Genome size is tightly coupled to morphology, ecology, and evolution among species [1-5], with one of the best-known patterns being the relationship between cell size and genome size [6, 7]. Classic theories, such as the "selfish DNA hypothesis," posit that accumulating redundant DNA has fitness costs but that larger cells can tolerate larger genomes, leading to a positive relationship between cell size and genome size [8, 9]. Yet the evidence for fitness costs associated with relatively larger genomes remains circumstantial. Here, we estimated the relationships between genome size, cell size, energy fluxes, and fitness across 72 independent lineages in a eukaryotic phytoplankton. Lineages with relatively smaller genomes had higher fitness, in terms of both maximum growth rate and total biovolume reached at carrying capacity, but paradoxically, they also had lower energy fluxes than lineages with relative larger genomes. We then explored the evolutionary trajectories of absolute genome size over 100 generations and across a 10-fold change in cell size. Despite consistent directional selection across all lineages, genome size decreased by 11% in lineages with absolutely larger genomes but showed little evolution in lineages with absolutely smaller genomes, implying a lower absolute limit in genome size. Our results suggest that the positive relationship between cell size and genome size in nature may be the product of conflicting evolutionary pressures, on the one hand, to minimize redundant DNA and maximize performance-as theory predicts-but also to maintain a minimum level of essential function. VIDEO ABSTRACT.