After four growing seasons, elevated CO2 did not significantly alter surface soil C pools in two intact annual grasslands. However, soil C pools in these systems are large compared to the likely changes caused by elevated CO2. We calculated statistical power to detect changes in soil C, using an approach applicable to all elevated CO2 experiments. The distinctive isotopic signature of the fossil-fuel-derived CO2 added to the elevated CO2 treatment provides a C tracer to determine the rate of incorporation of newly-fixed C into soil. This information constrains the size of the possible effect of elevated CO2 on soil C. Even after four years of treatment, statistical power to detect plausible changes in soil C under elevated CO2 is quite low. Analysis of other elevated CO2 experiments in the literature indicates that either CO2 does not affect soil C content, or that reported CO2 effects on soil C are too large to be a simple consequence of increased plant carbon inputs, suggesting that other mechanisms are involved, or that the differences are due to chance. Determining the effects of elevated CO2 on total soil C and long-term C storage requires more powerful experimental techniques or experiments of longer duration.