What Is the Hubble Tension?

Take the most carefully measured number describing how fast the universe is expanding. Now measure it a second way. The two answers do not match. They miss by about nine percent. That does not sound like much. But both measurements are now precise to about one percent, so the gap between them is far too large to be an accident. Physicists have spent a decade hunting for the mistake that would make it go away. They have built sharper instruments, rechecked every step, and ruled out the obvious culprits one by one. The gap has not closed. If anything, it is getting harder to explain.

One constant, two independent roads to it

The Hubble constant, written H₀, sets the present-day expansion rate of the universe. It is measured in kilometres per second of recession velocity, for every megaparsec of distance. H₀ is one of cosmology’s most fundamental numbers, and there are two structurally different ways to measure it.

The early-universe route starts from the cosmic microwave background. This is the relic radiation released when the universe first became transparent, 380,000 years after the Big Bang. Its faint pattern of temperature ripples encodes a precise sound horizon, a standard ruler baked in at that moment. Combine that ruler with the standard cosmological model, and H₀ falls out as a derived quantity.

The late-universe route measures H₀ directly, with no cosmological model required. It builds a distance ladder. Parallax fixes distances to nearby stars. Cepheid variable stars, calibrated against those distances, extend the ladder to galaxies hosting Type Ia supernovae. The supernovae extend it further still, out to where their redshift and known brightness give H₀ directly from Hubble’s law.

The two numbers, and the gap between them

Planck’s cosmic microwave background analysis gives H₀ near 67.4 kilometres per second per megaparsec. The SH0ES collaboration’s distance-ladder measurement gives roughly 73.0 to 73.5. Each side quotes uncertainties near one percent. A gap this size is not something either method can shrug off as noise. As of the 2025 assessments the tension has passed 6 standard deviations, well beyond the 5-sigma threshold physicists use to declare a discovery.

A third, largely independent route uses baryon acoustic oscillations. This is the same sound-horizon physics as the cosmic microwave background, but read off the clustering of galaxies rather than the microwave sky. The Dark Energy Spectroscopic Instrument gives H₀ = 68.53 ± 0.80, close to Planck and in tension with SH0ES, despite using none of Planck’s data.

The obvious suspect: crowded starfields

For years, the leading systematic explanation was Cepheid crowding. Cepheids sit in busy, star-forming regions of distant galaxies. The Hubble Space Telescope’s resolution could not always separate a Cepheid from its unresolved neighbours. Unrecognised crowding makes a star look brighter than it is, which makes it appear closer than it is, which biases H₀ upward.

NASA built the James Webb Space Telescope partly to settle this. Its near-infrared vision resolves roughly three times finer than Hubble at comparable wavelengths. That is sharp enough to pick apart individual Cepheids that Hubble could only see as a single smear of light.

JWST checked the crowding hypothesis and it failed

Adam Riess and collaborators compared more than a thousand Cepheids across five supernova host galaxies, observed with both telescopes. The mean difference between Hubble and JWST distances came out to −0.01 ± 0.03 magnitudes, consistent with zero. The team ruled out distance-dependent crowding bias as the explanation, at 8.2 standard deviations, a higher confidence than the tension itself.

A 2025 follow-up added a particularly clean low-background host galaxy, NGC 3447A, and again found no hidden bias. Combining 24 Type Ia supernovae across 19 JWST-observed Cepheid hosts gave H₀ = 73.49 ± 0.93, matching the original Hubble-based SH0ES result almost exactly. The telescope built to catch the error instead confirmed the number.

Ruling out one bad measurement, systematically

A further line of work tested whether any single method, distance indicator, or calibration step could be dropped to make the tension disappear. It could not. Excluding any one ingredient from the distance-ladder calculation leaves the local H₀ value essentially unchanged. The tension is spread across the entire measurement, not concentrated in one identifiable weak link.

That matters, because it forecloses the easiest kind of explanation. A single flawed calibration, or an overlooked systematic in one step, would have been the tidiest resolution available. The data no longer support that story.

Where the search for a fix stands

With the leading systematic explanations weakening, attention has shifted toward new physics that changes the expansion history itself. Early dark energy models add a brief burst of energy in the young universe. That burst shrinks the sound horizon and raises the cosmic microwave background’s inferred H₀, meeting the local value partway. One such model, combining cosmic microwave background, baryon acoustic oscillation and JWST data, raises H₀ to 71.58 ± 1.05. That eases the tension to roughly 1 standard deviation in that particular fit. The same DESI data drives a parallel standoff over the summed mass of the neutrinos.

Easing a tension inside one model is not the same as showing that model is correct. Early dark energy has to thread a needle. It must shift H₀ without disturbing the rest of the cosmic microwave background’s tightly constrained fit. It must also match how galaxies clustered as the same physics played out. No proposed extension of the standard model has yet done all of that cleanly and won broad acceptance.

What would actually settle it

A fully independent local measurement would carry the most weight, one that bypasses Cepheids entirely. Gravitational-wave standard sirens are the most promising candidate. The wave itself supplies the distance, with no stellar calibration step needed. But the sample of well-localised events is still too small to compete with the precision of the distance ladder or the cosmic microwave background.

Until that arrives, the field is left with two robust, independently checked numbers for the same quantity. They disagree by more than either measurement’s stated uncertainty allows. That is what makes this a tension rather than an error, and why it now sits among the sharpest open problems in cosmology.