Background: Head and Neck Squamous Cell Carcinoma (HNSCC) is a common head and neck malignancy that originates from lips, mouth, paranasal sinuses, oropharynx, larynx, nasopharynx, and other pharyngeal cancers. Since high-throughput expression data is now available, it is feasible to use global gene expression data to analyze the relationship between metabolic-related gene expression and clinical outcomes in HNSCC patients. Method: In this study, we used RNA sequencing (RNA-seq) data from cancer genomic maps (TCGA) and validated in the GEO dataset to shine the dysfunctional metabolic microenvironment and given potential biomarkers for metabolic therapy.Results: TCGA database contained 529 patients and 327 metabolic genes (198 upregulated and the largest logFC is CA9; 129 downregulated and the largest logFC is CA6. Cox regression model found 51 prognosis‐related genes and we found the most high-risk gene was LCLAT1 (HR=1.144, 95% CI=1.044~1.251); the most low-risk gene was CHDH (HR=0.580, 95% CI=0.400~0.839). We next study whether metabolic-related genes patterns could serve as an early predictor of incidence of HNSCC by ROC curve and the model demonstrated an AUC of 0.745 in TCGA and 0.618 in GEO. Meanwhile, the predictor of clinicopathological in HNSCC was also analyzed and we found the AUC for age, gender, grade, stage, T, M and N were 0.520, 0.495, 0.568, 0.606, 0.577, 0.476 and 0.673, respectively in TCGA and the AUC for age, gender, stage, T, M, N, smoking and HPV16-pos were 0.599, 0.531, 0.622, 0.606, 0.616, 0.550, 0.614, 0.519 and 0.397 respectively in GEO.Conclusion: Our research found a novel metabolic signature gene for HNSCC prognosis prediction based on the TCGA dataset. Our signatures may reflect metabolic microenvironment disorders and provide useful biomarkers for metabolic therapy and response treatment prediction.