This dataset provides estimates of aboveground biomass (AGB) for salt marshes in the contiguous United States for 2020 and includes all coastal watersheds across the contiguous United States at 10-m resolution. Estimates were generated by machine learning regression models, including XGBoost, random forests, and support vector machines, trained from in situ data within salt marshes identified by the National Wetland Inventory (NWI). The data are organized by Hydrologic United Code (HUC) 6-digit basin. Within each HUC, the spatial extent of salt marsh and its uncertainty were estimated by machine learning and input data from NWI maps, the National Elevation Dataset, along with Sentinel-1 and Sentinel-2 imagery. Estimates were compared to in situ biomass data from salt marshes in Georgia and Massachusetts. The data are provided in cloud-optimized GeoTIFF format.