Earth's hydrological cycle is expected to intensify in response to global warming, with a wet-gets-wetter, dry-gets-drier response anticipated over the ocean. Subtropical regions (similar to 15 degrees-30 degrees N/S) are predicted to become drier, yet proxy evidence from past warm climates suggests these regions may be characterized by wetter conditions. Here we use an integrated data-modeling approach to reconstruct global and zonal-mean rainfall patterns during the early Eocene (similar to 56-48 million years ago). The Deep-Time Model Intercomparison Project (DeepMIP) model ensemble indicates that the mid-(30 degrees-60 degrees N/S) and high-latitudes (>60 degrees N/S) are characterized by a thermodynamically dominated hydrological response to warming and overall wetter conditions. The tropical band (0 degrees-15 degrees N/S) is also characterized by wetter conditions, with several DeepMIP models simulating narrowing of the Inter-Tropical Convergence Zone. However, the latter is not evident from the proxy data. The subtropics are characterized by negative precipitation-evaporation anomalies (i.e., drier conditions) in the DeepMIP models, but there is surprisingly large inter-model variability in mean annual precipitation (MAP). Intriguingly, we find that models with weaker meridional temperature gradients (e.g., CESM, GFDL) are characterized by a reduction in subtropical moisture divergence, leading to an increase in MAP. These model simulations agree more closely with our new proxy-derived precipitation reconstructions and other key climate metrics and imply that the early Eocene was characterized by reduced subtropical moisture divergence. If the meridional temperature gradient was even weaker than suggested by those DeepMIP models, circulation-induced changes may have outcompeted thermodynamic changes, leading to wetter subtropics. This highlights the importance of accurately reconstructing zonal temperature gradients when reconstructing past rainfall patterns. As the world warms, the atmosphere is able to hold more moisture however, this moisture will not fall evenly across the globe. Some regions are expected to become wetter, whereas other regions will become drier. This is the basis of the familiar paradigm wet-gets-wetter, dry-gets-drier and is largely supported by future model projections. However, evidence from the geological record contradicts this hypothesis and suggests that a warmer world could be characterized by wetter (rather than drier) subtropics. Here, we use an integrated data-modeling approach to investigate the hydrological response to warming during an ancient warm interval (the early Eocene, 56-48 million years ago). We show that models with weaker latitudinal temperature gradients are characterized by a reduction in subtropical moisture divergence. However, this was not sufficient to induce subtropical wetting. If the meridional temperature gradient was weaker than suggested by the models, circulation-induced changes may have lead to wetter subtropics. This work shows that the latitudinal temperature gradient is a key factor that influences hydroclimate in the subtropics, especially in past warm climates.