Nitrogen (N) plays a dual role in the agri-food system: it is an essential nutrient for all life forms, yet also an environmental pollutant causing a range of environmental and human health impacts. As the plant nutrient needed in greatest quantities, and as a building block of proteins and other biomolecules, N is a necessary part of all life. In the last century, an enormous increase of N turnover in the agri-food system has enabled increasing per-capita food supply for a growing world population, but as an unintended side effect, N pollution has increased to levels widely agreed in science and policy to be far beyond sustainable limits. There is no such thing as perfectly circular N supply. Losses of N to the environment inevitably arise as N is transformed and used in the food system, for example in soil processes, in manure storage, and in fertilizer application. This lost N must be replaced by ‘new’ N, which is N converted to bioavailable forms from the vast atmospheric pool of unreactive dinitrogen (N2). New N comes mainly as synthetic N fertilizer and through a process known as biological N fixation (BNF). In addition, there is a large internal flow of recycled N in the food system, mainly in the form of livestock excreta. This recirculated N, however, is internal to the food system and cannot make up for the inevitable losses of N. The introduction of synthetic N fertilizer during the 20th century revolutionized the entire food system. The industrial production of synthetic N fertilizer was a revolution for agricultural systems because it removed the natural constraint of N scarcity. Given sufficient energy, synthetic N fertilizer can be produced in limitless quantities from atmospheric dinitrogen (N2). This has far-reaching consequences for the whole agri-food system. The annual input of synthetic N fertilizer today is more than twice the annual input of new N in pre-industrial agriculture. Since 1961, increased N input has enabled global output of both crop and livestock products to roughly triple. During the same time period, total food-system N emissions to the environment have also more than tripled. Livestock production is responsible for a large majority of agricultural N emissions. Livestock consume about three-quarters of global cropland N output and are thereby responsible for a similar share of cropland N emissions to air and water. In addition, N emissions from livestock housing and manure management systems contribute a substantial share of global N emissions to air. There is broad political agreement that global N emissions from agriculture should be reduced by about 50%. High-level policy targets of the EU and of the UN Convention on Biological Diversity are for a 50% reduction in N emissions. These targets are in line with a large body of research assessing what would be needed to stay within acceptable limits as regards ecosystem change and human health impacts. In the absence of dietary change towards less N-intensive diets, N emissions from food systems could be reduced by about 30%, compared to business-as-usual scenarios. This could be achieved by implementing a combination of technical measures, improved management practices, improved recycling of wasted N (including N from human excreta), and spatial optimization of agriculture. Human dietary change, especially in the most affluent countries, offers a huge potential for reducing N emissions from food systems. While many of the world’s poor would benefit nutritionally from increasing their consumption of nutrient-rich animal-source foods, many other people consume far more nutrients than is necessary and could reduce consumption of animal-source food by half without any nutritional issues. Research shows that global adoption of healthy but less N-polluting diets might plausibly cut future food-system N losses by 10–40% compared to business-as-usual scenarios. There is no single solution for solving the N challenge. Research shows that efficiency improvements and food waste reductions will almost certainly be insufficient to reach agreed environmental targets. To reach agreed targets, it seems necessary to also shift global average food consumption onto a trajectory with less animal-source food.