Nitrous oxide (N2O) emissions from inland waters remain a major source of uncertainty in global greenhouse gas budgets. N2O emissions are typically estimated using emission factors (EFs), defined as the proportion of the terrestrial nitrogen (N) load to a water body that is emitted as N2O to the atmosphere. The Intergovernmental Panel on Climate Change (IPCC) has proposed EFs of 0.25% and 0.75%, though studies have suggested that both these values are either too high or too low. In this work, we develop a mechanistic modeling approach to explicitly predict N2O production and emissions via nitrification and denitrification in rivers, reservoirs, and estuaries. In particular, we introduce a water residence time dependence, which kinetically limits the extent of denitrification and nitrification in water bodies. We revise existing spatially‐explicit estimates of N loads to inland waters to predict both lumped watershed and half‐degree grid cell emissions and EFs worldwide, as well as the proportions of these emissions that originate from denitrification and nitrification. We estimate global inland water N2O emissions of 10.6‐19.8 Gmol N yr−1 (148‐277 Gg N yr−1), with reservoirs producing most N2O per unit area. Our results indicate that IPCC EFs are likely overestimated by up to an order of magnitude, and that achieving the magnitude of the IPCC's EFs is kinetically improbable in most river systems. Denitrification represents the major pathway of N2O production in river systems, whereas nitrification dominates production in reservoirs and estuaries