The sequencing-driven revolution in microbial ecology unveiled the complexity of microbial communities that shape the health of our planet, and our own. Despite the ever-growing amount of data, our understanding of these systems remains extremely limited. Natural communities routinely harbor hundreds of species, and the intuition derived from analysis of low-dimensional models (with few species) may not extend to the high-dimensional case (many species, metabolites). Remarkably, the high-diversity limit of a classic resource competition model, introduced by MacArthur in 1969, proves analytically tractable and provides a platform for investigating the implications of high dimensionality for both ecological and evolutionary dynamics. At high diversity, this classic model exhibits a phase transition into a curious collective regime, leading to an effective “cohesion” of the community even in the absence of any cooperative interactions. These results highlight the promise of using statistical physics of disordered systems to go beyond the intuitive picture of discrete species and individual organisms, bringing our theory up to speed with the recent data challenging these foundational notions.