Grassland restoration through grazing exclusion is a key strategy for enhancing soil organic carbon (SOC) sequestration, yet the dynamic contributions of plant- versus microbial-derived carbon (C) remain incompletely understood. We hypothesized that with increasing restoration duration, microbial-derived C would become a major contributor to SOC relative to plant-derived C, and that the relative proportion of bacterial necromass would increase compared to fungal necromass. To explore this, we investigated a 25-year restoration chronosequence (3, 10, 19, 25 years) of a degraded typical steppe on Kastanozem soil in Inner Mongolia, China. While acknowledging the inherent limitations of a space-for-time substitution approach, such as potential unquantified variations in initial pre-enclosure soil conditions and plant species composition, we used lignin phenols, amino sugars, and PLFA analysis to estimate the dynamics of plant- and microbial-derived C. Grassland restoration was associated with significant increases in total PLFAs (15.4-58.8%), bacterial PLFAs (14.5-82.4%), lignin phenols (16.9-91.8%), and estimated microbial-derived C (5.0-8.8 g kg-1). Based on these specific biomarker estimates, which track only a subset of total C and do not equal 100% of the SOC pool, microbial-derived C accounted for 52.8-63.3% of SOC, compared to 10.1-15.5% for plant-derived C. Within the estimated microbial-derived C, the bacterial fraction increased over the restoration chronosequence, while the fungal fraction declined. Correlational analyses, including structural equation modeling, indicated that soil pH, bulk density, SOC, and microbial biomass were key factors closely associated with both C sources. Our findings suggest that microbial-necromass C, particularly from bacteria, is a major contributor to SOC accumulation during long-term grassland restoration in this semi-arid typical steppe, and that grazing exclusion can enhance SOC sequestration under the studied conditions and biomarker-based estimations.