The estimation of critical heat flux (CHF) in nuclear reactors is based largely on empirical relations that have aphysteal limiting conditions, a narrow range of applicability, and are inadequate for transient conditions. It is generally agreed that a more physically based approach is needed. Evidence is presented supporting the importance of boiling-induced fluid flow on the CHF process. Computational fluid dynamics (CFD) is used to model the microscale, transient dynamics of a vapor bubble growing in a subcooled liquid, resulting in qualitative reproduction of vapor blanket growth and CHF. The same CFD techniques are used to evaluate the macroscale thermal diffusion caused by spacers, resulting in qualitative reproduction of previous empirical results. This work forms the basis for a systematic investigation of CHF that could result in improved and less costly procedures for nuclear fuel design.