Crystallographic, mutagenesis, kinetic, and computational studies on Rubisco over three decades have revealed much about its catalytic mechanism and the role played by several active-site residues. However, key questions remain unanswered. Specific details of the carboxylase and oxygenase mechanisms, required to underpin the rational re-engineering of Rubisco, are still speculative. Here we address critical gaps in knowledge with a definitive comprehensive computational investigation of the mechanism of carboxylase activity at the Rubisco active site. Density functional theory calculations (B3LYP/6-31G(d,p)) were performed on active-site fragment models of a size up to 77 atoms, not previously possible computationally. All amino acid residues suspected to play roles in the acid-base chemistry in the multistep reaction, and interacting directly with the central Mg (2+) atom and the reactive moiety of substrate and intermediates, were included. The results provide a firm basis for us to propose a novel mechanism for the entire sequence of reactions in the carboxylase catalysis and to define precise roles for the active-site residues, singly and in concert. In this mechanism, the carbamylated LYS201 plays a more limited role than previously proposed but is crucial for initiating the reaction by acting as a base in the enolization. We suggest a wider role for HIS294, with involvement in the carboxylation, hydration, and C2-C3 bond-scission steps, consistent with the suggestion of Harpel et al. (1998) but contrary to the consensus view of Cleland et al. (1998). In contrast to the common assumption that the water molecule for the hydration step comes from within the active site, we propose that the Mg-coordinated water is not dissociated at the start of the gas-addition reaction but rather remains coordinated and is used for the hydration of the C3 carbon atom. New roles are also proposed for LYS175, GLU204, and HIS294. The mechanism suggests roles in the gas-addition step for residues in three spatially distinct regions of the active site, HIS294 and LYS334 in the C-terminal domain of the large subunit (LSU), but also hitherto unsuspected roles for a cluster of three residues (ASN123, GLU60, and TYR20) in the N-terminal domain of the partner LSU of the dimer containing the active site. Our new mechanism is supported by existing experimental data, provides new convincing interpretations of previously puzzling data, and allows new insights into mutational strategies for improving Rubisco activity.