Soil microorganisms dominate terrestrial biogeochemical cycles; however, we know very little about their spatial distribution and how changes in the distributions of specific groups of microbes translate into landscape and global patterns of biogeochemical processes. In this paper, we use a nested sampling scheme at scales ranging from 2 to 2,000 m to show that bacteria have significant spatial autocorrelation in community composition up to a distance of 240 m, and that this pattern is driven by changes in the relative abundance of specific bacterial clades across the landscape. Analysis of clade habitat distribution models and spatial co-correlation maps identified soil pH, plant abundance and snow depth as major variables structuring bacterial communities across this landscape, and revealed an unexpected and important oligotrophic niche for the Rhodospirillales in soil. Furthermore, our global analysis of high-elevation soils from the Andes, Rockies, Himalayas and Alaskan range shows that habitat distribution models for bacteria have a strong predictive power across the entire globe.