We linked a leaf-level CO2 assimilation model with a model that accounts for light attenuation in the canopy and measurements of sap-flux-based canopy conductance into a new canopy conductance-constrained carbon assimilation (4C-A) model. We estimated canopy CO2 uptake (AnC) at the Duke Forest free-air CO2 enrichment (FACE) study. Rates of AnC estimated from the 4C-A model agreed well with leaf gas exchange measurements (Anet) in both CO2 treatments. Under ambient conditions, monthly sums of net CO2 uptake by the canopy (AnC) were 13% higher than estimates based on eddy-covariance and chamber measurements. Annual estimates of AnC were only 3% higher than carbon (C) accumulations and losses estimated from ground-based measurements for the entire stand. The C budget for the Pinus taeda component was well constrained (within 1% of ground-based measurements). Although the closure of the C budget for the broadleaf species was poorer (within 20%), these species are a minor component of the forest. Under elevated CO2, the C used annually for growth, turnover, and respiration balanced only 80% of the AnC. Of the extra 700 g C m−2 a−1 (1999 and 2000 average), 86% is attributable to surface soil CO2 efflux. This suggests that the production and turnover of fine roots was underestimated or that mycorrhizae and rhizodeposition became an increasingly important component of the C balance. Under elevated CO2, net ecosystem production increased by 272 g C m−2 a−1: 44% greater than under ambient CO2. The majority (87%) of this C was sequestered in a moderately long-term C pool in wood, with the remainder in the forest floor–soil subsystem.