Carbohydrate-active enzymes (CAZymes) are families of essential and structurally related enzymes, which catalyze the creation, modification, and degradation of glycosidic bonds in carbohydrates to maintain essentially all kingdoms of life. CAZymes play a key role in many biological processes underpinning human health and diseases (e.g., cancer, diabetes, Alzheimer's diseases, AIDS) and have thus emerged as important drug targets in the fight against pathogenesis. The realization of the full potential of CAZymes remains a significant challenge, relying on a deeper understanding of the molecular mechanisms of catalysis. Considering numerous unsettled questions in the literature, while with a large amount of structural, kinetic, and mutagenesis data available for CAZymes, there is a pressing need and an abundant opportunity for collaborative computational and experimental investigations with the aim to unlock the secrets of CAZyme catalysis at an atomic level. In this review, we briefly survey key methodology development in computational studies of CAZyme catalysis. This is complemented by selected case studies highlighting mechanistic insights provided by computational glycobiology. Implication for inhibitor design by mimicking the transition state is also illustrated for both glycoside hydrolases and glycosyltransferases. The challenges for such studies will be noted and finally an outlook for future directions will be provided.