Archive · 17 Nov 2025
AI Enables First Milky Way Simulation Tracking Over 100 Billion Stars
Imagine trying to count every single grain of sand on a beach, not just the piles, but each tiny grain.
That's what scientists try to do with stars in our galaxy, the Milky Way.
It's super hard because there are billions and billions of stars!
Older computer models could only count small piles of stars, not individual ones.
It was too slow and used too much computer power.
Now, scientists have created a super smart computer program using AI, which is like a thinking machine.
This AI learned how gas moves in space really well.
It can now watch over 100 billion stars in our galaxy, like keeping track of every single grain of sand.
It does this much, much faster than before.
This new way of simulating the galaxy is like a giant leap forward.
It helps us understand our Milky Way better.
This smart AI trick might also help scientists study other big things, like our planet's weather and climate.
Researchers have developed the first Milky Way simulation capable of tracking over 100 billion stars using AI and physics.
The new simulation is hundreds of times faster than previous models and uses 100 times more stars for greater accuracy.
The breakthrough combines deep learning with high-resolution physics, allowing the AI to learn gas behavior and predict its spread efficiently.
This advancement overcomes previous limitations in galactic modeling that struggled with the scale and computational demands of simulating individual stars.
The AI-powered method has potential applications beyond astrophysics, including climate and weather research.
- Who
- Researchers led by Keiya Hirashima from institutions in Japan and Spain
- What
- Created the first simulation of the Milky Way capable of tracking over 100 billion stars using a combination of deep learning and high-resolution physics.
- When
- The breakthrough was showcased at the SC ’25 supercomputing conference. The simulation tracks stars over a period of 10,000 years.
- Where
- RIKEN Centre for Interdisciplinary Theoretical and Mathematical Sciences (Japan) in collaboration with the University of Tokyo and Universitat de Barcelona (Spain).
- Why
- To overcome the limitations of existing galactic models that struggle to simulate the vast number of stars in the Milky Way with sufficient detail and speed.
Previous Limitations
New AI Approach
Detail and Scale of Simulation
Previous Limitations
Existing simulations could handle systems with about one billion suns, falling short of the Milky Way's more than 100 billion stars and struggling to track individual stars.
New AI Approach
The new AI simulation can track over 100 billion stars and is hundreds of times faster than older methods.
Computational Power and Time
Previous Limitations
A full star-by-star simulation would take decades due to the tiny time steps required to capture fast events, demanding immense computing power and making adding more cores impractical.
New AI Approach
The AI's trained surrogate model predicts gas behavior efficiently, allowing for the capture of small events without slowing down the main simulation, making it faster and more practical.
Representation of Stars
Previous Limitations
Each particle in older models represented nearly 100 stars, hiding individual behavior and weakening accuracy.
New AI Approach
The new simulation uses 100 times more stars than earlier work, allowing for a more accurate representation and tracking of individual stellar behavior.
- Researchers Led By
- Keiya Hirashima
- Institutions Involved
- RIKEN Centre for Interdisciplinary Theoretical and Mathematical Sciences (Japan), University of Tokyo (Japan), Universitat de Barcelona (Spain)
- Simulation Capacity
- Over 100 billion stars tracked
- Simulation Time Period
- 10,000 years
- Key Technology
- Deep learning combined with high-resolution physics
- Speed Improvement
- Hundreds of times faster than older methods
- Showcased At
- SC ’25 supercomputing conference