Prognosis for esophageal squamous cell carcinoma (ESCC) is poor, so it is essential to develop a more complete understanding of the disease. The purpose of this study was to explore metabolic biomarkers and potential therapeutic targets for ESCC. An ultra-high-performance liquid chromatography coupled with high resolution mass (UPLC/MS)-based metabolomic analysis was performed in 141 ESCC cancerous tissue samples and 70 non-cancerous counterparts. The results showed that 41 differential metabolites were annotated in the training set, and 37 were validated in the test set. Single-metabolite-based receiver operating characteristic (ROC) curves as well as metabolite-based machine learning models, including Partial Least Squares (PLS), Support Vector Machine (SVM), and Random Forest (RF), were investigated for cancerous and non-cancerous tissue classification. Six most prevalent diagnostic metabolites-adenylsuccinic acid, UDP-GalNAc, maleylacetoacetic acid, hydroxyphenylacetylglycine, galactose, and kynurenine-showed testing predictive accuracies of 0.89, 0.95, 0.97, 0.89, 0.84, and 0.84, respectively. Moreover, the metabolite-based models (PLS, SVM, and RF) had testing predictive accuracies of 0.95, 0.95, and 1.00, respectively. Kaplan-Meier survival analysis and Cox proportional hazards regression analysis demonstrated that 2-hydroxymyristoylcarnitine (HR: 0.55, 95 % CI: 0.32 to 0.92), 3-hydroxyhexadecanoylcarnitine (HR: 0.49, 95 % CI: 0.29 to 0.83), and 2,3-Dinor-TXB1 (HR: 0.56, 95 % CI: 0.33 to 0.95) to be significantly associated with OS. Based on the observation of accumulation in amino acids, immunohistochemistry (IHC) staining revealed that the amino acid transporters SLC7A5/LAT1, SLC1A5/ASCT2, and SLC16A10/MCT10 were up-regulated in ESCC cancerous tissues when compared to non-cancerous equivalents. Consistently, the same panel of amino acids were downregulated in cells with SLC1A5 knockdown. Herein, it is concluded that this study not only identified several metabolites with diagnostic and/or prognostic value, but also provided accurate metabolite-based prediction models for ESCC tissue classification. Furthermore, the three up-regulated amino acid transporters were identified as potential therapeutic targets for ESCC, especially SLC1A5.