Soil moisture profiles can affect species composition and ecosystem processes, but the effects of increased concentrations of atmospheric carbon dioxide ([CO2]) on the vertical distribution of plant water uptake have not been studied. Because plant species composition affects soil moisture profiles, and is likely to shift under elevated [CO2], it is also important to test whether the indirect effects of [CO2] on soil water content may depend on species composition. We examined the effects of elevated [CO2] and species composition on soil moisture profiles in an annual grassland of California. We grew monocultures and a mixture of Avena barbata and Hemizonia congesta– the dominant species of two phenological groups – in microcosms exposed to ambient (∼370 μmol mol−1) and elevated (∼700 μmol mol−1) [CO2]. Both species increased intrinsic and yield-based water use efficiency under elevated [CO2], but soil moisture increased only in communities with A. barbata, the dominant early-season annual grass. In A. barbata monocultures, the [CO2] treatment did not affect the depth distribution of soil water loss. In contrast to communities with A. barbata, monocultures of H. congesta, a late-season annual forb, did not conserve water under elevated [CO2], reflecting the increased growth of these plants. In late spring, elevated [CO2] also increased the efficiency of deep roots in H. congesta monocultures. Under ambient [CO2], roots below 60 cm accounted for 22% of total root biomass and were associated with 9% of total water loss, whereas in elevated [CO2], 16% of total belowground biomass was associated with 34% of total water loss. Both soil moisture and isotope data showed that H. congesta monocultures grown under elevated [CO2] began extracting water from deep soils 2 weeks earlier than plants in ambient [CO2].