The cold grasslands of the Qinghai‐Tibetan Plateau form a globally significant biome, which represents 6% of the world's grasslands and 44% of China's grasslands. Yet little is known about carbon cycling in this biome. In this study, we calibrated and applied a process‐based ecosystem model called Organizing Carbon and Hydrology in Dynamic Ecosystems (ORCHIDEE) to estimate the C fluxes and stocks of these grasslands. First, the parameterizations of ORCHIDEE were improved and calibrated against multiple time‐scale and spatial‐scale observations of (1) eddy‐covariance fluxes of CO2 above one alpine meadow site; (2) soil temperature collocated with 30 meteorological stations; (3) satellite leaf area index (LAI) data collocated with the meteorological stations; and (4) soil organic carbon (SOC) density profiles from China's Second National Soil Survey. The extensive SOC survey data were used to extrapolate local fluxes to the entire grassland biome. After calibration, we show that ORCHIDEE can successfully capture the seasonal variation of net ecosystem exchange (NEE), as well as the LAI and SOC spatial distribution. We applied the calibrated model to estimate 0.3 Pg C yr−1 (1 Pg = 1015 g) of total annual net primary productivity (NPP), 0.4 Pg C of vegetation total biomass (aboveground and belowground), and 12 Pg C of SOC stocks for Qinghai‐Tibetan grasslands covering an area of 1.4 × 106 km2. The mean annual NPP, vegetation biomass, and soil carbon stocks decrease from the southeast to the northwest, along with precipitation gradients. Our results also suggest that in response to an increase of temperature by 2°C, approximately 10% of current SOC stocks in Qinghai‐Tibetan grasslands could be lost, even though NPP increases by about 9%. This result implies that Qinghai‐Tibetan grasslands may be a vulnerable component of the terrestrial carbon cycle to future climate warming.