Extreme arid regions in the worlds' major deserts are typified by quartz pavement terrain. Cryptic hypolithic communities colonize the ventral surface of quartz rocks and this habitat is characterized by a relative lack of environmental and trophic complexity. Combined with readily identifiable major environmental stressors this provides a tractable model system for determining the relative role of stochastic and deterministic drivers in community assembly. Through analyzing an original, worldwide data set of 16S rRNA-gene defined bacterial communities from the most extreme deserts on the Earth, we show that functional assemblages within the communities were subject to different assembly influences. Null models applied to the photosynthetic assemblage revealed that stochastic processes exerted most effect on the assemblage, although the level of community dissimilarity varied between continents in a manner not always consistent with neutral models. The heterotrophic assemblages displayed signatures of niche processes across four continents, whereas in other cases they conformed to neutral predictions. Importantly, for continents where neutrality was either rejected or accepted, assembly drivers differed between the two functional groups. This study demonstrates that multi-trophic microbial systems may not be fully described by a single set of niche or neutral assembly rules and that stochasticity is likely a major determinant of such systems, with significant variation in the influence of these determinants on a global scale.