Legume–grass intercropping and phosphorus (P) fertilization are recognized strategies for enhancing forage productivity, but their interactive effects on soil microbial processes and plant phosphorus nutrition in a semi-arid climate remain poorly understood. We conducted a field experiment with common vetch (Vicia sativa) and oat (Avena sativa) under two monocultures and three intercropping treatments (legume–grass ratios of 1:3, 2:3, and 1:1), combined with three P fertilization rates (0, 60, and 120 kg P ha−1). The results showed that common vetch/oat intercropping with moderate legume–grass proportions (1:3 and 2:3) significantly outyielded monocultures across all P levels and exhibited a stronger net biodiversity effect than the 1:1 intercropping at P fertilization. Plant P concentration was primarily increased by P fertilization. Crucially, all intercropping treatments showed a significantly lower microbial biomass carbon/phosphorous ratio than the monoculture in the absence of P fertilization. However, this difference disappeared when P was applied, indicating P fertilization overrode the intercropping-induced stoichiometric shift. Correlation analyses further showed that forage yield and plant P uptake were positively linked to microbial biomass P and negatively to the microbial biomass carbon/phosphorous ratio. Together, our findings reveal that a common vetch/oat intercropping system combined with P fertilization may improve the nutrient use efficiency through microbial pathway. This improvement in nutrient use efficiency leads to higher nutrient uptake by plants, thereby causing more rapid soil reserve depletion.