Every Method of Measuring the Universe’s Expansion Rate Agrees, and That’s a Problem

Two numbers have been circling each other for more than a decade, refusing to converge. One comes from the very young universe: the faint glow of microwaves left over from roughly 380,000 years after the Big Bang, carefully decoded to predict how fast space should be stretching today. That number lands at about 67 kilometres per second per megaparsec. The other comes from the universe we can actually see around us, measured by tracking how quickly nearby galaxies are flying apart. That number keeps landing around 73. The gap is small in absolute terms, barely 9 percent, but in a field where uncertainties have been whittled down to single-digit percentages, it is enormous. And it will not go away.

Now an international team of 37 astronomers has thrown everything they have at the problem. Their conclusion, published this week in Astronomy & Astrophysics, is both reassuringly precise and profoundly unsettling.

The group, calling itself the H0 Distance Network Collaboration, assembled at the International Space Science Institute in Bern, Switzerland, in March 2025 with a single goal: combine every credible method of measuring cosmic distances into one unified framework and see what falls out. The result is a Hubble constant of 73.50 plus or minus 0.81 kilometres per second per megaparsec, a precision of just over one per cent. It is the most accurate direct measurement of the local expansion rate ever produced. And it sits 7.1 standard deviations away from the value predicted by the standard cosmological model calibrated to the cosmic microwave background. In statistics, five sigma is conventionally treated as a discovery threshold. This is well past it.

What makes the new result particularly hard to dismiss is not just the number itself but how it was built. Previous measurements of the Hubble constant have tended to rely on a single chain of overlapping techniques, a so-called distance ladder. You measure nearby stars with one method, use those to calibrate brighter objects further out, and eventually reach galaxies distant enough to sit in the smooth expansion of space. The problem, critics have argued, is that any unrecognised error in one rung could contaminate everything above it.

The H0DN collaboration took a different approach. Instead of a ladder, they constructed what they call a distance network, linking roughly a dozen different measurement techniques into a web of cross-checks. Pulsating Cepheid stars. The sharp cutoff in brightness when red giant stars ignite helium in their cores (a feature astronomers call the tip of the red giant branch, or TRGB). Mira variables. Carbon-rich stars on the asymptotic giant branch. Type Ia supernovae. Type II supernovae. Surface brightness fluctuations in elliptical galaxies. The fundamental plane relation. The Tully-Fisher relation. Even geometric distances from water masers orbiting supermassive black holes. Each technique has its own quirks, its own systematic uncertainties, its own vocal advocates and sceptics. By weaving them together with a covariance-weighted statistical framework, the team could check whether any single method was dragging the result in an odd direction.

None was. When the group removed Cepheids from the analysis entirely, the central value barely shifted; the apparent drop turned out to be driven by losing twenty supernova calibrators, not the Cepheids themselves. Taking out the tip of the red giant branch changed almost nothing. Swapping one supernova dataset for another changed almost nothing. Even splitting the entire network into two fully independent paths that shared no data whatsoever produced answers that differed by a trivial 0.17 sigma. The internal consistency is, frankly, remarkable.

“This work effectively rules out explanations of the Hubble tension that rely on a single overlooked error in local distance measurements,” the collaboration writes. “If the tension is real, as the growing body of evidence suggests, it may point to new physics beyond the standard cosmological model.”

Before anyone ran a single calculation, the team held anonymous votes to decide which methods were mature enough for inclusion in the baseline analysis and which should be tested as variants. This was deliberate. By locking in the rules before seeing the outcome, they avoided the kind of post hoc tinkering that can (even unconsciously) nudge a result toward a preferred answer. Around forty specialists in Cepheids, supernovae, surface brightness fluctuations, masers, and cosmological theory participated in the process.

Not everyone in the field was at the table, though. Members and leaders of the Carnegie-Chicago Hubble Program, one of the most prominent groups working on local distance measurements, were invited but declined to participate. Their published results, which have sometimes suggested a somewhat lower Hubble constant when relying primarily on TRGB measurements, were still included in the analysis wherever they met the agreed quality criteria. It is worth noting that when the H0DN framework uses the same calibrator sample and methods as the Carnegie-Chicago group, it reproduces their published values closely; the remaining differences trace largely to sample size and the version of light-curve fitting software used for supernovae.

So where does this leave cosmology? The standard model, known as Lambda-CDM, has been spectacularly successful at describing the large-scale universe. It accounts for the cosmic microwave background, the distribution of galaxies, the abundance of light elements forged in the first minutes after the Big Bang. But it predicts a present-day expansion rate of about 67 or 68 kilometres per second per megaparsec, and the local measurements stubbornly refuse to agree. If neither side is wrong (and this new work makes it considerably harder to argue the local side is), then something about the model may be incomplete. Perhaps dark energy behaves differently than assumed. Perhaps there are additional relativistic particles that cosmologists have not yet accounted for. Perhaps gravity itself works slightly differently on cosmological scales.

The collaboration has released all its code and data publicly, inviting anyone to rerun the analysis, swap in different datasets, or extend the framework with future observations. The James Webb Space Telescope is already refining Cepheid and TRGB measurements with sharper resolution than Hubble could manage, and forthcoming surveys from the Vera C. Rubin Observatory and other next-generation instruments will add thousands more distance tracers in the coming years.

Whether those new observations will finally close the gap or prise it wider still is, for now, an open question. But after a decade of prodding, poking, and systematically dismantling every plausible source of error, the universe’s expansion rate remains stubbornly, persistently, inconveniently mismatched. Whatever is causing it probably is not a mistake.

Source: H0DN Collaboration, “The Local Distance Network: A community consensus report on the measurement of the Hubble constant at ~1% precision,” Astronomy & Astrophysics, 708, A166 (2026). DOI: 10.1051/0004-6361/202557993