Over half of the human genome is comprised of transposable elements (TE). Despite large-scale studies of the transcriptome in cancer, a comprehensive look at TE expression and its relationship to various mutations or prognosis has not been performed. We characterized the expression of TE in 178 adult acute myeloid leukemia (AML) patients using transcriptome data from The Cancer Genome Atlas. We characterized mutation specific dysregulation of TE expression using a multivariate linear model. We identified distinct patterns of TE expression associated with specific mutations and transcriptional networks. Genes regulating methylation was not associated with significant change in TE expression. Using an unpenalized cox regression analysis we identified a TE expression signature that predicted prognosis in AML. We identified 14 candidate prognostic TE transcripts (TEP) that classified AML as high/low-risk and this was independent of mutation-based and coding-gene expression based risk-stratification. TEP was able to predict prognosis in independent cohorts of 284 pediatric AML patients and 19 relapsed adult AML patients. This first comprehensive study of TE expression in AML demonstrates that TE expression can serve as a biomarker for prognosis in AML, and provides novel insights into the biology of AML. Studies characterizing its role in other cancers are warranted.