The origin of a large spread in the activation energy of crystallization, &Dgr;E, in metallic glasses is explained based on a structural relaxation model. The calculated &Dgr;Efrom the available thermal and viscosity data is in fair agreement with the experimental results. It is the structural relaxation which leads to high &Dgr;E(⩾100 kcal/mole) for stable glasses. It is also suggested that the structural relaxation is the rate‐controlling factor during the crystallization of the unstable glasses. Thus &Dgr;Efor unstable glasses is low (&Dgr;E⩽20 kcal/mole), and is expected to be lower for amorphous metals prepared by low‐temperature deposition than for the corresponding melt‐quenched metallic glasses.