DATE2026.09.28 #Awards & Prizes
Milky Way Simulation Utilizing Exascale Supercomputers and AI Selected as a Finalist for the 2026 Gordon Bell Prize
Disclaimer: machine translated by DeepL which may contain errors.
Led by Professor Michiko Fujii and Visiting Researcher Keiya Hirashima (RIKEN Center for Mathematical Creativity) of the Department of Astronomy, Graduate School of Science, The University of Tokyo, and involving Assistant Professor Kana Moriwakiand Naoto Harada, a Project Researcher has been selected as a finalist for the 2026 ACM Gordon Bell Prize for its research on simulations of the Milky Way Galaxy (Note 1).
The research team conducted numerical simulations of the Milky Way galaxy model at the world’s highest resolution, using up to 5 trillion particles, and achieved high computational performance on three exascale supercomputers (Note 2) with different configurations. In addition, the team accelerated computations using a new method that combines existing techniques with AI-based surrogate models (Note 3).
The ACM Gordon Bell Prize is awarded for achievements in the field of high-performance computing that involve the highly efficient execution of scientifically significant computations, and is one of the most prestigious awards in this field. The 2026 award results will be announced at SC26 ( International Conference for High-Performance Computing, Networking, Storage, and Analysis ), to be held in Chicago, Illinois, USA, from November 15 to 20.
Visualization of the Milky Way simulation
(Credit: Keiya Hirashima)
Related Links
High-Resolution Galaxy Simulations Enabled by AI
The Spread of a Supernova Explosion as Depicted by AI
Glossary
Note 1: The Milky Way Galaxy
The galaxy in which our solar system resides. A galaxy is a celestial object in which numerous stars and interstellar gas are clustered within dark matter and bound together by mutual gravitational forces.
Note 2: Exascale Supercomputer
A supercomputer capable of performing 1 exa (10¹⁸) floating-point operations per second.
Note 3: Surrogate Model
A model that uses machine learning and other techniques to provide approximate results in lieu of experiments or numerical simulations.