The accurate simulation and prediction of grassland aboveground biomass (AGB) and theoretical livestock carrying capacity are key steps for maintaining ecosystem balance and sustainable grassland management. The AGB in fenced grassland is not affected by grazing and its variability is only driven by climate change, which can be regarded as the grassland potential AGB (AGB_p). In this study, we compiled the data for 345 AGB field observations in fenced grasslands and their corresponding climate data, soil data, and topographical data on the Qinghai-Tibetan Plateau (TP). We further simulated and predicted grassland AGB_p and theoretical livestock carrying capacity under the climate conditions of the past (2000–2018) and future two decades (2021–2040) based on a random forest (RF) algorithm. The results showed that simulated AGB_p matched well with observed values in the field (R² = 0.76, P < 0.001) in the past two decades. The average grassland AGB_p on the Tibetan Plateau was 102.4 g m^–2, and the inter-annual changes in AGB_p during this period showed a non-significant increasing trend. AGB_p fluctuation was positively correlated with growing season precipitation (R² = 0.57, P < 0.001), and negatively correlated with the growing season diurnal temperature range (R² = 0.51, P < 0.001). The average theoretical livestock carrying capacity was 0.94 standardized sheep units (SSU) ha^–1 on the TP, in which about 54.1% of the areas showed an increasing trend during the past two decades. Compared with the past two decades, the theoretical livestock carrying capacity showed a decreasing trend in the future, which was mainly distributed in the central and northern TP. This study suggested that targeted planning and management should be carried out to alleviate the forage-livestock contradiction in grazing systems on the Tibetan Plateau.