Desert steppes are fragile ecosystems, which generally suffer severe wind erosion hazards, especially when they experiencing external disturbances such as grazing. Therefore, the effects of grazing on the wind erosion process are very important for improving our understanding of the effects of the stocking rate on grassland degradation and enacting reasonable grazing strategies in grasslands. However, few studies have discussed the wind erosion changes induced by different grazing intensities thus far. Herein, a series of grazing experiments was carried out in a desert steppe in Urad Back Banner, Inner Mongolia. These experiments included 2 types of grasslands that are shrub dominated grassland (SDG) and grass dominated grassland (GDG). Each grassland type had 3 grazing rate treatments: no grazing as the control (CK), moderate grazing (MG), and heavy grazing (HZ). After the grazing experiments lasting for 4 years, sand traps were set in the grazing plots to detect the horizontal sand flux at different grazing intensities. In addition to the wind erosion observed, the vegetation and soil particles were also surveyed and sampled. Our results showed that the vegetation cover (VC) and aboveground biomass (AGB) in the MG and HG plots all had obvious differences compared to the CK plot, regardless of whether it was SDG or GDG. The soil particle analysis showed that the fractions of erodible particles (60–200 μm) in the MG and HG plots decreased significantly. In the GDG, the soil particles exhibited a coarsening trend from CK to HG plots for the sorting action of the wind. Suffering a grazing disturbance, the wind erosion in a desert steppe exhibited significant changes. For instance, in the SDG, the horizontal sand flux in the MG and HG plots were 1.78 and 2.06 times the sand flux in the CK plots, respectively, and, in the GDG, they were 17.86 and 56.53 times the flux in the CK plots, respectively. Except in the MG and HG plots of the GDG, the saltation height of sand was generally lower than 20 cm. In the SDG, the particle size of the entrained sand mainly ranged from 80 to 200 μm, and, in the GDG, it ranged from 80 to 300 μm. This finding also indicated the validity of Shao’s sand entrainment equation. We also found that particle size of saltating sand increased with their saltation height for the tough surface of the MG and HG plots in the GDG. Based on the different fractions of horizontal flux in different directions, we also discussed the threshold wind velocity for sand entrainment, and we confirmed that the threshold wind velocities for no grazing in SDG and GDG were approximately 16.5 m·s−1 and 15 m·s−1, respectively. Using our previous wind erosion model, we estimated the dust emission in the total desert steppe of China and found that different grazing scenarios would change the magnitude and distribution of dust emissions.