Growing agricultural water use, driven by cropland expansion and a warming climate, has substantially altered the Earth's hydrological cycle. Despite its importance, accurate estimation of crop evaporation at a global scale remains an outstanding scientific challenge. Here, we introduce CropET, a global dataset of daily evaporation from 23 irrigated and rainfed crops at 5 arc-minute spatial resolution for the years 2000–2025. The dataset is developed by combining a FAO-56-based model with recent developments in dynamic crop mapping, satellite-derived precipitation, and reference evaporation based on atmospheric reanalysis. Our dataset compares favorably with existing products with a global crop evaporation estimate of 5,531 km3 against 5,868 km3 from WATNEEDS for the year 2000. It also shows a median correlation of 0.77 and Kling-Gupta Efficiency of 0.64 when validated against 10 eddy covariance sites. Given its explicit representation of temporal variability in crop areas and observation-based climate inputs, CropET will enable a wide range of applications, including benchmarking Earth system models, assessing irrigation withdrawals, and quantifying how agricultural expansion has reshaped regional water and energy budgets.