For decades the story of the late universe has been told as a slow-motion fuel shortage. Star formation peaked billions of years ago, has been declining ever since, and the tidy explanation was that galaxies simply ran low on the raw material: cold hydrogen gas. Burn through the tank, and the engine sputters.

A new measurement from China's FAST radio telescope, published Sept. 1 in Nature Astronomy, says that story does not hold up. Over the past 4.5 billion years the cosmic star-formation rate has dropped by roughly a factor of 2.5. The universe's reservoir of neutral atomic hydrogen, over the same interval, has barely moved. The gas is still there. Galaxies have just become worse at turning it into stars.

The paper, "Weak Evolution of Cosmic Atomic Hydrogen over the Past 4.5 Billion Years" (DOI 10.1038/s41550-026-02965-9), lists Chuan-Peng Zhang and Hong Guo as its first authors, part of a team led by scientists from the Chinese Academy of Sciences with roughly 50 co-authors, including the FASHI Collaboration. The work involved the National Astronomical Observatories of China, Shanghai Astronomical Observatory, Shanghai Jiao Tong University, and the DESI project. A preprint was posted to arXiv on July 6.

What was actually measured

The quantity at the center of the paper is Omega_HI, the cosmic density of neutral atomic hydrogen, the diffuse gas from which molecular clouds and eventually stars condense. Measuring it directly across cosmic time is hard, because the 21-cm radio line that atomic hydrogen emits is faint, and individual galaxies at even modest distances are too dim in that line for any single telescope to detect one by one.

The team's answer was to not try. Instead they stacked. The Dark Energy Spectroscopic Instrument (DESI) has catalogued millions of galaxies with precise positions and redshifts. Using FAST, the Five-hundred-meter Aperture Spherical radio Telescope, the researchers pointed at the sky positions of about 2.5 million DESI galaxies spread across roughly 12,000 square degrees, nearly one-third of the sky, and co-added the 21-cm signal at each galaxy's known redshift. Any single galaxy's signal is buried in noise. Add up 2.5 million of them, aligned by redshift, and the noise averages down while the hydrogen signal adds up.

The result is an average HI content for galaxies as a function of cosmic time, from the present day back to redshift 0.41, about 4.5 billion years ago. Combine that with the known number of galaxies at each epoch, and you get the cosmic HI density.

What the numbers say

The raw stacked measurement shows Omega_HI declining by a factor of 1.35 ± 0.10 from redshift 0.41 to today. After corrections, that shrinks to 1.12 ± 0.10, a change that is barely distinguishable from no change at all given the uncertainties.

Set that beside the cosmic star-formation rate density, which fell by a factor of 2.46 over the same interval, and the mismatch is stark. Star formation dropped to less than half of what it was. Atomic hydrogen dropped by about a quarter on the raw measurement, and possibly by only about a tenth once corrections are applied.

The team also checked whether individual galaxies were drying out. At fixed stellar mass, the HI gas fraction, the amount of atomic hydrogen a galaxy holds relative to its stars, evolved by less than 0.2 dex across the whole interval. Galaxies of a given size 4.5 billion years ago had roughly the same proportion of atomic gas as galaxies of that size do now.

The authors' conclusion is direct: the evolution of atomic hydrogen is "far weaker than the CSFRD decline," so gas depletion cannot be the primary driver of falling star formation.

Why the fuel-crisis explanation fails

The appeal of the depletion story was its simplicity. Stars form from gas, so if star formation is falling, the gas must be running out. But that logic only works if the bottleneck is the total amount of gas. It says nothing about how efficiently that gas is being used.

The FAST result breaks the chain at that link. If the atomic hydrogen reservoir is roughly constant while stars form at less than half the old rate, then the ratio of star formation to available atomic gas has itself fallen by roughly a factor of two. Something in the process that converts diffuse atomic hydrogen into dense, star-forming molecular hydrogen is working less well than it used to.

The paper offers two candidate mechanisms, both about flow rather than stock. The first is the supply chain from outside: galaxies are fed by gas streaming in along the filaments of the cosmic web, and the team points to that inflow weakening over time. The second is internal. As the density of gas in galaxies declines, the efficiency of converting atomic hydrogen into molecular hydrogen likely drops. Either mechanism, or both, could throttle star formation without draining the HI reservoir.

As the team put it, "the most important late-time changes may occur not in the total HI reservoir itself, but rather in the gas flow through the baryon cycle."

The baryon cycle, and why the question changes

Astronomers describe the movement of ordinary matter through a galaxy as the baryon cycle: gas falls in from the intergalactic medium, cools and settles into the disk, collapses into molecular clouds, forms stars, and is then partly ejected again by stellar winds and supernovae before, in some cases, raining back down. The total mass of atomic hydrogen is a single snapshot of one stage in that loop.

What this measurement suggests is that the snapshot is the wrong place to look. The reservoir is not the problem; the plumbing is. As ScienceDaily's coverage framed it, the research question shifts from whether the gas is depleting to why it is increasingly difficult to form stars despite abundant neutral hydrogen reserves.

That reframing matters for models. The authors describe their result as a stringent benchmark for models of gas accretion, phase conversion and star-formation regulation. Simulations of galaxy evolution that reproduce the decline in cosmic star formation by draining gas will need to reproduce it some other way, by getting the inflow rates and the HI-to-H2 conversion physics right instead. And a near-constant HI reservoir sets a hard constraint: whatever the models do, they cannot spend the atomic gas.

Why It Matters

The decline of cosmic star formation is one of the largest-scale trends in the observable universe, and until now its cause was assumed more than measured. This is the largest sample yet used to track the atomic-hydrogen budget back 4.5 billion years, and the answer contradicts the default explanation. Galaxies are not starving. They are sitting on fuel they can no longer efficiently use.

The method is as significant as the result. Stacking 2.5 million optical-survey galaxies in a radio telescope's data turns two instruments built for different purposes, DESI's spectroscopic redshift machine and FAST's enormous collecting area, into a single probe of cosmic gas. As DESI's catalog grows and FAST accumulates sky coverage, the same technique can push to higher redshift and finer time bins, and can be split by galaxy type, mass, and environment to test the two proposed mechanisms directly. Whether it is weakening inflow from the cosmic web, falling gas density, or something not yet on the list, the answer to why the universe stopped making stars is now a question about how gas moves, not how much of it is left.

Sources