The Universe Won't Agree With Itself

What We Know

Every galaxy we can see is moving away from us, and the farther away it is, the faster it’s receding. That relationship has a name, the Hubble constant, and a value: roughly how many kilometers per second faster a galaxy recedes for every additional megaparsec of distance. Cosmologists have two independent ways to measure it. One works backward from the cosmic microwave background, the afterglow of the Big Bang, using the standard model of cosmology to predict what today’s expansion rate should be: about 67 kilometers per second per megaparsec. The other measures it directly, using the brightness of specific nearby stars called Cepheid variables and a chain of exploding stars called Type Ia supernovae to build a distance ladder out to galaxies still close enough to calibrate precisely: about 73.

Those two numbers should match. They don’t. Both measurements have gotten more precise over the past decade, and the gap between them hasn’t closed. It’s widened, past the point where statisticians are willing to call it chance. By late 2025, independent teams working the numbers with different statistical methods were reporting the disagreement at past six standard deviations, the kind of confidence physicists usually reserve for a genuine discovery rather than a mistake waiting to be found.

The Roadblock

In 2023, the James Webb Space Telescope took a direct look at exactly the kind of measurement error skeptics had proposed for years: that crowded star fields or interstellar dust were quietly corrupting the Cepheid distance measurements. JWST checked. The Hubble Space Telescope’s numbers held up. That was supposed to be the systematic error hiding in plain sight, and it wasn’t there. Ruling out your leading suspect should feel like progress. Here it mostly closed off the easiest way out, and left two of the most carefully checked numbers in physics simply refusing to agree.

That leaves cosmologists in an uncomfortable spot. The standard model of the universe, the one built from a century of otherwise successful predictions, doesn’t have room in it for two correct expansion rates at once. Either one measurement chain has an error nobody has found after a decade of dedicated searching, or the model itself is missing something. Neither option is small.

Best Guesses

New physics before recombination

The theoretical favorite among many cosmologists: an extra component of energy, active only in the universe’s first few hundred thousand years, that current models leave out entirely. Adding it would shrink a specific early-universe distance called the sound horizon, which would shift the CMB-based prediction toward the higher, locally measured number. It fits the data reasonably well. It also requires inventing a new physical ingredient that no experiment has independently detected.

The Hubble bubble

A more modest idea: what if the disagreement isn’t universal, but local? If our region of the cosmos happens to be somewhat emptier than average, gravity from denser surrounding regions would pull outward on it, locally speeding up the apparent expansion rate without changing the true cosmic average at all. Some surveys of galaxy density around us are consistent with this. Others aren’t, and the size of underdensity required is larger than most theorists are fully comfortable with.

Early-universe systematic error

Perhaps the CMB-based prediction is the one that’s subtly wrong, not the local measurement, buried somewhere in the immense chain of calibration and modeling required to turn microwave background data into a number. Multiple independent teams have re-run that analysis looking for exactly this. None have found an error large enough to close the gap.

Late-universe systematic error

The mirror-image possibility: something in the nearby distance ladder is still off, despite JWST’s check. This was the leading suspicion for years and remains the hardest to fully rule out, since it would require a flaw that survives multiple independent verification methods. It’s a harder position to hold today than it was five years ago.

The Verdict

Cosmology has a decade of failed error-hunting behind it and one specific referee ahead. Gravitational-wave detections yield a distance measurement sharing no calibration steps with either the microwave background or the Cepheid ladder, which is exactly what a tiebreaker requires. There simply aren’t enough of them yet, and the ones on hand are too imprecise to rule. Every new detection tightens that a little.

So the resolution is coming on a schedule nobody controls, and it has only two available forms. Either a decade of extraordinarily careful work hides a flaw that dozens of teams have hunted and missed, or the model that has described this universe successfully for a century is missing an ingredient. Most cosmologists have quietly stopped expecting the first one.