A comprehensive understanding of the characteristics of soil phosphorus (P) fractions and their driving factors is essential for enhancing the productivity of subtropical forest ecosystems. However, standardized quantification of soil P fractions across different layers in diverse plantation types remains limited. Additionally, the drivers of soil P fraction transformations, particularly microbial factors, are poorly understood. This study investigated four representative plantation ecosystems, including Chinese fir (Cunninghamia lanceolata), Eucalyptus (Eucalyptus robusta), Moso bamboo (Phyllostachys edulis), and Masson pine (Pinus massoniana). Utilizing a modified Hedley P fractionation scheme, we systematically analyzed the spatial differentiation patterns of soil P fractions across three soil layers: surface (0-20 cm), intermediate (20-40 cm), and deep (40-60 cm). Combined with quantitative real-time PCR (qPCR) to characterize functional gene expression profiles of soil microbial communities and comprehensive measurements of soil physicochemical properties, this research elucidates the biogeochemical mechanisms driven by microbial activity that govern P transformation in distinct soil horizons of plantation ecosystems. Occluded P and organic P dominated soil P fractions (> 70.0% across all four plantations), with surface soils exhibiting significantly higher P fractions, microbial biomass carbon, microbial biomass nitrogen, microbial biomass P, and copy numbers of the phoX gene compared to deep soils. Moso bamboo plantations exhibited higher contents of available P (5.6-40.3 mg/kg) and secondary mineral P (8.7-57.5 mg/kg) and higher proportions (proportion of available P: 3.9%-6.7%; proportion of secondary mineral P: 3.3%-9.9%) in each soil layer compared to other plantations (P<0.05). Chinese fir plantations exhibited the highest content of primary mineral P (surface layer: 23.4 mg/kg; middle layer: 17.5 mg/kg; deep layer: 17.0 mg/kg), while eucalyptus plantations showed the highest content of occluded P (surface layer: 632.2 mg/kg; middle layer: 585.7 mg/kg; deep layer: 535.2 mg/kg), and Masson pine plantations showed the lowest content of available P (surface layer: 8.8 mg/kg; middle layer: 6.9 mg/kg; deep layer: 6.8 mg/kg). Microbial biomass P and the phoX gene significantly affected soil P fractions, acid P activity and the phoC gene were significantly correlated with soil available P (P<0.05). These results indicated that the soil P supply capacity in Moso bamboo plantations was significantly superior to that in Chinese fir, eucalyptus, and Masson pine plantations. Soil layer merged as a key factor affecting soil P fractions. We thus recommend implementing species-specific management practices for different plantations and prioritizing the soil layer as a regulatory role in regulating P cycle.