The intensification of global change has led to frequent atmospheric and soil drought events, posing severe threats to global ecosystems. Although soil drought (characterized by soil moisture, SM) and atmospheric drought (characterized by vapor pressure deficit, VPD) often co-occur, their combined effects are rarely quantified as compound droughts. This study integrates observational data and Coupled Model Intercomparison Project Phase 6 (CMIP6) simulations with correlation analysis, copula models, and machine learning to investigate the occurrence and impacts of compound extreme droughts. Our findings reveal that SM and VPD exhibit bimodal distributions, with synchronized extreme soil droughts and extreme atmospheric droughts occurring more frequently than expected from individual extreme events. Compared to the historical simulations (1920-1999) (-0.22 gC center dot m- 2 center dot day- 1), the impact of compound extreme droughts on gross primary productivity (GPP) are projected to be more severe in the future simulations (2021-2100) (SSP126:-0.26 gC center dot m- 2 center dot day- 1; SSP370:-0.33 gC center dot m- 2 center dot day- 1), with particularly pronounced impacts in semi-arid regions. The compound drought stress on GPP exhibits significant variations across vegetation types and along the climatic aridity gradient. With increasing carbon emission scenarios, CO2 becomes a crucial regulatory factor in compound drought stress on vegetation productivity. The negative impact intensity and spatial extent of extreme soil drought on GPP far exceed those of extreme atmospheric drought, indicating that SM will play a more critical role in extreme drought stress on vegetation productivity. These findings highlight evidence that extreme drought events weaken vegetation carbon sequestration, providing essential insights for accurately assessing the interactions between vegetation and climate in China under climate change scenarios.