We examine the ability of atmospheric general circulation models to reproduce seasonal and interannual variations in the distribution of moisture and moisture fluxes over the continental United States, using results from the second Atmospheric Model Intercomparison Project. Currently, output from 20 models is available from the AMIP-2 experiment, which covers the period 1979-1995. The model moisture flux divergence is calculated from the difference between evaporation and precipitation fields using the water balance equation, and it is compared against the vapor divergence in the NCEP-NCAR reanalysis. Concerning seasonal variability, summer is the only season in which evaporation exceeds precipitation in the models over the conterminous US, leading to an overall moisture flux divergence in the models that is also evident in the reanalysis. Relevant to the North American Monsoon Experiment, model and reanalysis results indicate a northward advance of moisture flux divergence from May to August, with a subsequent retreat southward by October. Model precipitation has a similar signature, though it fails to reach as far north as it does in both reanalysis and the Global Precipitation Climatology Project. We place the large-scale moisture budgets of the United States in a global context by using model-produced precipitable water, evaporation, and precipitation to examine the cycling times of moisture into and out of the atmosphere. We use these moistening and drying rates to examine model-based estimates of the intensity of the hydrological cycle. Preliminary results for the globe indicate that the mean ensemble global water vapor residence time is 8.1 days, though it can range from around 7 to 10 days, depending on model. The median global averaged climatological precipitable water is 24.4 mm, and precipitation and evaporation are near 3.0 mm/day. A seasonal signal and an El Niņo-associated interannual signal in both these quantities are evident in most models. The physical parameterizations that may influence the performance of individual GCMs in the simulations of the various hydrologic cycle parameters over the United States and the globe are examined.