Multivesicular bodies (MVBs) are endosomes that have internalized portions of the limiting membrane into the compartment, thereby forming intralumenal vesicles. This vesicle formation is unusual in that it is directed away from the cytoplasm, which requires a unique mechanism unlike any mechanism described for other vesicle formation events. The best contenders for the machinery that drives MVB vesicle formation are the ESCRT protein complexes. However, increasing evidence suggests that lipids may play a key role in this membrane-deformation process. This review attempts to combine the seemingly contradictory findings into a MVB vesicle formation model that is based on a lipid-driven and ESCRT-regulated mechanism.