The activities of soil ecoenzymes in relation to the mineralization of soil organic carbon, nitrogen, and phosphorus can reveal stoichiometric and energy limitations on the growth and metabolism of soil microbes. Despite the fact that arid and semi-arid ecosystems have received considerable attention in recent years, studies on the patterns of soil ecoenzymatic stoichiometry and the relationships among soil physiochemical and microbial community factors are still rare. In this study, we investigated the variations of soil physiochemical properties, microbial community composition, and EEAs by comparing soils in dominant shrubs and in the shrub-interspaces at four sites (Lianghekou, Feihong, Cuoji, and Moutuo). These study sites were expected to differ with respect to the level of water stress along the semi-arid valley in the upper Minjiang River, Sichuan province. The patterns of ecoenzymatic stoichiometry of C : N : P were different in the semi-arid Minjiang River valley as compared to those in the global pattern (C : N : P=3 : 2 : 2 vs. 1 : 1 : 1). Overall, there may be some degree of nitrogen limitation of microbial metabolism in this semi-arid region, which was supported by the relative activities of β-glucosidase compared to β-N-acetylglucosaminidase + leucine aminopeptidase and β-N-acetylglucosaminidase + leucine aminopeptidase to phosphatase. The different patterns of enzymatic stoichiometry between the soils in shrubs and in the shrub interspaces were not obvious. However, the limitation of C, N, and P resources in each individual site was somewhat site specific and may be due to the soil water availability. The two relatively drier sites (Lianghekou and Feihong) were more constrained by nitrogen availability, while the two relatively wetter sites (Cuoji and Moutuo) were more constrained by phosphorus availability. The variations of soil enzymatic stoichiometric patterns were affected by nutrient availabilities and soil water content as shown by the Pearson product-moment correlation analysis. The results of stepwise regression further revealed that the variations of enzyme ratios, such as C : N and N : P, were driven primarily by soil water content. Soil water content was positively correlated with the C : N enzyme ratio and negatively correlated with the N : P enzyme ratio. We also observed that arbuscular mycorrhizal fungi (AMF) and the fungi-to-bacteria ratio positively correlated with the C : N enzyme ratio, but negatively with the N : P enzyme ratio, which may imply that the enrichment of fungi and AMF communities will benefit C and P cycling by influencing enzyme activities. Our results suggest that any predicted decrease in soil water with climate change will probably further constrain the availability of N and P in this N-limited ecosystem in the semi-arid region. Moreover, this study also highlights the important role of fungi and AMF communities in meditating C and P cycling in this semi-arid region.