In the High Performance Computing Infrastructure (HPCI) program promoted by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), the “K Computer” and “Fugaku” have served as National Flagship Systems (NFS). These systems were implemented as ultra-large-scale parallel computers consisting of tens of thousands to over one hundred thousand nodes, equipped with high-performance processors based on general purpose multi-core or many-core CPUs. However, to achieve a significant improvement in performance per watt and to enable new approaches such as “AI for Science,” which advances research through active use of artificial intelligence, the capability of general purpose many-core CPUs has reached its limit. Consequently, leading supercomputers around the world now predominantly adopt accelerators such as GPUs.
Based on this background, MEXT initiated the development plan for the post-Fugaku system as the next NFS. The RIKEN Center for Computational Science (R-CCS) began system research and development for this project in FY2025. R-CCS has designated the system with the code name “FugakuNEXT,” conducting basic design in FY2025–2026, followed by detailed design beginning in FY2027, with the goal of system operation starting in FY2030.
In parallel, national university supercomputing centers that provide National Infrastructure Systems (NIS), HPCI resources other than the NFS, have already begun introducing GPU-based systems ahead of the NFS development. At several universities, GPU supercomputers have become core computational resources.
Until now, however, the program development in MEXT projects such as “Fugaku” Flagship Project has primarily focused on general purpose CPU-based applications, and there has been insufficient national support for nation-wide GPU code development. To address this issue, MEXT launched a call for proposals in July 2025 under the “Next-Generation HPC and AI Development Support Center Formation” program, aiming to establish a national hub to support code development centered on computational accelerators. The program requires that activities supporting software development for ultra-high-performance systems, particularly those using accelerators such as GPUs, and community development be conducted over a 4.5 year period starting in October 2025. This period aligns with the anticipated completion of the post-Fugaku system in 2030, and the program aims to support research on code development and performance tuning for GPU-based NIS systems prior to NFS deployment. Through these efforts, Japan aims to develop domestic code and research capabilities in HPC and AI that are competitive with GPU computing efforts in EU, USA and other countries.
Following the review process conducted by MEXT, a proposal led by the Research Organization for Information Science and Technology (RIST), with University of Tsukuba, the University of Tokyo, and Institute of Science Tokyo as core universities, was selected for the program. As a result, the Advanced HPC-AI Research and Development Support Center (HAIRDESC) was established. HAIRDESC was launched in October 2025 at Kobe Port Island and began its activities in collaboration with the three core universities.