The holobiont approach proposes that species are most fully understood within the context of their associated microbiomes, and that both host and microbial community are locked in a mutual circuit of co-evolutionary selection. Bees are an ideal group for this approach, as they comprise a critical group of pollinators that contribute to both ecological and agricultural health worldwide. Metagenomic analyses offer comprehensive insights into an organism’s microbiome, diet, and viral load, but remain largely unapplied to wild bees. Here, we present metagenomic data from three species of carpenter bees sampled from around the globe, representative of the first ever carpenter bee core microbiome. Machine learning, co-occurrence, and network analyses reveal that wild bee metagenomes are unique to host species. Further, we find that microbiomes are likely strongly affected by features of their local environment, and feature evidence of plant pathogens previously known only in honey bees. Performing the most comprehensive comparative analysis of bee microbiomes to date we discover that microbiome diversity is inversely proportional to host species social complexity. Our study helps to establish some of the first wild bee hologenomic data while offering powerful empirical insights into the biology and health of vital pollinators.