This dataset contains density functional theory (DFT) computations on point defects and impurities in several technologically relevant zinc blende semiconductors belonging to groups IV, III-V, and II-VI. Calculation folders include bulk semiconductor supercell calculations, calculations on defect-containing supercells in 7 different charge states, as well as calculations on reference elemental standard states and compounds. Shell and python-based scripts are applied on the converged calculation output files to determine the complete charge-, chemical potential-, and Fermi level-dependent formation energies of ~ 2000 native point defects and extrinsic impurity atoms across a total of 34 compounds, resulting in the largest computational defect dataset to date. This data forms the basis of machine learning models developed to facilitate quick prediction and screening of consequential impurities in semiconductors for a variety of optoelectronic applications.