Tumors depend on a blood supply to deliver oxygen and nutrients, making tumor vasculature an attractive anti-cancer target. However, only a fraction of cancer patients benefits from angiogenesis inhibitors. Whether anti-angiogenic therapy would be more effective if targeted to individuals with specific tumor characteristics is unknown. To better characterize the tumor vascular environment both within and between cancer types, we developed a standardized metric – the Endothelial Index (EI) – to estimate vascular density in over 10,000 human tumors, corresponding to 31 solid tumor types, from transcriptome data. We then used this index to compare hyper- and hypo-vascular tumors, enabling the classification of human tumors into six vascular microenvironment signatures (VMSs) based on the expression of a panel of 24 vascular hub genes. EI and VMS correlated with known tumor vascular features and were independently associated with prognosis in certain cancer types. Retrospective testing of clinical trial data identified VMS2 classification as a powerful biomarker for response to bevacizumab. Our studies thus provide an unbiased picture of human tumor vasculature which may enable more precise deployment of anti-angiogenesis therapy.