Workshop: Sovereign AI Supercomputing Cloud Workshop
Authors: Toshihiro Hanawa (The University of Tokyo)
Abstract: In December 2025, the Japanese government adopted its first AI Basic Plan under the AI Promotion Act, positioning AI research and development, including the computing infrastructure behind it, as a national priority. In line with this, MEXT is promoting AI for Science (AI4S) and strengthening Japan’s computing infrastructure for it through two major initiatives. First, the RIKEN Center for Computational Science (R-CCS) introduced RIKYU in 2026, a supercomputer for AI4S development with 1,600 NVIDIA Blackwell GPUs. RIKYU complements Fugaku in developing and applying scientific foundation models, and its resources are also available to MEXT’s AI4S research programs, including ARiSE and SPReAD1000. Second, MEXT’s FY2025 supplementary budget supports four university centers that provide National Infrastructure Systems (NIS) in HPCI, the framework for sharing computing resources across Japanese universities and research institutes, to expand shared GPU resources and make their use more efficient for AI4S. The four centers at Hokkaido Univ., Univ. of Tsukuba, UTokyo, and Kyushu Univ. have agreed to share operational know-how and standardize practices, and UTokyo will serve as an HPCI hub that links jobs across resource-providing centers. This talk presents both efforts and discusses how national infrastructure for AI4S can be operated and shared across institutions.
Back to Sovereign AI Supercomputing Cloud Workshop Archive Listing
Back to Full Workshop Archive Listing