DESI (the Dark Energy Spectroscopic Instrument) has produced the strongest observational hint to date that dark energy is not a cosmological constant. In April 2024, DESI DR1 BAO measurements combined with CMB and Type Ia supernova data preferred the evolving dark energy parameterization w(a) = w₀ + wa(1−a) over ΛCDM, with best-fit values w₀ > −1 and wa < 0 at 2.5 to 3.9σ depending on the supernova compilation (arXiv:2404.03002). The March 2025 DESI DR2 release strengthened the signal: DESI BAO plus CMB alone favors w₀waCDM at about 3.1σ, rising to 2.8, 3.8, or 4.2σ when the Pantheon+, Union3, or DES-Y5 supernova samples are added (arXiv:2503.14738). Taken at face value, the data describe a dark energy component that was in a phantom phase (w < −1) in the past, crossed the w = −1 divide near z ≈ 0.4–0.5, and is weakening today.
ΛCDM has no room for this. The model hard-codes w = −1 exactly, at all times and in all environments, because its dark energy is a constant vacuum term in the field equations; there is no mechanism by which the equation of state can drift. A confirmed w₀ > −1 with wa < 0 would therefore falsify the cosmological constant outright. Worse, the implied phantom crossing cannot be produced by the simplest replacement, a single minimally coupled quintessence scalar field, which is mathematically confined to one side of the w = −1 divide. Reproducing the DESI signal requires multi-field models, modified gravity, or an interacting dark sector, each introducing new degrees of freedom and new fine-tuning. The same fits also entangle with the neutrino sector: allowing evolving dark energy relaxes the cosmological neutrino mass bound to roughly 0.16 eV, easing a separate emerging tension with oscillation experiments.
The standing of the result remains contested. Independent reanalyses argue the preference is dominated by low-redshift supernova calibration systematics; correcting the low-z sample can shrink the significance to 0.5 to 1.5σ (arXiv:2502.04212), and sensitivity to Planck low-ℓ polarization and the optical depth to reionization has also been flagged. Whether the signal is new physics or residual systematics should be settled by DESI DR3, Euclid, Rubin-LSST, and Roman later this decade. As of 2026 it stands among the most statistically significant hints of physics beyond ΛCDM, and it is the first dataset-driven indication that the dark energy sector itself, not just its calibration, may be dynamical.