The universe is expanding at two different speeds

The Hubble constant is the rate at which the universe is expanding today, usually written in the unwieldy units of kilometers per second per megaparsec. It sets the age of the universe, the scale of everything in it and the clock against which cosmologists check their models. For most of the twentieth century, astronomers argued about whether it was closer to 50 or 100. They would be delighted to have today’s problem: two methods, each precise to about one percent, that disagree by about eight.

One method starts in the early universe. Using the pattern of hot and cold spots in the cosmic microwave background, together with the standard cosmological model, the Planck satellite’s final analysis predicts a value of 67.4 kilometers per second per megaparsec. The other method measures nearby galaxies directly. A community consensus published in 2026 by the Local Distance Network, combining many independent distance techniques, gives 73.50 ± 0.81. The two differ by about seven standard deviations.

What eight percent looks like

It helps to translate the units. A galaxy 100 megaparsecs away, some 326 million light-years, should be receding at 6,740 kilometers per second if the early-universe number is right, and 7,300 if the local number is. The reciprocal of the constant gives a rough timescale for the expansion, the so-called Hubble time: 14.5 billion years for the lower value and 13.4 billion for the higher.

That second figure is less alarming than it looks, since the actual age of the universe depends on its whole expansion history. But it shows what is at stake. And there is one number that frames the problem better than any other. In the magnitude system astronomers use for brightness, an 8 percent error in distance amounts to about 0.17 magnitudes. If the tension is a mistake in the local measurements, the mistake must be about that large, and it must affect the calibration systematically. Astronomers now routinely measure the rungs of the ladder more precisely than that.

A ladder of distances

Measuring the expansion directly requires two things for each galaxy: how fast it is receding, which is easy, and how far away it is, which is very hard. Astronomers build distances in steps. Parallax gives distances to nearby stars in the Milky Way. Some of those stars are Cepheid variables, which pulse with a period tied to their true brightness, a relationship discovered by Henrietta Leavitt in 1912. Cepheids in nearby galaxies then calibrate the brightness of Type Ia supernovae, which are bright enough to be seen far into the expanding universe.

Every step carries its own uncertainties, and an error in an early rung propagates to everything above it. For years the most popular explanation for the tension was a subtle bias in the Cepheid measurements, perhaps from the light of neighboring stars blending into the Cepheids’ images in crowded galactic disks.

What Webb was supposed to settle

The James Webb Space Telescope offered a test. Its sharper infrared images can separate Cepheids from their neighbors far better than Hubble could. When the SH0ES team, led by Adam Riess, observed Cepheids in several supernova host galaxies with Webb, the measurements matched Hubble’s. Crowding, it seemed, was not the answer.

But Webb also made room for independent checks, and those did not line up neatly. A team led by Wendy Freedman used other kinds of stars as distance markers, the brightest red giants and a class of carbon-rich stars, and its latest combined analysis arrives at 70.39, which that team argues is consistent with the early-universe value. Others disagree, and at a 2025 meeting of the American Astronomical Society one prominent researcher declared the tension a crisis. A separate approach using gravitationally lensed quasars, which involves no distance ladder at all, gives about 71.6 with an uncertainty of 4.6 percent.

Two methods, each precise to about one percent, disagree by about eight.

The ruler that is not measured

A point often lost in coverage: the Planck number is not a measurement of today’s expansion. What the microwave background measures with exquisite precision is an angle, the apparent size of the sound-wave pattern on the sky. To turn an angle into a distance you need the true size of the pattern, and that has to be calculated from the physics of the first 380,000 years.

This is why the most discussed theoretical fixes target the early universe. If something made that sound-wave ruler about 8 percent shorter than the standard calculation, the inferred expansion rate would rise to match. A brief burst of extra dark energy before the microwave background was released could do it, as could new light particles or interactions among dark components. Each idea eases the tension, but most create new problems in other data, which the standard model had fit very well.

Mistake or discovery

If the local measurements are right, then the standard cosmological model, which has passed almost every other test, is missing something about the early universe. If the early-universe value is right, then something is still wrong with the local ladder, and finding it would mean revising a century of methods. The consensus report’s finding that removing either Cepheids or red giants barely changes its answer narrows the room for a single faulty rung, though it does not close the argument with the teams who get lower numbers.

New data are arriving quickly. Gravitational-wave events with visible counterparts offer a distance measurement with no ladder, and large surveys are mapping the expansion history in fine detail. The same surveys are now raising questions about dark energy itself. For a field used to agreement, the next few years may be unusually argumentative.

Sources

• “The Local Distance Network: a community consensus report on the measurement of the Hubble constant”, Astronomy & Astrophysics, 2026
• Planck Collaboration, “Planck 2018 results”, Astronomy & Astrophysics, 2020
• Freedman et al., “Status Report on the Chicago-Carnegie Hubble Program”, The Astrophysical Journal, 2025
• Riess et al., “JWST Observations Reject Unrecognized Crowding of Cepheid Photometry”, 2024