Our recent perplexing findings that polyploid giant cancer cells (PGCCs) acquired embryonic-like stemness and were capable of tumor initiation raised two important unanswered questions: how do PGCCs acquire such stemness, and to which stage of normal development do PGCCs correspond. Intriguingly, formation of giant cells due to failed mitosis/cytokinesis is common in the blastomere stage of the preimplantation embryo. However, the relationship between PGCCs and giant blastomeres has never been studied. Here, we tracked the fate of single PGCCs following paclitaxel-induced mitotic failure. Morphologically, early spheroids derived from PGCCs were indistinguishable from human embryos at the blastomere, polyploid blastomere, compaction, morula and blastocyst-like stages by light, scanning electron or three-dimensional confocal scanning microscopy. Formation of PGCCs was associated with activation of senescence, while budding of daughter cells was associated with senescence escape. PGCCs showed time- and space-dependent activation of expression of the embryonic stem cell markers OCT4, NANOG, SOX2 and SSEA1 and lacked expression of Xist. PGCCs acquired mesenchymal phenotype and were capable of differentiation into all three germ layers in vitro. The embryonic-like stemness of PGCCs was associated with nuclear accumulation of YAP, a key mediator of the Hippo pathway. Spheroids derived from single PGCCs grew into a wide spectrum of human neoplasms, including germ cell tumors, high-grade and low-grade carcinomas and benign tissues. Daughter cells derived from PGCCs showed attenuated capacity for invasion and increased resistance to paclitaxel. We also observed formation of PGCCs and dedifferentiation in ovarian cancer specimens from patients treated with chemotherapy. Taken together, our findings demonstrate that PGCCs represent somatic equivalents of blastomeres, the most primitive cancer stem cells reported to date. Thus, our studies reveal an evolutionarily conserved archaic embryonic program in somatic cells that can be de-repressed for oncogenesis. Our work offers a new paradigm for cancer origin and disease relapse.