Dark Energy & Hubble Tension
The universe expands at an accelerating rate (1998), caused by 'dark energy', and two ways of measuring the expansion rate disagree.
Open in the interactive tree →Two supernova teams found in 1998 that the expansion is accelerating (Nobel Prize 2011). In the standard model the cause is a constant 'cosmological constant' making up about 68% of the energy content. In addition, measurements in the nearby universe (about 73 km/s/Mpc) and from the cosmic microwave background (67.4 km/s/Mpc) give different expansion rates: the Hubble tension.
As of October 2026
Local measurements (SH0ES, about 73 km/s/Mpc) and the cosmic microwave background (Planck, 67.4) disagree by roughly 5 sigma on the SH0ES team's estimate. JWST data support Hubble's Cepheid distances, while the CCHP team (Freedman) gets about 70.4 with a different method (red-giant stars) and argues the tension may be milder; the dispute continues. DESI Data Release 2 (2025) hints at time-varying dark energy at 2.8 to 4.2 sigma, below the 5-sigma discovery threshold, and DESI finished its planned survey in April 2026 with the full analysis still to come. No explanation is accepted.
What is missing
- Independent measurements of the expansion rate (gravitational-wave standard sirens, lensing time delays) with better precision
- More and better data: DESI DR3, Euclid, the Rubin Observatory, the Roman telescope
- A theory of dark energy (calculated vacuum energy is off by about 10^120)
- Clarity on systematic errors in supernova calibration
Becomes possible once solved
- Knowing whether the expansion continues forever, slows or tears everything apart
- A corrected standard model of cosmology
- Hints to new physics in the early universe or in gravity
Open steps
- Distance-ladder calibration errors Medium AI leverageFind out whether hidden systematics in Cepheid, red-giant-branch or supernova calibration explain the gap between local and microwave-background values.
- Expansion rate from standard sirens High AI leverageMeasure the expansion rate from gravitational-wave events with known distance, while controlling selection effects such as Malmquist bias.
- Is dark energy evolving? Medium AI leverageTest with the full DESI data set, Euclid and new supernova samples whether dark energy weakens over time at 5 sigma, and resolve disagreement between supernova samples.
- A theory of the dark-energy value Low AI leverageExplain why vacuum energy is about 10^120 times smaller than naive quantum-field estimates, or replace the cosmological constant with a dynamic field.
Where AI could help
Medium AI leverage. AI emulators and neural inference make cosmology fits much faster, but the tension hinges on calibration systematics and new surveys, not on model speed.
- Emulate theory predictions so tests of many dark-energy models run orders of magnitude faster
- Classify supernovae and model gravitational lenses from Rubin, Euclid and Roman survey data
- Re-analyze distance-ladder calibrations at scale to hunt for hidden systematic errors
- Fit time-varying dark-energy models to DESI DR3 and other data sets jointly
Shown so far
- A June 2026 benchmark on DESI DR2, Pantheon+ and cosmic-chronometer data found neural simulation-based inference matched exact MCMC to better than 0.3 sigma on the simpler dataset with order-of-magnitude shorter runtime. source
Prerequisites
- Expansion & Big Bang1929
- Cosmic Microwave Background1964CMB measurements fix the dark-energy and Hubble-tension numbers
- Accelerating Universe1998Dark energy was inferred from the 1998 supernova result
- DESI: Evolving Dark Energy?2025