Land surface all-wave net radiation (R_n) is a key parameter in many land processes. Current products have drawbacks of coarse resolutions, large uncertainty, and short time spans. A deep learning method was used to obtain global surface R_n. A long-term R_n product was generated from 1981 to 2019 using AVHRR data. The product has the highest accuracy and a reasonable spatiotemporal variation compared to three other products. Our product will play an important role in long-term climate change.