We present a comprehensive gridded GDP per capita dataset downscaled to the Admin 2 level (43,501 units) covering 1990–2022. It updates existing outdated datasets, which use reported subnational data only up to 2010. Our dataset, based on reported subnational GDP per capita data from 89 countries and 2,708 units, employed various novel methods for extrapolation and downscaling. Downscaling with machine learning algorithms showed high performance (R2 = 0.73 for test, R2 = 0.86 for entire dataset) and accuracy against reported datasets (Pearson R = 0.88). The dataset includes reported and downscaled annual data for three administrative levels—0 (national; 237 units), 1 (provincial), and 2 (municipality)—in polygon and 5 arc-min resolution raster formats. The dataset has a higher spatial resolution and wider temporal range than the existing data. We also provide total GDP by combining GDP per capita with population count. This new dataset can contribute to global or regional spatial analyses such as socioenvironmental modelling and economic resilience evaluation. The data are available in http://doi.org/10.5281/zenodo.10976734.