Wildfires destabilize biocrust, requiring decades for most biological constituents to regenerate, but bacteria may recover quickly and mitigate the detrimental consequences of burnt soils. To evaluate the short-term recovery of biocrust bacteria, we tracked shifts in bacterial community form and function in Cyanobacteria/lichen-dominated (shrub interspaces) and Cyanobacteria/moss-dominated (beneath Artemisia tridentata) biocrusts 1 week, 2 months, and 1 year following a large-scale burn manipulations in a cold desert (Utah, USA). We found no evidence of the burned bacterial community recovering to a burgeoning biocrust. The foundational biocrust phyla, Cyanobacteria, dominated by Microcoleus viginatus (Microcoleaceae), disappeared from burned soils creating communities void of photosynthetic taxa. One year after the fire, the burned biocrust constituents had eroded away and the bare soils supported the formation of a convergent community of chemoheterotrophic copiotrophs regardless of location. The emergent community was dominated by a previously rare Planococcus species (family Planococcaceae, Firmicutes) and taxa in the Cellulomonadaceae (Actinobacteria), and Oxalobacteraceae (Betaproteobacteria). Previously burnt soils maintained similar levels of bacterial biomass, alpha diversity, and richness as unburned biocrusts, but supported diffuse, poorly-interconnected communities with 75% fewer species interactions. Nitrogen fixation declined at least 3.5-fold in the burnt soils but ammonium concentrations continued to rise through the year, suggesting that the exhaustion of organic C released from the fire, and not N, may diminish the longevity of the emergent community. Our results demonstrate that biocrust bacteria may recover rapidly after burning, albeit along a different community trajectory, as rare bacteria become dominant, species interconnectedness diminishes, and ecosystem services fail to rebound. Keywords: biological soil crust, Bromus tectorum, disturbance, Great Basin Desert, rare biosphere, network co-occurrence model.