Agricultural application of reactive nitrogen (N) for fertilization is a cause of massive negative environmental problems on a global scale. However, spatially explicit and crop-specific information on global N losses into the environment and knowledge of trade-offs between N losses and crop yields are largely lacking. We use a crop growth model, Python-based Environmental Policy Integrated Climate (PEPIC), to determine global N losses from three major food crops: maize, rice, and wheat. Simulated total N losses into the environment (including water and atmosphere) are 44 Tg N yr− 1. Two thirds of these, or 29 Tg N yr− 1, are losses to water alone. Rice accounts for the highest N losses, followed by wheat and maize. The N loss intensity (NLI), defined as N losses per unit of yield, is used to address trade-offs between N losses and crop yields. The NLI presents high variation among different countries, indicating diverse N losses to produce the same amount of yields. Simulations of mitigation scenarios indicate that redistributing global N inputs and improving N management could significantly abate N losses and at the same time even increase yields without any additional total N inputs