Epichloë bromicola is an endophytic fungal species that systemically and perennially colonizes intercellular spaces of leaf blades, leaf sheaths and culms of Bromus grass species. E. bromicola causes choke disease in B. erectus, suppressing maturation of most, if not all, host inflorescences. In an investigation of the interaction between fungus and host, we used a quantitative polymerase chain reaction technique to estimate the amount of fungal DNA, and thereby fungal concentration, in host plants. Fungal concentration was directly correlated with vegetative vigour of the plant, as measured by longest leaf length, number of tillers and vegetative above-ground biomass, suggesting that, during vegetative growth, the endophytic fungus is most beneficial for the plant when present in high concentrations. In contrast, the reproduction of the plant, as measured by the number of functional inflorescences, was inversely correlated with fungal concentration: the majority of infected plants, and all that were associated with high concentrations of fungi, were diseased. Thus, the benefit of endophyte infection for the plant is coupled with the disadvantages of infertility. Fungal concentration was shown to be at least in part genetically determined because fungal concentration differed significantly in different plant–endophyte genotype combinations (symbiotum). In a field experiment with normal and CO2-enriched environments, elevated CO2 levels favoured fungal reproductive vigour over host reproductive vigour, suggesting that these plant endophytes would be at a selective advantage in a corresponding environmental-change scenario. We conclude that a dynamic and complex relationship between fungal endophyte infection, fungal concentration, genotype and environment affects growth and fecundity of B. erectus and should contribute to the evolution of these plant–fungal interactions.