The main objective of this paper is to study the impacts of climate change on the hydrological cycle of the main French river basins, including the different uncertainties at stake. In particular, the relative importance of modeling uncertainty versus that of downscaling uncertainty is investigated. An ensemble of climate scenarios are statistically downscaled in order to force a hydrometeorological model over France. Then, the main changes in different variables of the hydrological cycle are studied. Despite large uncertainties linked to climate models, some robust signals already appear in the middle of the 21st century. In particular, a decrease in mean discharges in summer and fall, a decrease in soil moisture, and a decrease in snow cover, especially pronounced at the low and intermediate altitudes, are simulated. The low flows become more frequent but generally weak, and uncertain changes in the intensity of high flows are simulated. To evaluate downscaling uncertainties and assess the robustness of the results obtained with the statistical downscaling method, two other downscaling approaches are used. The first one is a dynamical downscaling methodology based on a variable resolution atmospheric model, with a quantile‐quantile bias correction of the model variables. The second approach is based on the so‐called anomaly method, that simply consists of perturbing present climate observations by the climatological change simulated by global climate models. After hydrological modeling, some discrepancies exist among the results from the different downscaling methods. However they remain limited and to a large extent smaller than climate model uncertainties, which raises important methodological considerations.