The common empirical idea that the dissolution of silica is controlled by the hydrolysis of the first Si-O-Si bond of Si surface species is checked against experimental activation energies by DFT (B3LYP) calculations using a polarized continuum model. The calculated energy barrier for the hydrolysis of a Si-O-Si bridge of a double-linked Si atom of beta -cristobalite at the water-silica interface (29 kcal/mol) comes out to be higher than the measured activation energy of silica dissolution at the point of zero net proton charge of the surface (PZPC) by at least 7 kcal/mol. This discrepancy is significantly outside the estimated error bars of the calculations. Therefore, we propose a new mechanism, which is based on the assumption that the breakage of the first Si-O-Si bond is followed by the very fast reverse reaction of dehydroxylation of the formed Si-OH HO-Si defect. Because of this "self-healing" effect, the probability of both Si-O-Si bonds of the double-linked Si atoms being dissociated is very low, which explains the very small rate of dissolution of silica at PZPC. This mechanism also allows us to interpret the experimental fact that the preexponential factor of the reaction is extremely small. Within the new mechanism, the measured activation energy is associated with the hydrolysis of the last Si-O-Si bond of the Si atoms. Unlike the first Si-O-Si bond, the hydrolysis. of the last bond is not hindered by the lattice resistance, which leads the theoretical activation energy (20 kcal/mol) to be in good agreement with experiment.