How AI slashed the time to simulate our entire galaxy from decades to months
In short: For years, modeling the Milky Way star-by-star was computationally impossible. To track every individual star, traditional supercomputers would have needed 36 years to simulate just one billion years of history. Now, a new deep learning approach has broken this bottleneck, delivering a simulation 100 times faster.
The fundamental challenge in galactic modeling lies in the massive disparity of scales. To accurately capture rapid events like supernova explosions, simulations must use incredibly small time increments. Using current physics-based methods, simulating one million years of galactic evolution requires approximately 3-15 hours of computation. Under such conditions, modelling a billion years of evolution would require upwards of 36 years of actual time, a process rendered unfeasible by the rising energy demands and falling efficiency associated with increasing supercomputer cores.
Researchers from Japan's RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences (iTHEMS), headed by Keiya Hirashima and joined by collaborators from Universitat de Barcelona and The University of Tokyo, have overcome this obstacle. They developed a method that integrates a deep learning surrogate model with standard numerical simulations. This AI component was trained using high-resolution supernova simulations to predict how gas behaves and spreads during the 100,000 years following an explosion, effectively removing a major computational bottleneck.
The results, presented at the SC '25 conference, represent a massive leap in resolution. While previous state-of-the-art models were limited to a mass of about one billion suns—where a single particle represented a cluster of roughly 100 stars—this new model tracks more than 100 billion individual stars. The efficiency gain is profound: simulating one million years now takes just 2.78 hours, meaning one billion years of evolution can be completed in approximately 115 days. This breakthrough in astrophysics may also offer new strategies for modeling complex Earth systems, such as climate and weather research.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: Simulating ALL 100 billion stars in the Milky Way for the first time (with the help of AI?!)
- Also: AI creates the first 100-billion-star Milky Way simulation
- Also: The simulated Milky Way: 100 billion stars using 7 million CPU cores
To cite this card, link to https://intellectplus.co.uk/app/space-simulating-all-100-billion-stars-in-the-milky-way-for-the-fi.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
Related
More in Space · Latest · A–Z · All topics
