Hexavalent uranium (U(VI)) is reduced to tetravalent uranium (U(IV)) by microorganisms (e.g., Geobacter or Shewanella spp.) as well as by abiotic reductants such as sulfide or Fe(II) species. During reduction, the heavy isotope (238U) is typically enriched in U(IV), while abiotic reduction exhibits variable isotope fractionation directions. The use of U isotope signatures in rocks and sediments is an attractive tool for probing paleo-redox conditions and deconvoluting modern processes in the subsurface. However, these signatures are being used with little understanding of the mechanistic underpinnings of U isotopic fractionation. Here, we contribute a theoretical elucidation of U isotope fractionation during the biological reduction of U(VI) to U(IV) by introducing a steady-state model for the multi-step reduction reaction. This model was derived based on the requirement of the Rayleigh distillation model that the isotope fractionation coefficient εRayleigh is time-independent, and the final product is removed from the system. In this model, ε Rayleigh depends on the equilibrium isotope fractionation coefficient ε eq for each reaction step, and hence, we calculated ε eq using ab-initio methods. Our calculations revealed that ε eq is largest for redox steps (1.44–1.60‰ for U(VI) to U(V), 0.76–0.79‰ for U(V) to U(IV)) and for the binding of U(VI) to a cytochrome (0.42–0.73‰). Using experimentally-derived ε Rayleigh and the calculated ε eq, we determined that U isotopic fractionation associated with the binding of U(VI) to a cytochrome or with the reduction from U(VI) to U(V) could not be achieved solely with equilibrium fractionation and must include a kinetic isotope fractionation component. We also interpreted previously reported ε Rayleigh for abiotic U reduction by FeS, which followed the Rayleigh model. The abiotic ε Rayleigh depended on the rate of removal of U(VI) from the solution and the amounts of neutrally charged species (i.e., Ca2UO2(CO3)3). These experimental trends can be explained consistently using the steady-state model. Hence, we propose that the present steady-state model can be used more generally for any U multi-step reaction for which the experimental data follow the Rayleigh model.