Abstract Precipitation extremes are expected to intensify under climate warming, but substantial uncertainty remains in quantifying this intensification. A widely used approach, Clausius‐Clapeyron (CC) scaling, is limited by an inexplicit definition of extremes and by its neglect of factors other than local temperature. We propose a framework using the Metastatistical Extreme Value Distribution that precisely defines the extremes being analyzed, and explores, through a large observational data set, their dependence on local thermodynamics and large‐scale atmospheric circulation at different temporal scales. Our results show that thermodynamics predominantly controls changes in hourly precipitation extremes and that their rate of increase must necessarily depend on the return period, in contrast with CC‐scaling arguments. On the contrary, daily precipitation is seen to be mostly controlled by large‐scale circulation, in ways that cannot be captured by CC‐based approaches. These findings clarify the physical mechanisms responsible for future changes in hydrologic extremes and possible methods to quantify them.