Something remarkable is happening at the edge of the observable universe—and our standard models of cosmology can’t explain it.

Using the newly released COSMOS2025 catalog from the James Webb Space Telescope’s COSMOS-Web survey, I’ve analyzed 26,288 high-redshift galaxies and found that massive galaxies at z > 6 aren’t just slightly more abundant than predicted—they exceed our theoretical limits by factors of hundreds.

Let me be clear: this isn’t about models being off by 10% or 20%. At redshifts z > 8, we’re seeing galaxy abundances that exceed even the maximum physically possible number—the hard ceiling where every single baryon in every dark matter halo is converted into stars with 100% efficiency.

That ceiling is unphysical. Real galaxies can’t reach it. Yet the observations blow past it anyway.

The Numbers

The COSMOS2025 catalog—the definitive JWST COSMOS-Web data release covering 0.54 square degrees—provides photometric redshifts and stellar masses for roughly 700,000 galaxies. From this, I selected 7,837 massive galaxies (log M★/M☉ > 9) at z > 5.

Here’s what the comparison to standard ΛCDM predictions looks like:

The excess grows monotonically with redshift. This isn’t a statistical fluctuation in one bin—it’s a systematic trend across the entire high-redshift regime.

Why This Can’t Be Explained Away

The immediate response to findings like this is typically: “You must have systematic errors.” I’ve spent considerable effort stress-testing these results.

Photometric redshift contamination? Even if 50% of the z > 8 sample are actually lower-redshift interlopers (a pessimistic assumption inconsistent with spectroscopic validation rates), the excess still exceeds 5× the halo limit.

Stellar mass overestimation? Even if stellar masses are systematically overestimated by a full dex (factor of 10), the tension at z > 8 remains >7× above the physical ceiling.

Cosmic variance? The 0.54 deg² survey volume corresponds to ~30% uncertainty from cosmic variance at z ~ 8. The observed excess is 100-500×.

Duty cycle effects? Arguments about galaxies being visible only briefly actually worsen the tension—if galaxies are visible for only a fraction of their existence, there must be even more total halos to produce the observed counts.

Under the most pessimistic combined scenario—50% contamination + 0.5 dex mass shift + 3σ cosmic variance—the excess at z > 8 still exceeds 20×.

Independent Confirmation: Star Formation Rates

Number counts can be affected by mass calibration issues. But specific star formation rate (sSFR = SFR/M★) is largely independent of absolute mass normalization, since systematic mass errors partially cancel in the ratio.

The sSFR shows a striking positive correlation with redshift (Spearman ρ = 0.33, p < 10⁻¹⁰). High-redshift galaxies aren’t just more numerous than expected—they’re forming stars approximately twice as efficiently per unit mass compared to their z ~ 5 counterparts.

This corroborating trend is crucial: it tells us we’re seeing genuine enhanced activity in the early universe, not merely calibration artifacts.

What Does This Mean?

I want to be careful here. The paper I’ve published (DOI: 10.6084/m9.figshare.31059964) is intentionally model-agnostic. It establishes an empirical tension without advocating for any particular resolution.

The tension admits three broad classes of explanation:

1. Unidentified systematic errors: Catastrophic photo-z failures, unknown selection effects, or SED modeling failures we haven’t accounted for.
2. Non-standard baryonic physics: More efficient early star formation, reduced feedback, or modified initial mass functions—all requiring substantial revisions to galaxy formation physics.
3. Modified cosmology: Enhanced early structure formation through mechanisms like early dark energy, modified gravity, or increased primordial power on relevant scales.

From the perspective of Energy-Flow Cosmology, the findings are intriguing. If entropy gradients in the early universe were steeper than in later epochs, this could drive more rapid energy flow and accelerated structure assembly. The monotonic increase in galaxy excess with redshift would then reflect the evolution of entropy gradient magnitude over cosmic time.

But I emphasize: this interpretation is offered as one consistent framework, not as a proven explanation. The empirical results stand independently of any theoretical interpretation.

The Bottom Line

Whatever the ultimate explanation, the COSMOS-Web data establish a firm observational fact: the early universe produced massive galaxies far more efficiently than our standard models predict. At z > 8, this efficiency exceeds even theoretical maximum limits.

This is either telling us something profound about early galaxy formation, or pointing to observational systematics we haven’t yet identified. Either outcome advances our understanding.

The data and analysis code are fully available for independent verification. Science works best when extraordinary claims can be checked.

Citation: Magnusson, M. (2026). Observed Galaxy Abundances at z > 6 Exceed Halo-Limited Predictions in COSMOS-Web. DOI: 10.6084/m9.figshare.31059964

About the Author

Morten Magnusson

Administrator

With a background in thermodynamics and systems theory, his work focuses on unifying cosmology, physics, and cognition under a continuous energy–entropy framework. Magnusson’s publications explore how energy flow and entropy shape spacetime, structure, and awareness across physical and informational domains. He has authored over thirty papers archived through Figshare, including the Energy-Flow Cosmology (EFC v2.1) series, the Grid–Higgs Framework, and the CEM–Cosmos Model of Cognitive Thermodynamics. His current research develops computational implementations of EFC using CLASS, RAMSES, and GADGET-4 to test energy–entropy coupling in large-scale cosmological structures. Magnusson leads the long-term project to integrate physics, biology, and artificial intelligence into a single thermodynamic continuum — where energy, entropy, and information are seen as expressions of the same underlying flow.

Visit Website View All Posts