We present a computational model of mitochondrial deoxynucleotide metabolism and mitochondrial DNA (mtDNA) synthesis. The model includes the transport of deoxynucleosides and deoxynucleotides into the mitochondrial matrix space, their phosphorylation and polymerization into mtDNA. Different simulated cell types (cancer, rapidly dividing, slowly dividing, and postmitotic) were represented in this model by different cytoplasmic deoxynucleotide concentrations. We calculated the changes in deoxynucleotide concentrations within the mitochondrion over the course of an mtDNA replication event and the time required for mtDNA replication in the different cell types. Based on the model we define three steady states of mitochondrial deoxynucleotide metabolism, the phosphorylating state (net import of deoxynucleosides and export of phosphorylated deoxynucleotides), the desphosphorylating state (the reverse), and the efficient state (net import of both deoxynucleosides and deoxynucleotides). We present five testable hypotheses based on this simulation. (1) The deoxynucleotide pools within a mitochondrion are sufficient to support only a small fraction of even a single mtDNA replication event. (2) The mtDNA replication time in postmitotic cells is much longer than that in rapidly dividing cells. (3) Mitochondria in dividing cells are net sinks of cytoplasmic deoxynucleotides, while mitochondria in postmitotic cells are net sources. (4) The deoxynucleotide carrier exerts the most control over the mtDNA replication rate in rapidly dividing cells, but in postmitotic cells the NDPK and TK2 enzymes have the most control. (5) Following from the previous point, rapidly dividing cells derive almost all of their mtDNA precursors from the cytoplasmic deoxynucleotides, not from phosphorylation within the mitochondrion.