Cancer cells display heterogeneous and dynamic states in glioblastoma, but how these malignant states arise and whether they follow a tractable cellular trajectory across tumours is poorly understood. Here, we generate a deep single cell and spatial multi-region atlas of glioblastoma that integrates transcriptomic, epigenomic and genomic analysis to comprehensively characterise their tumour heterogeneity. We describe spatially-patterned transitions of malignant cells from dev-like towards glial injury response- and hypoxia-defined states during tumour expansion. This malignant cell trajectory dominates glioblastoma, manifesting across tumours and genetically distinct subclonal lineages that are finely spatially intermixed within tumours. Moreover, this trajectory unfolds across specialised myeloid signalling environments that mirror the spatial compartmentalization of malignant cells. Our findings define a stereotyped trajectory of cancer cells in glioblastoma and unify glioblastoma tumour heterogeneity into a tractable cellular and tissue framework.