Mitochondrial DNA (mtDNA) mutations and copy number changes are among the most common genetic events in all tumors and can directly impact genetic machinery critical to metabolic homeostasis. Despite the central role mitochondria play in energy metabolism and cell physiology, the degree to which mtDNA copy number and genotype vary from cell to cell in human tissues is poorly characterized. Moreover, the phenotypic and evolutionary consequences of this variation are poorly characterized. This dissertation addresses three main aims surrounding mitochondrial genome at the single cell level: (1) to characterize the factors that drive the large variation in per-cell mtDNA copy number and determine the transcriptional phenotype of the mtDNA copy number, (2) to determine how clonal structure of the tumor defined by nuDNA copy number variation and ploidy shape mtDNA copy number distribution, and (3) to investigate the phenotypic effects of mtDNA mutations as changes in mtDNA copy number and mtDNA gene expression. To achieve these aims, we combine the analysis of amplification-free single-cell whole genome sequencing (DLP+) on 121,099 single cells derived from immortalized cell lines, patient-derived xenografts, and primary human tumors as well as population-scale analysis of mtDNA in 14,439 tumors. Cells typically contained thousands of mtDNA copies, but variation in mtDNA copy number in otherwise isogenic cells was extensive and strongly associated with biophysical adaptations in cell size. Pervasive whole-genome doubling events in nuclear DNA (nuDNA) associated with stoichiometrically balanced adaptations in mtDNA copy number, implying that the mtDNA-to-nuDNA ratio (MNR), rather than mtDNA copy number itself, mediated downstream phenotypes. Through multimodal analysis of DLP+ and single cell RNA sequencing, we found that clones with high MNR exhibit increased expression of mtDNA-encoded genes and subsequent enrichment in innate immune response and depletion of hypoxia. Finally, we observed that both loss-of-function and non-coding variants to mtDNA elicited heteroplasmy-dependent changes in mtDNA copy number and mitochondrial transcription. Thus, co-registered measures of nuDNA and mtDNA in single cells illuminate adaptive states that disrupt the stoichiometric balance of nuclear and mitochondrial components of genomically unstable cancers.