RIKEN has announced a partnership with Fujitsu and NVIDIA to develop Japan´s next flagship supercomputer, development codename FugakuNEXT. The three organizations will form an international initiative to design and build the system. NVIDIA will lead the design of the GPU infrastructure, marking the first adoption of GPUs as accelerators in a Japanese flagship supercomputing system, while Fujitsu will lead the basic design of the overall system, including compute nodes and CPUs.
FugakuNEXT is intended to function as a unified Artificial Intelligence-HPC platform that tightly integrates simulation and Artificial Intelligence workloads. The project will extend the CPU technologies developed for Fugaku and incorporate GPUs as accelerators to maximize execution performance for both simulation and Artificial Intelligence applications. The partners emphasize an integrated approach that aims to deliver world-leading capabilities in simulation and Artificial Intelligence rather than focusing on simulation performance alone.
Aiming for operation around 2030, the FugakuNEXT initiative stresses that sustained scientific output will depend not only on advances in hardware but also on comprehensive innovations in software and algorithms. The three organizations plan to collaborate on the broader high-performance computing infrastructure and supporting software ecosystem to enable new computational solutions. Further details about system specifications, procurement, or deployment phases are not stated in the article.
