Pathogens and parasites of solitary bees have been studied for decades, but the microbiome as a whole is poorly understood for most taxa. Comparative analyses of microbiome features such as composition, abundance, and specificity, can shed light on bee ecology and the evolution of host–microbe interactions. Here we study microbiomes of ground-nesting cellophane bees (Colletidae: Diphaglossinae). From a microbial point of view, the diphaglossine genus Ptiloglossa is particularly remarkable: their larval provisions are liquid and smell consistently of fermentation. We sampled larval provisions and various life stages from wild nests of Ptiloglossa arizonensis and two species of closely related genera: Caupolicana yarrowi and Crawfordapis luctuosa. We also sampled nectar collected by P. arizonensis. Using 16S rRNA gene sequencing, we find that larval provisions of all three bee species are near-monocultures of lactobacilli. Nectar communities are more diverse, suggesting ecological filtering. Shotgun metagenomic and phylogenetic data indicate that Ptiloglossa culture multiple species and strains of Apilactobacillus, which circulate among bees and flowers. Larval lactobacilli disappear before pupation, and hence are likely not vertically transmitted, but rather reacquired from flowers as adults. Thus, brood cell microbiomes are qualitatively similar between diphaglossine bees and other solitary bees: lactobacilli-dominated, environmentally acquired, and non-species-specific. However, shotgun metagenomes provide evidence of a shift in bacterial abundance. As compared with several other bee species, Ptiloglossa have much higher ratios of bacterial to plant biomass in larval provisions, matching the unusually fermentative smell of their brood cells. Overall, Ptiloglossa illustrate a path by which hosts can evolve quantitatively novel symbioses: not by acquiring or domesticating novel symbionts, but by altering the microenvironment to favor growth of already widespread and generalist microbes.