Adult and pediatric glioblastomas are incurable and extremely aggressive brain cancers. Their extensive intratumoral genomic heterogeneity and their hierarchical functional organization contribute to the biological complexity of these malignancies. Single-cell omics could assist in resolving the cellular complexity of glioblastoma. However, computational bottlenecks, especially for downstream analyses of single-cell epigenomic datasets, have been obstacles in direct investigations of surgical specimens. We have created a computational tool that uses single-cell epigenomic data to differentiate between tumor and non-malignant cells in surgical samples. Analyses of pure tumor cell populations enabled by our computational tool revealed complex epigenetic behavior in individual glioblastomas, both statically and during disease progression. Our data uncover epigenomic networks that contribute to tumorigenic function, including transcription factors that dominate the chromatin landscapes of cancer stem cell populations. All together, our computational tool and single-cell epigenomic datasets could significantly improve our understanding of the complex biology of pediatric and adult glioblastoma.