High-quality direct normal irradiance (DNI) observations are of vital importance for the optimal design, installation and profitability of CST plants. Numerous models have been developed for estimating DNI at regional or global scales. The goal of this study is to generate a gridded DNI dataset for all-sky conditions over mainland China during 1981–2014, based on a broadband DNI estimates for clear-sky conditions (obtained with the REST2_v9.1 model) and cloud transmittance estimates using sunshine observations. The results indicate that the REST2_v9.1 model can be used to estimate DNI with high accuracy and consistency, owing to its robust two-band parameterization of the radiation transfer processes. Comparing daily DNI modeled predictions to measurements at 6 BSRN (Baseline Surface Radiation Network) stations in East Asia results in relatively low errors statistics: RMSE, MAE, RMSER, MAER and R of 1.436 MJm⁻², 0.900 MJm⁻², 22.26%, 13.95% and 0.972. Somewhat lessened agreement is found at the 17 CMA (China meteorological administrations) stations: 4.064 MJm⁻², 2.864 MJm⁻², 34.41%, 24.26% and 0.914, respectively. A gridded DNI dataset is constructed using sunshine duration measurements at 2474 CMA meteorological stations and the MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, version 2) reanalysis products. The spatial and temporal variations of DNI in different climate zones throughout China are also investigated. The gridded DNI datasets generated in this study would assist in numerous solar resource studies and solar energy applications.