Deep inside a vast computational complex on an artificial island in Kobe, Japan, rows of towering black cabinets hum steadily, consuming enormous amounts of electricity and processing power. This is Fugaku, one of the fastest supercomputers on Earth — and it is here that scientists have achieved something once considered science fiction: a fully functioning digital simulation of an entire cerebral cortex.
In a newly published peer-reviewed study in the ACM International Supercomputing Conference proceedings, researchers demonstrated that it is now technically possible to reproduce a complete mammalian cerebral cortex — down to the electrical behavior of individual neurons — inside a computer. The achievement marks a major milestone in neuroscience, computational biology, and artificial intelligence, with implications that stretch from treating brain disorders to probing the deepest mysteries of consciousness itself.
“This is much more than just an animation,” said Anton Arkhipov, an investigator at the Allen Institute and co-author of the study. “The key point is demonstrating that it is technically feasible to reproduce the cerebral cortex of a mouse at this spatiotemporal resolution.”
A Brain Rebuilt in Silicon
The digital model represents a mouse cerebral cortex composed of roughly ten million neurons connected by billions of synapses. Electrical signals ripple through the simulated tissue exactly as they do in living brains, governed by the same biophysical rules that shape real neural activity. Researchers can pause the simulation, rewind it, zoom into individual synapses, rewire connections, and run alternative scenarios — all without touching a single living animal.
“It’s like having a slow-motion video of thought itself,” one researcher involved in the visualization work explained.
The reconstruction was made possible using ultra-detailed biological maps provided by the Allen Institute, which allowed scientists to rebuild the cortex layer by layer and cell type by cell type. Researchers at the University of Illinois at Urbana-Champaign then developed advanced visualizations that allow scientists to observe neural activity across 86 interconnected brain regions in real time.
Powered by Fugaku’s ability to perform approximately 400 quadrillion calculations per second, the simulation doesn’t collapse into chaos or silence — a common problem in simpler models. Instead, it stabilizes into rhythms that closely resemble those measured in real mouse brains.
“That biological fidelity is the breakthrough,” Arkhipov said. “Smaller simulations can sometimes reproduce similar patterns for the wrong reasons. What we’ve shown is that this model behaves like a real brain for the right reasons.”
Beyond Scale: Why Biology Matters
While the sheer size of the simulation is staggering, researchers stress that scale alone is not the achievement. What matters is that the model obeys the same physical laws as biological brains. Neurons fire, transmit signals, and interact based on real anatomical and electrical properties derived from living tissue.
“This is not a metaphorical brain or a simplified abstraction,” Arkhipov explained. “It is a biologically realistic simulation.”
That realism makes the model uniquely powerful as a scientific tool. Because researchers can manipulate it freely, they can test hypotheses that would be impossible, unethical, or impractical in live animals.
A New Weapon Against Brain Disease
One of the most immediate applications of the digital cortex lies in medicine. Brain disorders such as Alzheimer’s disease, epilepsy, autism, and schizophrenia often involve subtle changes that occur long before symptoms appear.
“Imagine certain components of the cortical network starting to change early in disease,” Arkhipov said. “Maybe specific cell types disappear, or connectivity patterns are altered. We can implement those changes in the simulation and ask what effect they have.”
Tiny shifts that might never be detectable in living brains can be amplified and examined in the digital model, allowing scientists to identify which changes actually matter — and which might serve as early targets for intervention or treatment.
For now, Arkhipov emphasizes, the project is pragmatic and mechanistic. “The core purpose is understanding disease,” he said. “But the same tools could eventually help us explore much bigger questions.”
Could a Machine Become Conscious?
Those bigger questions push the work into philosophical territory. Because the simulation already reproduces neural activity patterns associated with perception and cognition, some researchers believe future versions could one day help explain how consciousness itself emerges.
“The question of where awareness comes from is very deep,” Arkhipov said. “Our project can contribute in a big way to understanding the associated mechanisms.”
Future models may be able to sustain internal neural activity even without external input — a defining feature of conscious systems. If such a network maintained memory, momentum, and self-generated dynamics, would it still require a biological body?
Arkhipov doesn’t think so.
“All the phenomena involved are physical processes,” he said. “I’m not aware of any law of nature that requires them to arise only in biological systems. I would think it is entirely possible for a piece of hardware to be a thinking, feeling entity.”
Skepticism From Neuroscience
Not everyone agrees.
Peter Coppola, a visiting scholar at the University of Cambridge, argues that consciousness cannot be inferred from neural activity alone.
“We do not have a conclusive measure of consciousness,” Coppola said. “No test can tell us that a system is experiencing something. Even behavioral observation can fail.”
Coppola also questions whether a cortex alone — digital or biological — is sufficient for conscious experience. “It’s hard to imagine a truly accurate model of a brain that lacks a subcortex and a body,” he said, noting evidence that experience can persist without the most advanced cortical structures.
He also points to missing biological features in the current simulation, including neural plasticity — the brain’s ability to rewire itself — and neuromodulation, the chemical systems that fine-tune neural activity.
“All models are wrong, but some are useful,” Coppola said, quoting statistician George E. P. Box. “This is a remarkable model — but it may still miss the mechanisms that matter most for consciousness.”
A Bridge to the Unknown
Arkhipov does not dismiss these criticisms. Instead, he sees the project as a bridge — one that connects present-day neuroscience to future discoveries that may reshape how humans understand minds, machines, and themselves.
Whether digital brains will ever become conscious remains unknown. But even if they never cross that threshold, the ability to see every neuron, every synapse, and every signal in action represents a revolutionary tool.
“In animals, we see tiny windows into neural activity,” Arkhipov said. “In the simulation, nothing is hidden.”
For now, the digital mouse brain remains firmly rooted in science rather than speculation — studying disease, testing treatments, and illuminating the physical basis of thought. But as computing power grows and biological detail deepens, the line between simulation and mind may become increasingly difficult to define.
And if consciousness ever does emerge from silicon, Arkhipov believes, it would be nothing short of extraordinary.
