This study aims to evaluate the effect of different afforestation models on soil microbial composition in the Loess Plateau in China. In particular, we determined soil physicochemical properties, enzyme activities, and microbial community structures in the top 0 cm to 10 cm soil underneath a pure Hippophae rhamnoides (SS) stand and three mixed stands, namely, H. rhamnoides and Robinia pseucdoacacia (SC), H. rhamnoides and Pinus tabulaeformis (SY), and H. rhamnoides and Platycladus orientalis (SB). Results showed that total organic carbon (TOC), total nitrogen, and ammonium (NH4(+)) contents were higher in SY and SB than in SS. The total microbial biomass, bacterial biomass, and Gram+ biomass of the three mixed stands were significantly higher than those of the pure stand. However, no significant difference was found in fungal biomass. Correlation analysis suggested that soil microbial communities are significantly and positively correlated with some chemical parameters of soil, such as TOC, total phosphorus, total potassium, available phosphorus, NH4(+) content, nitrate content (NH3(-)), and the enzyme activities of urease, peroxidase, and phosphatase. Principal component analysis showed that the microbial community structures of SB and SS could clearly be discriminated from each other and from the others, whereas SY and SC were similar. In conclusion, tree species indirectly but significantly affect soil microbial communities and enzyme activities through soil physicochemical properties. In addition, mixing P. tabulaeformis or P. orientalis in H. rhamnoides plantations is a suitable afforestation model in the Loess Plateau, because of significant positive effects on soil nutrient conditions, microbial community, and enzyme activities over pure plantations.