The cancer stem cell hypothesis is an evolving concept of oncogenesis that has recently gained wide acceptance. In its simplest form this hypothesis suggests that many if not all tumors arise by consecutive genetic changes in a small subpopulation of cells termed cancer stem cells. These cells have the capacity to sustain tumor growth and are defined by three features: self-renewal, differentiation into the cell types of the original cancer and potent tumor formation. The definition of a cancer stem cell does not necessarily imply its origin from a stem, progenitor or differentiated cell. Hence, the term tumor or cancer initiating cell is often used instead to avoid any implications. Here, we propose a model suggesting that tumor cells progressively acquire stem cell properties as a consequence of oncogene-induced plasticity. The basis of our proposal are data from several experimental in vitro and in vivo models demonstrating reprogramming events triggered by specific combinations of oncogenes. These oncogene combinations not only induce cell lineage switches but also drive the reversal of ontogeny within cell lineages during tumor progression to metastasis. In this perspective article we will summarize the experimental evidence that illustrates our concept and discuss its implications for tumor formation and tumor therapy.