By the end of this century, most climate models predict that the eastern equatorial Pacific (EEP) will warm more than the western Pacific—a pattern similar to El Niño. This shift weakens the Walker circulation, a key atmospheric system in the tropics. However, the causes of this warming pattern vary across models, and we are still unsure which mechanisms matter most. To investigate, we ran targeted climate experiments using two versions of a sophisticated climate model (CESM2) that differ only in their atmospheric physics. We imposed an increase in CO2 while holding surface wind and solar heating constant. The two models responded very differently: one (CAM6) showed a strong El Niño-like pattern, while the other (CAM4) showed a weaker response. This difference was traced to how the models simulate cloud feedbacks and surface evaporation. These results suggest that how climate models represent clouds and evaporation has a major impact on projections of Pacific warming, and could help explain why models differ in their predictions of a future El Niño-like warming pattern.