Transient starch production is thought to strongly control plant growth and response to elevated CO2. We tested this hypothesis with an experimentally based mechanistic model in Arabidopsis thaliana. Experiments were conducted on wild-type (WT) A. thaliana, starch-excess (sex1) and starchless (pgm) mutants under ambient and elevated CO2 conditions to determine parameters and validate the model. The model correctly predicted that mutant growth is approx. 20% of that in WT, and the absolute response of both mutants to elevated CO2 is an order of magnitude lower than in WT. For sex1, direct starch unavailability explained the growth responses. For pgm, we demonstrated experimentally that maintenance respiration is proportional to leaf soluble sugar concentration, which gave the necessary feedback mechanism on modelled growth. Our study suggests that the effects of sugar-starch cycling on growth can be explained by simple allocation processes, and the maximum rate of leaf growth (sink capacity) exerts a strong control over the response to elevated CO2 of herbaceous plants such as A. thaliana.