The simulation shows us the movie of how the universe assembled itself.
In a computational act of profound ambition, Chinese scientists have reconstructed the entire lifespan of the universe — 13.8 billion years — by setting 4.2 trillion virtual dark matter particles in motion inside the ORISE supercomputer. The HyperMillennium simulation addresses a fundamental limitation of astronomy: telescopes can only capture frozen moments from the distant past, never the living arc of cosmic change. By modeling the invisible gravitational scaffolding that shapes all visible matter, researchers have built a mirror precise enough to reflect both the universe's grandest structures and its rarest extremes. The data will soon belong to the world, offered as a shared foundation for the next era of cosmic discovery.
- The universe is too vast and too old to observe in motion — so scientists have rebuilt it from scratch inside a machine, particle by particle, across all of time.
- 16,000 accelerator computing cards ran without pause for 18 days, generating 13 petabytes of data in a computational effort without precedent in scale and resolution combined.
- When tested against Abell 2744, one of the most complex galaxy clusters known, the simulation's predictions matched real telescope observations with striking accuracy — a critical validation.
- HyperMillennium now stands as the only simulation to achieve both the vast volume and fine resolution demanded by next-generation sky surveys like ESA's Euclid and the China Space Station Telescope.
- Leading cosmologists from Texas to Munich have called it a redefinition of what numerical cosmology can do — and the data is being opened to the entire global scientific community.
Inside China's ORISE supercomputer, 4.2 trillion virtual dark matter particles are clustering and colliding across a simulated space stretching 12 billion light-years — recreating the full 13.8 billion-year history of the cosmos. This is HyperMillennium, a project led by the National Astronomical Observatories of the Chinese Academy of Sciences that confronts one of astronomy's deepest frustrations: telescopes see only frozen moments from the past, never the living evolution of galaxies across time.
Dark matter — invisible, undetectable by any instrument, yet comprising roughly 85 percent of all matter — is the gravitational skeleton upon which the visible universe is built. Lead researcher Gao Liang describes ordinary matter as merely the surface of the cosmos, with dark matter forming its hidden architecture. Simulating its behavior across cosmic time is the only way to understand how galaxies were born and grew.
The computational scale is staggering. Sixteen thousand accelerator cards worked continuously for 18 days, producing approximately 13 petabytes of data. Early results, published in the Monthly Notices of the Royal Astronomical Society, focused on Abell 2744, a famously complex galaxy cluster — and the simulation matched real observations with remarkable precision.
What distinguishes HyperMillennium from its global peers is its dual ambition. Japan's Uchuu excels at fine resolution; Europe's Flagship 2 at sheer volume. HyperMillennium achieves both simultaneously — a combination essential for the next generation of sky surveys, including ESA's Euclid mission and the China Space Station Telescope. Cosmologists Volker Springel and Mike Boylan-Kolchin have each described it as redefining what is possible in the field. The simulation's data will soon be shared globally, turning a national scientific achievement into a common resource for understanding how the universe assembled itself.
Inside a Chinese supercomputer, the universe is being born again. Not metaphorically—4.2 trillion virtual particles of dark matter are clustering, colliding, and coalescing under the weight of simulated gravity, recreating 13.8 billion years of cosmic history in a digital realm. This is HyperMillennium, a cosmological simulation so vast it encompasses a cubic volume of space stretching 12 billion light-years on each side—a distance so immense that if you lined up 120,000 Milky Way galaxies end to end, you'd barely cross it.
The project, led by the National Astronomical Observatories of the Chinese Academy of Sciences, represents a fundamental shift in how scientists approach the universe's deepest mysteries. When astronomers point their telescopes at distant galaxies, they are not watching them evolve in real time. They are seeing frozen moments from billions of years ago, static images that reveal nothing about how those galaxies changed, grew, or transformed. The universe is too large and too old for direct observation. But it is not too large for calculation.
