The spread among the predictions by climate models for the strengthening of the global hydrological cycle [i.e., the global mean surface latent heat flux (LH), or, equivalently, precipitation] at a given level of CO2-induced global warming is of the same magnitude as the intermodel mean. By comparing several climate models from the World Climate Research Programme (WCRP) Coupled Model Intercomparison Project phase 3 (CMIP3) database under idealized CO2 forcings, it is shown that differences in the increase in global atmospheric shortwave heating (SWabs) induced by clear-sky absorption, presumably by water vapor, partly explains this spread. The increases in SWabs and LH present similar spreads across models but are anticorrelated, so the sumSWabs1LHincreases more robustly than either alone. This is consistent with a recently proposed theory (Takahashi) that predicts that this sum (or, equivalently, the net longwave divergence minus the surface sensible heat flux) is constrained by energy conservation and robust longwave physics. The intermodel scatter in SWabs changes is explained neither by differences in the radiative transfer models nor in intermodel differences in global water vapor content change, but perhaps by more subtle aspects of the changes in the water vapor distribution. Nevertheless, the fact that the radiative transfer models generally underestimate the increase in SWabs relative to the corresponding line-by-line calculation for a given change in water vapor content suggests that the climate models might be overestimating the rate of increase in the global hydrological cycle with global warming. © 2009 American Meteorological Society.