The Advanced General Intelligence for Science Program (AGIS), Transformative Research Innovation Platform of RIKEN platforms (TRIP) Headquarters, and RIKEN Center for Computational Science (R-CCS) have selected "RIKYU" as the name for their new AI for Science development supercomputer.

According to the selection committee, the name was chosen “because it reflects a computing platform designed to uncover the fundamental principles and laws governing nature and using AI and supercomputing, and to support a wide range of research from basic science to applied studies. The name was also considered to be well aligned with the purposes of AI for Science. In addition, its pronunciation resembles that of RIKEN, making it easy to pronounce and familiar to both Japanese and international users.” The two characters used in the name are both found in the Japanese name of RIKEN. The first, RI (理), means science or logic, and the second, KYU (究), means pursuit. Additionally, following the international trend of naming supercomputers after people, the name sounds like Sen no Rikyu, the 16^th century tea master who is known for having perfected the art of the tea ceremony.

In the Rikyu Doka, a collection of poems that are said to convey the teachings of Sen no Rikyu, there is a waka poem describing the well-known concept of shuhari (follow, break, transcend), which describes a Japanese philosophy of the stages of learning. The poem states: “follow the rules and forms completely; even if you break them, even if you transcend them, never forget the foundation”. Following what one has learned from a master is akin to AI extensively learning existing knowledge through machine learning. Breaking knowledge by re-examining what has been mastered in comparison with other insights and one’s own perspective, is analogous to the trained AI actively engaging in research and pioneering new knowledge. Transcending existing frameworks reflects a future in which humanity and AI jointly pioneer new frontiers in science through the broad exploratory scope of AI for Science. With this philosophy in mind, and while pursuing wide-ranging exploration, we have entrusted the name "RIKYU" with the meaning of “transcending while never forgetting the true purpose of science.”

RIKYU is a computing system being introduced at the RIKEN Kobe Campus on Port Island in Kobe, with the aim of innovating scientific research through AI. It delivers exascale-class computational performance, 10¹⁸ operations per second, using the 16-bit and 8-bit floating-point operations that are widely used in AI training and inference. RIKYU is currently undergoing the necessary preparations for full-scale operation, scheduled to begin in July 2026.

To advance AI for Science in the future, we will establish a user support framework that enables the system to be widely utilized not only for the development of scientific research infrastructure models under the AGIS program but also for AI for Science research through domestic and international collaboration.

RIKYU comprises 400 compute nodes equipped with the NVIDIA GB200 NVL4 (1,600 NVIDIA Blackwell GPUs in total). The nodes are interconnected by NVIDIA Quantum-X800 InfiniBand networking, enabling high-speed communication of up to 3.2 terabits per second (1 terabit = 10¹² bits per second). It delivers performance of over 64.16 PFLOPS (10¹⁵ operations per second) in double-precision floating-point arithmetic (FP64) and over 15.539 EFLOPS (10¹⁸ operations per second) in 8-bit floating-point arithmetic (FP8).

The collaboration between RIKYU, which specializes in massively parallel computing for AI, and Supercomputer Fugaku, which specializes in large-scale scientific computing, is expected to contribute to the development and utilization of advanced scientific research infrastructure models.

The naming process

The centers launched a public call for naming proposals for the new “AI for Science Development Supercomputer,” based on the following naming criteria: (1) it should convey that the supercomputer is used for research that benefits society; (2) it should be familiar and appealing to people in Japan and around the world; (3) it should contain or evoke an element related to science; and (4) it should evoke the image of AI assisting scientists in their research.

A total of 1,019 proposals were received during the submission period from November 20 to December 22, 2025. When duplicate submissions and identical-name entries were removed, a total of 856 valid candidates remained.

Subsequently, the entries were considered by the AI for Science Development Supercomputer Naming Committee, which was composed of (in alphabetical order) Yutaka Akiyama (Senior Visiting Scientist, RIKEN Center for Computational Science), William Dawson (Unit Leader, Materials Science Application Interface Platform Development Unit, AI for Science Platform Division, RIKEN Center for Computational Science), Koichi Takahashi (Team Director, Laboratory for Biologically Inspired Computing, RIKEN Center for Biosystems Dynamics Research/ Project Director, AGIS, TRIP Headquarters), Shuntaro Tani (RIKEN ECL Team Leader, Digital Twin for Light-Matter Interaction RIKEN ECL Research Team, RIKEN Center for Advanced Photonics), and Toshiji Hase (External Member, Science Fiction Writer), with Makoto Taiji (Director, AGIS, TRIP Headquarters) and Satoshi Matsuoka (Director, RIKEN Center for Computational Science/Division Director, AI for Science Platform Division, RIKEN Center for Computational Science/ Project Director, AGIS, TRIP Headquarters) acting as observers. The Committee selected the name RIKYU taking into account input from researchers involved in AGIS.

The entry submitted under this name was from Mr. Seizo Kawai, to whom we plan to send a commemorative gift as a token of our appreciation.

Contact

Inquiry contact

Department: AGIS / TRIP Headquarters, RIKEN
Email: agis_pr@ml.riken.jp

Institutional Contact Point

Department: Communication Division, RIKEN
Contact form