Dark matter—the invisible scaffolding that makes up roughly 85 percent of all matter in the cosmos—cannot be seen with any telescope. It emits no light, absorbs no light, interacts with nothing we can directly measure. Yet it is the gravitational skeleton upon which galaxies hang. Gao Liang, the project's lead researcher at the Academy of Sciences, explains that ordinary visible matter is merely the "visible surface" of the cosmos, while dark matter forms its hidden architecture. To understand how galaxies form and evolve, scientists must first understand dark matter's behavior across cosmic time. Simulations are the only tool that can do this. By calculating the gravitational pull among trillions of virtual particles, researchers can reconstruct the universe's evolution with precision, watching dark matter cluster and distribute itself across billions of years, then comparing those predictions against what telescopes actually observe.
The HyperMillennium simulation runs on China's ORISE supercomputer, processing 4.2 trillion particles through specially optimized software. The computational demand is staggering: 16,000 accelerator computing cards worked continuously for 18 days to complete the simulation, generating approximately 13 petabytes of data—roughly equivalent to 13 million high-definition movies. The first results, published recently in the Monthly Notices of the Royal Astronomical Society, focused on Abell 2744, a massive galaxy cluster with an extraordinarily complex structure. The simulation's predictions matched real observations with striking accuracy, validating the model's ability to capture even the rarest and most extreme cosmic systems.
Globally, three major simulations now track trillions of dark matter particles. Japan's Uchuu prioritizes resolution—the ability to see fine detail. Europe's Flagship 2 emphasizes volume—the ability to model vast regions of space. HyperMillennium does both simultaneously, combining unprecedented scale with exceptional precision. This dual strength matters because the next generation of sky surveys—including the European Space Agency's Euclid mission and the China Space Station Telescope—will observe the universe at scales and depths that demand simulations equally ambitious in scope and accuracy. Without HyperMillennium, scientists would lack a reliable computational mirror to test their theories against these new observations.
Mike Boylan-Kolchin, a cosmologist at the University of Texas, called the simulation "a computational marvel" that will unlock fundamental physics from cosmic observations. Volker Springel, director of the Max Planck Institute for Astrophysics in Germany, said it "redefines what is nowadays possible in numerical cosmology." The first batch of data will soon be made available to the global scientific community, transforming HyperMillennium from a Chinese achievement into a shared resource for understanding how the universe assembled itself across 13.8 billion years.
Notable Quotes
Cosmological simulations are the key to solving this puzzle. By calculating gravitational interactions among vast numbers of virtual dark matter particles, they can reconstruct the evolution of the universe from its early days to the present.— Gao Liang, lead researcher, National Astronomical Observatories of the Chinese Academy of Sciences
The HyperMillennium simulation will be a touchstone for the galaxy formation and cosmology communities for years to come.— Mike Boylan-Kolchin, University of Texas
The Hearth Conversation Another angle on the story
Why does it matter that we can simulate dark matter if we can't see it in the first place?
Because we can see what dark matter does. Galaxies cluster in specific patterns, they move in specific ways—all of that is shaped by dark matter's gravity. If our simulation predicts those patterns and they match what telescopes observe, then we've validated our understanding of something invisible. That's how physics works.
But couldn't you just observe the universe directly and skip the simulation?
The universe is 13.8 billion years old. The light from distant galaxies takes billions of years to reach us, so we're seeing them as they were in the past, frozen in time. We can't watch them evolve. A simulation lets us compress that evolution into days of computation.
So this is really about filling in the gaps between snapshots?
Exactly. Telescopes give us still frames from different eras. The simulation shows us the movie—how structures form, merge, and transform under gravity over cosmic time.
Why does China's version combine both scale and precision when other simulations chose one or the other?
Because the new telescopes coming online—Euclid, the China Space Station Telescope—will observe both vast regions of space and tiny, rare objects within them. You need a simulation that can handle both demands simultaneously, or your predictions won't match the observations.
What happens with all that data now that it's generated?
It gets released to scientists worldwide. The simulation becomes a testing ground. Researchers can compare their theories against it, refine their models, make new predictions about how galaxies form. It's a shared resource for understanding the universe.