It is critical to identify the community assembly patterns (i.e., deterministic or stochastic processes) of soil microbes and the potential driving factors to better predict the belowground biodiversity and functioning in forest ecosystems. Here, a combined approach of neutral model and multivariate analysis was employed to examine the soil bacterial communities in 12 undisturbed forests in China, spanning a wide latitudinal range from 21.6°N to 50.8°N. A clear divergent pattern was found for community composition, indicating that deterministic processes dominated the community assembly of soil bacteria. The α-diversity (richness) nonlinearly changed from tropical to cold temperate zones, with the lowest and highest values detected in subtropical and temperate zones, respectively. Although no latitudinal pattern was observed for β-diversity (community variation), there were clear climate zone patterns. Unlike the minor effects of mean annual precipitation (MAP) and temperature (MAT) on bacterial α-diversity, MAP and MAT were important factors affecting soil bacterial β-diversity. Soil pH was a strong driver of α- and β-diversity, but plant factors had only minor effects. Altogether, this study highlights the unexpected importance of climatic factors in shaping bacterial β-diversity in forest soils. Our findings have implications for future investigations of bacterial community dynamics in forest ecosystems, particularly the responses of community composition to global climate change scenarios across large geographical scales.