Abstract Background Metabolite-protein interactions (MPIs) are crucial regulators of cancer metabolism; however, their roles and coordination within the esophageal squamous cell carcinoma (ESCC) microenvironment remain largely unexplored. This study is the first to comprehensively map the metabolic landscape of the ESCC microenvironment by integrating an MPI network with multi-scale transcriptomics data. Methods First, we characterized the metabolic states of cells in ESCC using single-cell transcriptome profiles of key metabolite-interacting proteins. Next, we determined the metabolic patterns of each ESCC patient based on the composition of different metabolic states within bulk samples. Finally, the ESCC samples were clustered into unique subtypes. Results Sixteen ESCC metabolic states across 7 cell types were identified based on the re-analysis of single-cell RNA-sequencing data of 208,659 cells in 64 ESCC samples. Each of the 7 cell types within the tumor microenvironment exhibited distinct metabolic states, highlighting the high metabolic heterogeneity of ESCC. Based on differences in the compositions of the metabolic states, 4 ESCC subtypes were identified in two independent cohorts (n = 79 and 119), which were associated with significant variations in prognosis, clinical features, gene expression, and pathways. Notably, the inactivation of cellular detoxification processes may contribute to the poor prognosis of ESCC patients. Conclusions Overall, we redefined robust ESCC prognostic subtypes and identified key MPI pathways that link metabolism to tumor heterogeneity. This study provides the first comprehensive mapping of the ESCC metabolic microenvironment, offering novel insights into ESCC metabolic diversity and its clinical applications.