Hypertranscription is widespread in aggressive human cancers. However detection relies on mRNAs, which are heavily processed and have variable half-lives, and on accurate cell number estimations. Previously we introduced FFPE-CUTAC, a genome-wide method for mapping RNA Polymerase II in formalin-fixed paraffin-embedded (FFPE) sections. Here we apply FFPE-CUTAC on slides and curls to demonstrate hypertranscription at regulatory elements and replication-coupled histone genes. We find that hypertranscription differs between transgene-driven mouse gliomas and scales with enhanced proliferation and reduced mitochondrial DNA. We also apply FFPE-CUTAC to identify tumor-specific patterns in assorted human tumor-normal pairs. We analyze the top-ranked 100 annotated regulatory elements that are hypertranscribed in most of the tumors and identify multiple loci around ERBB2 on Chromosome 17q12-21 in the breast and colon cancer samples, mapping likely HER2 amplifications punctuated by selective sweeps. Our results demonstrate that FFPE-CUTAC measurement of hypertranscription provides an affordable and sensitive genome-wide strategy for cancer diagnosis.