIA GPUs via NVLink™ Fusion, targeting a 100x performance increase. This collaborative approach aims to integrate future quantum computers, establishing a unified platform for HPC, AI, and quantum computing innovation.
NVIDIA and RIKEN are also actively developing floating-point emulation software to fully leverage NVIDIA Tensor Core GPU performance for traditional scientific workloads. Combined with access to NVIDIA’s CUDA-X library – a suite of 400+ optimized GPU-accelerated tools – these advancements promise to accelerate discoveries in both AI and quantum computing. Both new RIKEN systems are planned to be operational in spring 2026, with FugakuNEXT following in 2030.
FugakuNEXT: Next-Generation Supercomputer Development
RIKEN, Japan’s leading research institute, is dramatically upgrading its supercomputing infrastructure with two new systems powered by 2,140 NVIDIA Blackwell GPUs. These aren’t replacements, but proxy machines designed to accelerate development for “FugakuNEXT,” the successor to the world-renowned Fugaku supercomputer. The new systems will focus on both AI for science and quantum computing, leveraging NVIDIA’s GB200 NVL4 platform and Quantum-X800 InfiniBand networking to tackle complex challenges in fields like life sciences and materials science.
FugakuNEXT is projected to deliver a staggering 100x performance increase over current systems, achieved by pairing FUJITSU-MONAKA-X CPUs with NVIDIA GPUs via NVLink Fusion. This combination aims to unlock breakthroughs across HPC, AI, and quantum computing—eventually integrating production-level quantum computers. RIKEN’s strategy isn’t solely hardware-focused; they’re also utilizing NVIDIA CUDA-X libraries and developing floating-point emulation software to maximize GPU performance for traditional scientific workloads.
Operational in spring 2026 (the proxy machines) with FugakuNEXT following in 2030, this initiative solidifies Japan’s commitment to sovereign AI and scientific innovation. By strategically using these new systems to codesign and develop for its next-generation flagship, RIKEN aims to establish a unified platform for pushing the boundaries of discovery—from climate modeling to quantum algorithm development—and maintain its position at the forefront of global scientific research.
Software and Technologies Accelerate Scientific Discovery
Japan is poised to become a leader in AI-driven scientific discovery thanks to a major investment in supercomputing infrastructure. RIKEN, Japan’s leading research institute, is deploying two new supercomputers powered by 2,140 NVIDIA Blackwell GPUs. These systems, utilizing the GB200 NVL4 platform and NVIDIA Quantum-X800 InfiniBand networking, will accelerate research across diverse fields – from life sciences and materials science to climate modeling and quantum computing – bolstering Japan’s sovereign AI strategy.
Beyond immediate computational power, RIKEN is integrating NVIDIA software to unlock even greater performance. Development of floating-point emulation software leveraging NVIDIA Tensor Cores will optimize traditional scientific applications for GPU acceleration. Utilizing the CUDA-X suite – boasting over 400 optimized libraries – will further boost HPC applications, specifically within AI and quantum computing initiatives. These combined efforts promise substantial gains in research efficiency.
This investment extends to the future with the development of FugakuNEXT, a successor to the renowned Fugaku supercomputer. Combining FUJITSU-MONAKA-X CPUs with NVIDIA GPUs via NVLink™ Fusion, FugakuNEXT is projected to deliver 100x greater application performance. Planned for operation by 2030, it will also integrate production-level quantum computers, creating a unified platform for HPC, AI, and quantum computing – cementing Japan’s commitment to innovation in these critical fields.
