Based on a retrospective study of the French Fast-breeder Reactor (FBR) programme, this paper aims to show the dynamics of FBR demonstrator evaluation, with methodological inputs from the “Science and Technology Studies” branch of sociology. Such a reactor has to demonstrate the feasibility - including safety, technical and economic viability - of a promising technology regarded as a potentially inexhaustible energy source for the future. Until the mid-seventies, the need for an FBR fleet was regarded as urgent, entailing a focus on demonstration reactors to prove “technical” feasibility. But after the mid-seventies, the purpose of evaluating FBR projects was to prove the technical and economic viability of the programme, as well as its safety. The analysis of the Superphénix case is used to illustrate the difficulty of reconciling the three elements of assessment in a changing context, where the respective weights of the various criteria evolve in a dynamic fashion: it calls for an examination of the implicit specifications of demonstrators.