Human endogenous retroviruses (HERVs) have been implicated in a variety of human diseases including cancers. However, technical challenges in analyzing HERV sequence data have limited locus-specific characterization of HERV expression. Here, we use the software Telescope (developed to identify expressed transposable elements from metatranscriptomic data) on 43 paired tumor and adjacent normal tissue samples from The Cancer Genome Atlas Program to produce the first locus-specific retrotranscriptome of head and neck cancer. Telescope identified over 3000 expressed HERVs in tumor and adjacent normal tissue, and 1078 HERVs were differentially expressed between the two tissue types. The majority of differentially expressed HERVs were expressed at a higher level in tumor tissue. Differentially expressed HERVs were enriched in members of the HERVH family. Hierarchical clustering based on HERV expression in tumor-adjacent normal tissue resulted in two distinct clusters with significantly different survival probability. Together, these results highlight the importance of future work on the role of HERVs across a range of cancers.