The history of human acute myeloid leukemia stem cells (AMLSCs) began in a seminal study performed by Lapidot and Dick, proving that only CD34+CD38- human primary acute myeloid leukemia (AML) cells can repopulate in severe combined immunodeficient mice. The concept of leukemic stem cells (LSCs) has impeded a huge change in the treatment strategy against AML from killing proliferating leukemic cells to eradicating quiescent/dormant LSCs. As next-generation sequencing technologies have developed, multiple and recurrent genetic mutations have been discovered in large cohorts of patients with AML, and the updated understanding of leukemogenesis has improved the old concept of LSC to a revised version of a serial developmental model of LSC; that is, pre-LSCs are generated as seeds by the first hit on epigenetic regulators, and then, leukemia-initiating LSCs emerge from seeds by the second hits on genes involved in transcription and signaling. Dreams for universal and targetable AMLSC biomarker sparing healthy hematopoietic stem cells have weakened after the confrontation of significant heterogeneity of AMLSCs from genomic and immunophenotypic viewpoints. However, there is still hope for effective targets for AMLSCs since there is evidence that grouped gene signatures, such as 17-gene LSC score, and common epigenetic signatures, such as clusters, independent of various gene mutations, exist. Recently, the LSC niche in the bone marrow has been actively investigated and has expanded our knowledge of the physiology and vulnerability of AMLSCs. Presently, an applicable treatment that always works in AMLSCs is lacking. However, we will find a way, we always have.