The Universe Has a Twist: DESI's 14 Million Galaxies Challenge the Cosmological Constant

There is a number embedded in the standard model of cosmology — the cosmological constant, denoted Λ — that cosmologists have quietly suspected might be too simple to be true. It was Einstein’s greatest blunder and then his greatest posthumous vindication. Today, that number is under siege. And the weapon is a telescope on a mountain in Arizona, equipped with 5,000 robotic fiber positioners and the patience to stare at 14 million galaxies.

In March 2025, the Dark Energy Spectroscopic Instrument (DESI) collaboration published its Data Release 2 results, representing three years of continuous observation. The headline finding: the universe’s dark energy may not be the static, featureless vacuum energy that ΛCDM (the standard “Lambda cold dark matter” model) predicts. It may be changing over time. And the statistical evidence for that change — ranging from 2.8 to 4.2 sigma depending on which supernova dataset is included — is getting hard to dismiss.

What Is the Cosmological Constant, and Why Does It Matter?

When Einstein first introduced Λ in 1917, it was a fudge factor to make the universe static. He removed it when Hubble showed the universe was expanding. Then, in 1998, two teams measuring distant Type Ia supernovae discovered that the expansion was accelerating — a result so counterintuitive it earned the 2011 Nobel Prize. Something was pushing the universe apart. That something was named “dark energy,” and in the simplest formulation, it corresponds to a constant energy density in empty space: the cosmological constant, resurrected.

In this picture, dark energy has an equation-of-state parameter w = −1, precisely. It doesn’t dilute as the universe expands. It doesn’t evolve. It just is, baked into the geometry of spacetime.

The question that has lurked beneath cosmology for the past 25 years is: what if w ≠ −1? What if dark energy evolves? What if what we call the cosmological constant is actually something more interesting — a dynamic field, perhaps, or a hint of physics beyond general relativity?

DESI was built, in part, to answer exactly that question.

The Instrument: 5,000 Eyes on the Universe

DESI sits atop the Nicholas U. Mayall 4-meter Telescope at Kitt Peak National Observatory in Arizona, operated by NOIRLab. Its heart is a focal plane packed with 5,000 individually addressable robotic fiber positioners, each capable of repositioning in about 5 seconds to acquire a new target. Each fiber feeds light to spectrographs that measure the redshift — the stretching of light due to cosmic expansion — of whatever object the fiber is pointed at.

This means DESI can measure the redshifts of 5,000 galaxies or quasars simultaneously. Night after night. For years.

The instrument surveys several types of objects across a huge range of cosmic distances: nearby bright galaxies, luminous red galaxies, emission line galaxies, and distant quasars — including measurements through the Lyman-alpha forest, the dense thicket of hydrogen absorption features imprinted on quasar light by intergalactic gas at high redshift. Together, these tracers allow DESI to map the large-scale structure of the universe across cosmic history.

The Cosmic Ruler: Baryon Acoustic Oscillations

DESI’s primary tool for measuring dark energy is the baryon acoustic oscillation (BAO) feature — a subtle, periodic ripple in the distribution of galaxies across the sky. Before the universe was about 380,000 years old, it was a hot plasma of protons, electrons, and photons, all tightly coupled and vibrating like a sound wave. When the universe cooled enough for electrons to combine with protons (the epoch of recombination), those sound waves froze in place, leaving a preferred clustering scale imprinted in the distribution of matter.

That scale is approximately 490 million light-years. Because it’s set by well-understood physics from the early universe, it acts as a cosmic standard ruler — a known physical scale that astronomers can measure at different distances and epochs to trace the expansion history of the universe with high precision.

DESI DR2’s galaxy BAO analysis covered more than 14 million galaxies and quasars in the redshift range 0.1 < z < 4.2. Combined with the Lyman-alpha forest BAO (which adds measurements at the highest redshifts, with statistical precision of 0.65% on the isotropic BAO scale at an effective redshift of z = 2.33), the collaboration assembled the most comprehensive BAO dataset ever compiled.

The Signal: Dark Energy Is Moving

In the standard ΛCDM model, dark energy is characterized by equation-of-state parameter w = −1 everywhere and for all time. The simplest generalization of this is the “w₀wₐ” parameterization, in which:

w(z) = w₀ + wₐ × z/(1 + z)

Here, w₀ is the value today and wₐ describes how much w changes over cosmic time. In ΛCDM, both w₀ = −1 and wₐ = 0.

The DESI DR2 results favor a solution with w₀ > −1 and wₐ < 0. This means dark energy was stronger in the past and is weakening — the opposite of many intuitions about a runaway cosmological constant. When combining DESI BAO data with the cosmic microwave background (CMB), this dynamical dark energy model is preferred over ΛCDM at 3.1 sigma. When the collaboration also folds in supernova data, the significance ranges from 2.8 to 4.2 sigma depending on which supernova compilation is used.

By the conventions of physics, 3 sigma is “evidence” for a new phenomenon; 5 sigma is “discovery.” So DESI is not yet claiming a confirmed detection of new physics. But 3.1 to 4.2 sigma is not noise. It is a persistent, coherent signal from three independent probes — BAO, CMB, and supernovae — all pointing in the same direction.

