Sup3rLoad is a dataset of hourly electricity load for the contiguous United States that varies jointly with electrification and weather. It spans 2 electrification scenarios, 5 CMIP6 Earth system models (Sup3rCC, SSP2-4.5), 6 scenario years between 2025 and 2050, 60 weather years (2000 through 2059), 14 end-use subsectors, and 48 states plus the District of Columbia, stored in a NetCDF4 file. Hourly profiles for weather-sensitive subsectors (heating, cooling, light-duty electric vehicle charging, and data center cooling) were produced by training neural network regressions on building-energy simulations, applying them to downscaled weather projections, and integrating the results through a bottom-up stock-turnover energy-demand model calibrated to state-level electricity sales and hourly system load patterns.