The origin of late Neogene Hippopotamidae (Artiodactyla) involves one of the most serious conflicts between comparative anatomy and molecular biology: is Artiodactyla paraphyletic? Molecular comparisons indicate that Cetacea should be the modern sister group of hippos. This finding implies the existence of a fossil lineage linking cetaceans (first known in the early Eocene) to hippos (first known in the middle Miocene). The relationships of hippos within Artiodactyla are challenging, and the immediate affinities of Hippopotamidae have been studied by biologists for almost two centuries without resolution. Here, we compare opposing hypotheses implicating several "suiform" families. This morphological analysis of a comprehensive set of taxa and characters offers a robust solution to the origins of Hippopotamidae. This family appears to be deeply nested within the otherwise extinct artiodactyl family Anthracotheriidae, most precisely within the most advanced selenodont forms. The proposed sister group of hippos is the middle to late Miocene African semiaquatic Libycosaurus. Any close relationships of hippos with suoids, particularly with Tayassuidae, are rejected. Furthermore, the clade (Hippopotamidae, Anthracotheriidae) is proposed as the sister group of the Cetacea, offering broad morphological support for a molecular phylogeny, such support being also consistent with the fossil record. Corroboration of this relationship requires an exploration of anthracothere affinities with other Paleogene artiodactyls. Among those, the position of Ruminantia is a central question, still to be solved. Further progress in this debate is likely to come from morphological studies of paleontological data, whether known or still to be discovered.