With the advent of the James Webb Space Telescope (JWST), observational probes of the structure of objects in the early universe are more readily available. In particular, the discovery of high-redshift (z ∼ 10) Supermassive Black Holes (SMBHs) challenges the typical formation channels of these objects, which cannot form sufficiently massive seeds or grow them within these timescales without significant periods of super-Eddington accretion. Similarly, in the past few years, a brand new class of high-redshift objects has been discovered, called “Little Red Dots” (LRDs), many of which show evidence for accreting SMBHs (10⁶ − 10⁸ M⊙) hosted in highly compact galaxies approximately 50-100 parsecs in size. 

To address the potential formation of these objects, we have introduced a model of self-interacting dark matter (SIDM), where a small fraction of the dark matter is ultra-strongly self-interacting (uSIDM). The simple dark-sector model realizes both a SIDM and a uSIDM component while reproducing the observed dark matter relic abundance. Typical uSIDM cross-sections are on the order of 10³ − 10⁴ cm²/g, several orders of magnitude larger than those typically considered for SIDM. With such high cross-sections, the uSIDM undergoes rapid gravothermal evolution leading to a full core-collapse of the innermost portion of the dark matter halo. The rapid timescale for this collapse allows for the formation of a heavy SMBH seed well within the available formation timescales; thus uSIDM is a natural candidate for the formation of LRDs. Subsequent sub-Eddington accretion grows the seeds to match the observed SMBH masses and LRD mass function, while the model also makes a falsifiable prediction for the clustering bias of LRDs at z ∼ 5.