A September 2025 Bayesian decomposition attributed 78% of the Hubble tension to real physical effects and only 13% to measurement error. That 78% number has been circulating as near-proof that new physics lurks behind the 9% gap between local and CMB expansion rates. I think it proves something narrower: the discrepancy is unlikely to be a single dumb mistake. That is a long way from proving the standard cosmological model is broken.
I cover energy and climate, not astrophysics. But I spend my working life watching people confuse statistical confidence with engineering certainty, and the Hubble tension debate has the same disease. A press release announcing 5 sigma does not deploy new physics any more than a press release announcing a gigawatt of solar capacity deploys electrons. The question is whether the measurement chain holds up under independent scrutiny at the precision the researchers themselves demand. By their own admission, it does not. Not yet.
The Chain Matters More Than the Sigma
Local measurements of H₀ rely on a distance ladder: anchor galaxies calibrated with Cepheids or JAGB stars, then Type Ia supernovae extending the ruler outward. Each rung introduces calibration assumptions. The CMB value of 67.4 ± 0.5 km/s/Mpc depends on the ΛCDM model being correct. Both sides carry baggage.
JWST confirmed that Hubble Space Telescope's Cepheid photometry was not grossly wrong. Good. That eliminates one specific error. It does not eliminate the possibility of correlated systematics across the ladder, or subtle model dependencies in how CMB data gets translated into an expansion rate. The December 2025 gravitational lensing measurement landed at 73 km/s/Mpc with 4.5% precision. Consistent with local values, yes. But 4.5% precision on a 9% discrepancy is not the independent confirmation people want it to be. It is suggestive. Suggestive is not conclusive.
The 2025 International Space Science Institute workshop confirmed the local value holds at roughly 1% precision across multiple techniques. That is impressive. It is also exactly the kind of convergence you would expect if the techniques share calibration heritage. The workshop's own participants noted that 1-2% cross-method precision is still needed to fully rule out unknown systematics. They set the bar. We should hold them to it.
What Would Change My Mind
Fair point to the other side: when 4 independent methods all land near 73 km/s/Mpc and the CMB sits stubbornly at 67.4, the simplest explanation starts to look like a real physical effect. I grant that. The pattern is hard to dismiss.
But "hard to dismiss" is not "proven." Early dark energy models that fix the Hubble tension tend to worsen the S8 tension on galaxy clustering. That is a red flag. If your proposed new physics solves one problem while creating another, you have not found the answer; you have moved the discrepancy. Cosmology has been here before. The Pioneer anomaly ran at roughly 8 sigma for years before thermal radiation pressure explained it. High sigma, wrong conclusion.
What would actually convince me: a measurement of H₀ from a method with zero shared calibration heritage with the distance ladder, at 1-2% precision, landing near 73. Gravitational wave standard sirens could do this. LIGO-Virgo-KAGRA's current constraint sits around 68 ± 6 km/s/Mpc, far too imprecise to settle anything. The next observing run, O5, expected to begin in 2027, should tighten that considerably. That is the measurement worth waiting for.
Adam Riess says we need to reexamine the foundations of the current cosmological model. I agree with the impulse. But reexamining foundations means stress-testing them, not replacing them with speculative alternatives that fail their own consistency checks. Fund the gravitational wave observations. Fund the next generation of lensing surveys. Get the cross-method precision to 1%. Then, if 73 still stands, I will be the first to say ΛCDM needs a rewrite.
Until then, 5 sigma built on a shared scaffolding is a reason to keep measuring, not a reason to start rewriting.
