A key objective for sustainable agriculture and forestry is to breed plants with both high carbon gain and water-use efficiency (WUE). At the level of leaf physiology, this implies increasing net photosynthesis (A N) relative to stomatal conductance (g s). Here, we review evidence for CO2 diffusional constraints on photosynthesis and WUE. Analyzing past observations for an extensive pool of crop and wild plant species that vary widely in mesophyll conductance to CO2 (g m), g s, and foliage A N, it was shown that both g s and g m limit A N, although the relative importance of each of the two conductances depends on species and conditions. Based on Fick's law of diffusion, intrinsic WUE (the ratio A N/g s) should correlate on the ratio g m/g s, and not g m itself. Such a correlation is indeed often observed in the data. However, since besides diffusion A N also depends on photosynthetic capacity (i.e., V c,max), this relationship is not always sustained. It was shown that only in a very few cases, genotype selection has resulted in simultaneous increases of both A N and WUE. In fact, such a response has never been observed in genetically modified plants specifically engineered for either reduced g s or enhanced g m. Although increasing g m alone would result in increasing photosynthesis, and potentially increasing WUE, in practice, higher WUE seems to be only achieved when there are no parallel changes in g s. We conclude that for simultaneous improvement of A N and WUE, genetic manipulation of g m should avoid parallel changes in g s, and we suggest that the appropriate trait for selection for enhanced WUE is increased g m/g s.