Context: DR1 Became DR2

These results didn’t come out of nowhere. DESI’s first year of data (DR1), released in April 2024, already showed hints of this behavior at the 2.6 sigma level for DESI + CMB, and up to 3.9 sigma when combined with supernovae. DR2, based on three years of data and more than twice as many objects, deepens those hints into something harder to shrug off.

Crucially, the DR2 results are consistent with DR1, with SDSS (the Sloan Digital Sky Survey, DESI’s predecessor), and with independent supernova datasets. The signal doesn’t go away when you use different data. It doesn’t vanish when you change analysis methods. It persists.

The paper is bracingly direct about the implications: “Unless there is an unknown systematic error associated with one or more datasets, it is clear that ΛCDM is being challenged by the combination of DESI BAO with other measurements and that dynamical dark energy offers a possible solution.”

That “unless” carries a lot of weight, and the collaboration has worked hard to earn the right to write it. The DR2 analysis includes extensive checks for systematic errors — from fiber assignment completeness to photometric calibration to redshift failures — and finds no smoking gun.

What Could It Mean?

If dynamical dark energy is real, several theoretical frameworks could explain it.

The most commonly discussed is quintessence — a hypothetical scalar field that permeates space and evolves slowly over cosmic time, somewhat like the Higgs field but at cosmological scales. Depending on the potential energy landscape of this field, it could produce a dark energy that was stronger in the early universe and weakens today, matching the w₀wₐ signal DESI sees.

Another possibility is coupled dark energy, in which dark energy interacts with dark matter, modifying the expansion history in subtle ways. Yet another is a modification to general relativity itself — not dark energy at all, but a breakdown of Einstein’s gravity at cosmic scales.

And then there’s the possibility that the cosmological constant is correct, and one of the datasets harbors an undetected systematic error. Supernovae in particular require careful calibration; different groups have constructed supernova compilations (Pantheon+, Union3, DES-SN5YR) that yield slightly different constraints, which is why the significance of DESI’s result spans a range rather than a single number.

Cosmologists are being appropriately cautious. This is exactly the right attitude: exciting results demand the highest scrutiny.

The Neutrino Bonus

Buried in the DR2 results is another headline-worthy finding. By combining DESI BAO with Planck CMB data, the collaboration derived a 95% upper limit on the sum of neutrino masses:

Σmᵥ < 0.064 eV (assuming ΛCDM)

This is one of the tightest constraints ever placed on neutrino masses from cosmological data. For context, neutrino oscillation experiments tell us that neutrinos have mass, and that the sum must be at least about 0.06 eV. DESI’s upper limit is kissing the minimum allowed by particle physics — a situation that will become even more interesting as the instrument collects more data, and as next-generation surveys push the sensitivity further.

What Comes Next

DESI is still running. The collaboration plans to collect data until at least 2026, accumulating a five-year dataset that will dwarf even DR2. The full survey will cover roughly one-third of the sky to a depth of about 3 billion light-years — a three-dimensional map of cosmic structure that will allow BAO measurements precise enough to test dark energy models at the sub-percent level.

Meanwhile, complementary surveys are gearing up. The Euclid space telescope, launched in 2023, is mapping billions of galaxies using both imaging and spectroscopy. The Nancy Grace Roman Space Telescope, set to launch in the late 2020s, will add near-infrared photometry and slitless spectroscopy across a wide field of view. Stage-IV surveys like the Vera C. Rubin Observatory will cross-check results using weak gravitational lensing and photometric redshifts. The CMB-S4 experiment will sharpen CMB constraints on dark energy and neutrino masses independently.

Together, these facilities will either confirm the DESI signal or rule it out. If it’s confirmed, cosmology has a new constant to explain — or, perhaps, a new fundamental field to add to the list of nature’s ingredients. If it evaporates with more data, that will tell us something important too: that ΛCDM is more robust than its cracks suggest.

A 25-Year Mystery, Still Open

There is something beautifully apt about the current moment in cosmology. The field spent the 1990s astonished to discover that the universe was accelerating at all. It spent the 2000s building ΛCDM into a precision machine, testing it against survey after survey, finding it consistent. And now, in 2025, the most sensitive instrument ever built to characterize dark energy is whispering that maybe — maybe — the simplest explanation was never the right one.

That whisper is getting louder. DESI’s robotic fibers keep repositioning, night after night, collecting light from galaxies billions of light-years away, encoding in their spectra the geometry of a cosmos that doesn’t quite fit the model.

The cosmological constant may yet survive. But it’s going to have to work for it.

---

Key papers:

• DESI Collaboration, “DESI DR2 Results II: Measurements of Baryon Acoustic Oscillations and Cosmological Constraints,” Physical Review D (2025). arXiv:2503.14738, DOI: 10.1103/tr6y-kpc6
• DESI Collaboration, “DESI DR2 Results I: Baryon Acoustic Oscillations from the Lyman Alpha Forest,” Physical Review D (2025). arXiv:2503.14739, DOI: 10.1103/2wwn-xjm5
• DESI Collaboration, “DESI 2024 VI: Cosmological Constraints from BAO Measurements,” JCAP 2025(02):021. arXiv:2404.03002, DOI: 10.1088/1475-7516/2025/02/021