Cytoplasmic male sterility (CMS) in plants is a classical example of genomic conflict, opposing maternally-inherited cytoplasmic genes (mitochondrial genes in most cases), which induce male sterility, and nuclear genes, which restore male fertility. In natural populations, this type of sex control leads to gynodioecy, that is, the co-occurrence of female and hermaphroditic individuals within a population. According to theoretical models, two conditions may maintain male sterility in a natural population: (1) female advantage (female plants are reproductively more successful than hermaphrodites on account of their global seed production); (2) the counter-selection of nuclear fertility restorers when the corresponding male-sterility-inducing cytoplasm is lacking. In this review, we re-examine the model of nuclear-mitochondrial conflict in the light of recent experimental results from naturally occurring CMS, alloplasmic CMS (appearing after interspecific crosses resulting from the association of nuclear and cytoplasmic genomes from different species), and CMS plants obtained in the laboratory and carrying mitochondrial mutations. We raise new hypotheses and discuss experimental models that would take physiological interactions between cytoplasmic and nuclear genomes into account.