Ecosystems may exist in alternative stable states and thereby extremely differ in ecosystem structure and functions, including gross primary productivity (GPP), which is crucial for assessing an ecosystem's ability to capture atmospheric carbon dioxide, especially in the context of climate change. This study applied the alternative stable states theory to evaluate GPP in global dryland forests, and analyzed multi-year average GPP data alongside environmental factors such as the Aridity Index and mean annual precipitation. Here, we found the existence of alternative stable states of GPP along the aridity gradient. The mean values of GPP were 893 and 1,540 gC m-2 year-1 under lower and higher branches of alternative stable states, respectively, compared to the current mean value of 1,203 gC m-2 year-1. Notably, we observed striking regional disparities in GPP, with Africa and Oceania predominantly in the higher alternative stable state, while North America and Asia were in the lower alternative stable state. However, GPP, along with mean annual precipitation, did not exhibit alternative stable states, but a significant variation during the medium range of mean annual precipitation (241-402 mm year-1). The relationships between GPP data and environmental factors were consistent across different forest types. This study sheds light on dryland forest productivity and indicates adaptive management strategies that should be used to bolster ecosystem function in the context of climate change.