Intensification of the Earth's hydrological cycle amplifies the interannual variability of precipitation, which will significantly impact the terrestrial carbon (C) cycle. However, it is still unknown whether previously observed relationship between soil respiration (Rs) and precipitation remains applicable under extreme precipitation change. By analyzing the observations from a much larger dataset of field experiments (248 published papers including 151 grassland studies and 97 forest studies) across a wider range of precipitation manipulation than previous studies, we found that the relationship of Rs response with precipitation change was highly nonlinear or asymmetric, and differed significantly between grasslands and forests, between moderate and extreme precipitation changes. Response of Rs to precipitation change was negatively asymmetric (concave‐down) in grasslands, and double‐asymmetric in forests with a positive asymmetry (concave‐up) under moderate precipitation changes and a negative asymmetry (concave‐down) under extreme precipitation changes. In grasslands, the negative asymmetry in Rs response was attributed to the higher sensitivities of soil moisture, microbial and root activities to decreased precipitation (DPPT) than to increased precipitation (IPPT). In forests, the positive asymmetry was predominantly driven by the significant increase in microbial respiration under moderate IPPT, while the negative asymmetry was caused by the reductions in root biomass and respiration under extreme DPPT. The different asymmetric responses of Rs between grasslands and forests will greatly improve our ability to forecast the C cycle consequences of increased precipitation variability. Specifically, the negative asymmetry of Rs response under extreme precipitation change suggests that the soil C efflux will decrease across grasslands and forests under future precipitation regime with more wet and dry extremes. The relationship of soil respiration response with precipitation change was highly nonlinear or asymmetric, and differed significantly between grasslands and forests. Soil respiration response was negatively asymmetric (concave‐down) in grasslands, and double‐asymmetric in forests with a positive asymmetry (concave‐up) under moderate precipitation changes and a negative asymmetry (concave‐down) under extreme precipitation changes.