Intercropping is a promising sustainable agriculture practice, but its adoption is hindered by the potential reduction in benefits when shifting from monoculture to intercropping, especially if there are trade-offs between yield, soil organic carbon (SOC) and nitrous oxide (N₂O) emissions. In this study, we synthesized 3843 paired observations from 510 field-based studies that compared the yields, SOC and N₂O emissions of intercropping systems (only crop-crop intercropping for grain yield and including crop-tree agroforestry in addition to crop-crop intercropping for SOC and N₂O) with monoculture systems. Our meta-analysis revealed that on average, intercropping increased grain yields by 22.3 %, calorie yields by 16.1 % and protein yields by 14.4 %, while reducing N₂O emissions by 14.3 %. Legume-based intercropping was found to be more effective than non-legume intercropping. Intercropping also significantly increased SOC content, with agroforestry outperforming crop-crop intercropping (average increases of 15.5 % and 6.1 %, respectively). However, the effectiveness of intercropping systems varies depending on climate, soil properties and management practices. Based on data-driven machine learning models, our upscaling analyses show that converting monoculture land to cereal-legume intercropping could increase grain yields by 5.35 × 10⁸ Mg yr ^-1, potentially feeding nearly 1013 million people. This conversion could also enhance SOC stocks by 2.84 Pg C yr ^-1 and reduce N₂O emissions by 625 Gg N yr ^-1, resulting in a reduction of 10.6 Pg CO₂-eq yr ^-1 for net global warming potential. Overall, converting monoculture systems to intercropping could generate approximately US$595 billion in economic benefits annually. These findings highlight the critical role of intercropping as a scalable nature-based solution for addressing the dual challenges of global food security and climate change.