Soil extracellular enzyme activity (EEA) and enzyme stoichiometry are indicators of soil nutrient availability and microbial substrate limitation. However, the patterns in soil enzyme activities and stoichiometry, and their key drivers remain unclear during vegetation restoration in degraded sandy land. Here, soil ecosystems at five restoration stages of degraded sandy land were selected for investigation and we quantified the microbial nutrient limitation using EEA stoichiometry. The average C:N:P ratio for the nutrient-acquisition enzymes was 1:0.88:0.86, which differed from the global ratio of 1:1:1, indicating stronger microbial C limitation than N or P limitation. With vegetation restoration, the enzyme N:P ratio decreased, but the angle term used in vector analysis increased, indicating that the restored grassland transitioned from N-restricted (angle < 45°) to P-restricted (angle > 45°). Plant C inputs increased the soil nutrient content and significantly decreased the microbial C and N limitation but increased P limitation due to nutrient competition between plants and soil microorganisms. Decreased soil water levels caused by vegetation transpiration may have shifted the microbial limitation from N to P. The decreased fungi:bacteria ratio strengthened the microbial C and N or P limitations. On the basis of these findings, as measures to alleviate the associated nutrient limitations, we would recommend supplementation with phosphorus in the potential and slight stages of rocky desertification and the supplementary application of nitrogen in the moderate and severe stages during the restoration of degraded sandy ecosystems.