A key indicator of climate change is the greater frequency and intensity of precipitation extremes across much of the globe. In fact, several studies have already documented increased regional precipitation extremes over recent decades. Future projections of these changes, however, vary widely across climate models. Using two generations of models, here we demonstrate an emergent relationship between the future increased occurrence of precipitation extremes aggregated over the globe and the observable change in their frequency over recent decades. This relationship is robust in constraining frequency changes in precipitation extremes in two separate ensembles and under two future emissions pathways (reducing intermodel spread by 20–40%). Moreover, this relationship is also apparent when the analysis is limited to near-global land regions. These constraints suggest that historical global precipitation extremes will occur roughly 32 ± 8% more often than at present by 2100 under a medium-emissions pathway (and 55 ± 13% more often under high emissions).