Bivalent chromatin is defined by the co-occurrence of otherwise opposing H3K4me3 and H3K27me3 modifications and is typically located at unmethylated promoters of lowly transcribed genes. In embryonic stem cells, bivalent chromatin has been proposed to poise developmental genes for future activation, silencing or stable repression upon lineage commitment. Normally, bivalent chromatin is kept in tight balance in cells, in part through the activity of the MLL/COMPASS-like and Polycomb repressive complexes that deposit the H3K4me3 and H3K27me3 modifications, respectively, but also emerging novel regulators including DPPA2/4, QSER1, BEND3, TET1 and METTL14. In cancers, both the deregulation of existing domains and the creation of de novo bivalent states is associated with either the activation or silencing of transcriptional programmes. This may facilitate diverse aspects of cancer pathology including epithelial-to-mesenchymal plasticity, chemoresistance and immune evasion. Here, we review current methods for detecting bivalent chromatin and discuss the factors involved in the formation and fine-tuning of bivalent domains. Finally, we examine how the deregulation of chromatin bivalency in the context of cancer could facilitate and/or reflect cancer cell adaptation. We propose a model in which bivalent chromatin represents a dynamic balance between otherwise opposing states, where the underlying DNA sequence is primed for the future activation or repression. Shifting this balance in any direction disrupts the tight equilibrium and tips cells into an altered epigenetic and phenotypic space, facilitating both developmental and cancer processes.