RIKEN Names 15.539-EFLOPS AI Supercomputer RIKYU Ahead of July 2026 Launch
Updated
Updated · HPCwire · Jun 29
RIKEN Names 15.539-EFLOPS AI Supercomputer RIKYU Ahead of July 2026 Launch
2 articles · Updated · HPCwire · Jun 29
Summary
RIKEN said its new AI-for-science supercomputer will be called RIKYU, with full-scale operation at its Kobe campus scheduled to begin in July 2026.
15.539 EFLOPS in FP8 and 64.16 PFLOPS in FP64 are the system’s stated performance levels, delivered by 400 compute nodes using 1,600 NVIDIA Blackwell GPUs and Quantum-X800 InfiniBand links up to 3.2 terabits per second.
1,019 public naming proposals were submitted in late 2025, narrowing to 856 valid candidates before a committee chose RIKYU, a name RIKEN said evokes science, pursuit and broad international familiarity.
RIKEN expects RIKYU to work alongside Fugaku—its flagship scientific supercomputer—to support AGIS infrastructure models and wider domestic and international AI-for-science research.
What specific scientific mysteries will RIKYU solve first by collaborating with the supercomputer Fugaku?
Amid a global AI race, can Japan’s new supercomputer unite rivals for shared scientific breakthroughs?
How will a tea master's philosophy guide Japan's new AI supercomputer to redefine scientific discovery?
RIKYU and FugakuNEXT: Japan’s Strategic Leap to Exascale AI and Quantum-Driven Supercomputing
Overview
Japan is launching RIKYU, a groundbreaking supercomputing initiative that marks a new era in high-performance computing and artificial intelligence. Drawing its name from RIKEN—where 'RI' means science or logic and 'KYU' means pursuit—RIKYU reflects a strong commitment to scientific exploration and logical progress. Inspired by the tradition of naming supercomputers after influential figures, RIKYU also honors Sen no Rikyū, the famed tea master, symbolizing precision and mastery. This initiative is set to redefine Japan’s technological landscape, blending cultural depth with cutting-edge innovation to drive the nation’s leadership in AI and supercomputing.