Skip to content

Japan's upcoming supercomputer, FugakuNEXT, is set to be powered by Nvidia GPUs and Fujitsu Arm CPUs, with a projected cost of $750 million. This machine aims to transform AI-driven science and global research.

Japan allocates over $750 million for FugakuNEXT, a zetta-scale supercomputer created by RIKEN and Fujitsu. Equipped with FUJITSU-MONAKA3 CPUs and advanced accelerators, the machine will incorporate AI into scientific studies, aiming to produce a performance 1,000 times greater than current models.

Nvidia graphics processors and Fujitsu Arm central processing units will be the driving force...
Nvidia graphics processors and Fujitsu Arm central processing units will be the driving force behind Japan's upcoming $750 million supercomputer, FugakuNEXT. The goal of this ambitious project is to redefine AI-driven science and research on a global scale.

Japan's upcoming supercomputer, FugakuNEXT, is set to be powered by Nvidia GPUs and Fujitsu Arm CPUs, with a projected cost of $750 million. This machine aims to transform AI-driven science and global research.

In a significant move towards global leadership in AI and High Performance Computing (HPC) innovation, Japan is collaborating with RIKEN, Fujitsu, and Nvidia to develop FugakuNEXT. This ambitious project, if successful, could redefine the role of national supercomputers in shaping both science and society.

RIKEN, a prominent Japanese research institute, has set ambitious goals for FugakuNEXT. The project aims to achieve a fivefold increase in hardware performance and a twentyfold improvement in software and algorithmic aspects. The goal is to surpass exascale and venture into zetta-scale territory, a significant leap in computational power.

At the heart of FugakuNEXT is a hybrid design, integrating GPUs as a central element. Fujitsu will develop the new MONAKA-X CPUs for the project, while Nvidia will supply GPUs and co-design the interconnect fabric through NVLink Fusion. Nvidia's software stack will also be fully integrated into FugakuNEXT.

The development budget for FugakuNEXT exceeds 110 billion Yen ($740 million USD), and the project is expected to cost more than 375 billion Yen ($2.5 billion USD) in total. The government of Japan views this as a strategic investment to strengthen domestic semiconductor technologies.

Satoshi Matsuoka, Director of RIKEN's Center for Computational Science (R-CCS), considers the partnership with Nvidia a major strategic move. Matsuoka frames FugakuNEXT as an integrated AI-HPC platform and a model for "AI for Science," aiming to accelerate discovery across diverse fields.

FugakuNEXT is intended to serve as a national platform for research and industry. It is projected to have a peak of around 600 exaFLOPS in FP8 sparse precision and is expected to come online around 2030 at RIKEN's Kobe campus.

Japan aims to position itself as a global leader in AI and HPC innovation through FugakuNEXT. If successful, FugakuNEXT could establish Japan as a central force in the zetta-scale era, pioneering the integration of HPC and AI at unprecedented scale.

For updates on FugakuNEXT and related news, we recommend following Tom's Hardware on Google News.

Read also: