Human activities have induced land degradation, loss of biodiversity, and the decline of the functions and services of ecosystems on the Loess Plateau. Vegetation restoration is an effective way to improve degraded ecosystems, and afforestation is an important measure to restore degraded land. Existing studies on ecological restoration have typically focused on the relationship between species diversity of microbial community and a single ecosystem function, neglecting the relationships between the interactions among microbial communities and ecosystem multifunctionality (EMF). In order to investigate the relationships of soil microbial biodiversity and the complexity of microbial networks with EMF in afforestation restoration, in this study, we used the space-time substitution method to investigate the effects of afforestation restoration on soil microbial diversity, soil microbial network complexity, and 10 variables of ecosystem function related to soil nutrient cycling and plant productivity along a 50-year afforestation chronosequence, with an aim to the relationships between the soil microbial community characteristics and EMF. The results indicated that microbial multi-diversity and network complexity of the soil microbial community increased significantly with the increase of time series of afforestation restoration, reaching a peak in the eighth year, and then decreased (P<0.05); EMF showed a trend of increasing and then decreasing, reaching a maximum in the twentieth year (P<0.05). With soil environmental factors not controlled, bacterial and archaeal diversity was not significantly correlated with EMF, while fungal diversity was significantly positively correlated with EMF (P<0.001). Both the soil microbial multi-diversity and network complexity showed significant positive correlations with EMF (P<0.001), with the soil microbial network complexity explaining more about EMF than microbial multi-diversity. Although the multi-diversity and network complexity of soil microorganisms explained less about EMF after controlling the effect of soil factors by partial least squares regression analysis (P<0.001), the microbial network complexity explained more about EMF than microbial multi-diversity. Furthermore, structural equation model (SEM) showed that soil microbial diversity had an indirect effect on EMF in afforested restoration sites through soil microbial network complexity. In summary, the microbial network complexity was a more effective predictor of EMF than soil microbial multi-diversity. Soil microbial multi-diversity positively influenced EMF in an indirect manner, mainly by contributing to network complexity. Our study revealed the relationship between microbial community characteristics and EMF in afforestation restoration ecosystems, providing a theoretical basis for improving the restoration of degraded restoration ecosystem functions.