We present the first comprehensive general circulation model study of water ice condensation and cloud formation in the Martian atmosphere. We focus on the effects of condensation in limiting the vertical distribution and transport of water and on the importance of condensation for the generation of the observed Martian water cycle. We do not treat cloud ice radiative effects, ice sedimentation rates are prescribed, and we do not treat interactions between dust and cloud ice. The model generates cloud in a manner consistent with earlier one‐dimensional (1‐D) model results, typically evolving a uniform (constant mass mixing ratio) vertical distribution of vapor, which is capped by cloud at the level where the condensation point temperature is reached. Because of this vertical distribution of water, the Martian atmosphere is generally very far from fully saturated, in contrast to suggestions based upon interpretation of Viking data. This discrepancy results from inaccurate representation of the diurnal cycle of air temperatures in the Viking Infrared Thermal Mapper (IRTM) data. In fact, the model suggests that only the northern polar atmosphere in summer is consistently near its column‐integrated holding capacity. In this case, the column amount is determined primarily by the temperature of the northern polar ice cap. Comparison of the water cycle generated by the model with and without atmospheric ice condensation and precipitation shows two major roles for water ice cloud. First, clouds are essential to the observed rapid return of atmospheric water to the surface in late northern summer, as ice sedimentation forces the water column to shrink in response to the downward motion of the condensation level, concentrating water near surface sinks. Second, ice sedimentation limits the amount of water that is transported between the hemispheres through the Hadley circulation. This latter effect is used to greatly improve the model simulation of the annual water cycle by increasing ice sedimentation rates. The model is thus shown to be able to reasonably reproduce the annual cycles of vapor and ice cloud as compared to Viking data. In addition, the model is shown able to reproduce near‐instantaneous maps of water ice derived from Hubble Space Telescope images. The seasonal evolution of the geographic distribution of water ice compares reasonably well with Viking and Mars Global Surveyor (MGS) Mars Orbiter Laser Altimeter (MOLA) observations, except in the prediction of a weak tropical cloud belt in southern summer. Finally, it is shown that the tropical cloud belt is generated in the model by the cooling of water vapor entrained in the upwelling branch of the Hadley cell. Decline of the tropical cloud belt in mid northern summer is shown to be related to an increase in air temperatures, rather than to decreases in water vapor supply or the vigor of Hadley cell ascent. By equinox, the cloud belt experiences a second major decline event, this time due to a reduction in vapor supply. The ability of the model to emulate many aspects of observed cloud behavior is encouraging, as is the ability of enhanced ice sedimentation to improve the overall quality of the water cycle simulation. However, significant work remains to be done before all observational constraints can be matched simultaneously. Specifically, in order for the generally good fit to all other data to be attained, cloud ice particle sizes about an order of magnitude too large must be used.