Cotton is the main cultivated cash crop in Northwest China. This study is motivated by the challenge of economically growing cotton in the face of climate change, which could lead to changes in cotton variety, phenology, water demand, heat requirement and yield. The APSIM-OzCot crop growth model is used to simulate cotton cultivation at two sites with different climatic conditions (Shihezi and Alaer, Xinjiang Province). The model is fully calibrated and validated using observations. Simulations forced with future climate data downscaled from the HadCM3 Global Climate Model show that the response of cotton phenology, yield and water use to climate change is different for different cultivation, sites, greenhouse gas emission scenarios and time horizons. Under the SRES A1B and B1 emissions scenarios, cotton yield and water use are greater in the future than in the 1961–1990 period while the growing season is shorter. Under the SRES A2 emissions scenario, however, yields at cold sites drop after 2070due mainly to the shortening of cotton growth periods. Thus, in the cold regions, varieties with short growth periods are replaced by those with long growth periods. The results show that, compared with current local varieties, cotton yields increase by 356 kg/ha with the medium maturity variety “K7” and473 kg/ha with the late maturity variety “ZM49” under the A2 scenario by 2070. Total evapotranspiration correspondingly increases by 69 mm (“K7”) and 92 mm (“ZM49”). However, water use efficiency increases by 0.32 kg/m3to 0.34 kg/m3(6.3%) and 0.35 kg/m3(9.4%) for “K7” and “ZM49”, respectively. A reasonable adaptive strategy to maintain cotton yields in the future may be to decrease the area over which cotton is planted and raise water use efficiency.