Plasmids are extra-chromosomal genetic elements that encode a wide variety of phenotypes and can be maintained in bacterial populations through vertical and horizontal transmission, thus increasing bacterial adaptation to hostile environmental conditions like those imposed by antimicrobial substances. To circumvent the segregational instability resulting from randomly distributing plasmids between daughter cells upon division, non-transmissible plasmids tend to be carried in multiple copies per cell, with the added benefit of exhibiting increased gene dosage. But carrying multiple copies also results in a high metabolic burden to the bacterial host, therefore reducing the overall fitness of the population. This trade-off poses an existential question for plasmids: What is the optimal plasmid copy number? In this manuscript, we address this question using a combination of population genetics modeling with microbiology experiments consisting of Escherichia coli K12 bearing a multi-copy plasmid encoding for bla TEM-1 , a gene conferring resistance to β -lactam antibiotics. We postulate and analyze a Wright-Fisher model to evaluate the interaction between selective pressure, the number of plasmid copies carried by each cell, and the energetic cost associated with plasmid bearing. By numerically determining the optimal plasmid copy number for constant and fluctuating selection regimes, we show that the stability of multi-copy plasmids is maximized at intermediate plasmid copy numbers. We conclude by arguing that plasmid copy number is a highly optimized evolutionary trait that depends on the rate of environmental fluctuation and balances the benefit between increased stability in the absence of selection with the burden associated with carrying multiple copies of the plasmid.