The world’s most powerful supercomputers can now run simulations of billions of neurons, and researchers hope such models will offer unprecedented insights into how our brains work

Scientists at Germany’s Jülich Research Centre are preparing to run one of the most ambitious brain simulations ever attempted using JUPITER, currently the world’s fourth most powerful supercomputer[1]. The simulation aims to model 20 billion neurons and 100 trillion connections - equivalent to the human cerebral cortex[1:1].

Led by neurophysics professor Markus Diesmann, the team demonstrated in late 2025 that their “spiking neural network” model could be scaled up to run on JUPITER’s thousands of graphical processing units[1:2][2]. “We know now that large networks can do qualitatively different things than small ones,” said Diesmann[1:3].

The simulation could help researchers:

  • Test theories of brain functionality impossible to study with real brains
  • Investigate how memories form
  • Model diseases like epilepsy and test potential treatments
  • Study learning processes at accelerated speeds[1:4]

However, University of Sussex professor Thomas Nowotny cautions about the limitations: “We can’t actually build brains. Even if we can make simulations of the size of a brain, we can’t make simulations of the brain.”[1:5]

This effort builds on recent progress in brain mapping, including the 2024 completion of the first fruit fly brain circuit map[3]. The JUPITER simulation represents a significant advance over previous attempts like the Human Brain Project, which struggled to achieve similar goals despite substantial funding[3:1].

  1. New Scientist - We’re about to simulate a human brain on a supercomputer ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎

  2. The Conversation - A new supercomputer aims to closely mimic the human brain ↩︎

  3. Futurism - Scientists Preparing to Simulate Human Brain on Supercomputer ↩︎ ↩︎

  • woodenghost [comrade/them]@hexbear.net
    14·
    2 days ago

    “We can’t actually build brains. Even if we can make simulations of the size of a brain, we can’t make simulations of the brain.”

    Seems like an important distinction. Still cool though.

    • originalucifer@moist.catsweat.com
      5·
      2 days ago

      kinda feels like a really good first step though… virtually replicate all the parts you can until you can reverse engineer the software to run on it

    • Xavienth@lemmygrad.ml
      3·
      2 days ago

      I imagine one of the most important factors is we don’t have a way to accurately scan a brain to correctly reproduce all those trillions of connections. We don’t have a blueprint to go off. Or if we do it’s probably prohibitive in scale to collect that.