Contrastive learning has been established as a powerful representation learning approach for unlabeled data. Models trained with contrastive learning approaches have shown to be robust and transferable across different tasks, and are currently used worldwide for applications like image generation with text prompting. In drug discovery, large amounts of unlabeled data are frequently encountered, but currently are often not used in standard machine learning methods. Therefore, contrastive learning is a suitable approach to empower drug discovery by exploiting the large amounts of unlabeled data. In this chapter, we introduce the contrastive learning framework and its most prominent applications for drug discovery. Further, we explain for which computer-aided drug discovery tasks contrastive learning methods can be used.