The amount of atmospheric nitrogen (N) deposited on the land surface has increased globally and by nearly five times in China from 1901 to 2005. Little is known about how elevated reactive N input has affected the carbon (C) sequestration capability of China's terrestrial ecosystems, largely due to the lack of reliable data on N deposition. Here we have used a newly developed data set of historical N deposition at a spatial resolution of 10 km × 10 km in combination with other gridded historical information on climate, atmospheric composition, land use, and land management practices to drive a process-based ecosystem model, the dynamic land ecosystem model (DLEM) for examining how increasing N deposition and its interactions with other environmental changes have affected C fluxes and storage in China's terrestrial ecosystems during 1901–2005. Our model simulations indicate that increased N deposition has resulted in a net C sink of 62 Tg C/yr (1 Tg = 1012 g) in China's terrestrial ecosystems, totaling up to 6.51 Pg C (1 Pg = 1015 g) in the past 105 years. During the study period, the N-induced C sequestration can compensate for more than 25% of fossil-fuel CO2 emission from China. The largest C sink was found in southeast China, a region that experienced the most significant increase of N deposition in the period 1901–2005. However, the net primary productivity induced by per-unit N deposition (referred to as ecosystem N use efficiency, ENUE, in this paper) has leveled off or declined since the 1980s. This indicates that part of the deposited N may not be invested to stimulate plant growth, but instead leave the ecosystem by various pathways. Except shrubland and northwest/southwest China, signs of N saturation are apparent in the rest major biome types and regions, with ENUE peaking in the 1980s and leveling off or declining thereafter. Therefore, to minimize the excessive N pollution while keeping the N-stimulated C uptake in China's terrestrial ecosystems, optimized management practices should be taken to increase N use efficiency rather than to keep raising N input level in the near future.