Patients with advanced stomach adenocarcinoma (STAD) commonly show high mortality and poor prognosis. Increasing evidence has suggested that basic metabolic changes may promote the growth and aggressiveness of STAD; therefore, identification of metabolic prognostic signatures in STAD would be meaningful. An integrative analysis was performed with 407 samples from The Cancer Genome Atlas (TCGA) and 433 samples from Gene Expression Omnibus (GEO) to develop a metabolic prognostic signature associated with clinical and immune features in STAD using Cox regression analysis and least absolute shrinkage and selection operator (LASSO). The different proportions of immune cells and differentially expressed immune-related genes (DEIRGs) between high- and low-risk score groups based on the metabolic prognostic signature were evaluated to describe the association of cancer metabolism and immune response in STAD. A total of 883 metabolism-related genes in both TCGA and GEO databases were analyzed to obtain 184 differentially expressed metabolism-related genes (DEMRGs) between tumor and normal tissues. A 13-gene metabolic signature (GSTA2, POLD3, GLA, GGT5, DCK, CKMT2, ASAH1, OPLAH, ME1, ACYP1, NNMT, POLR1A, and RDH12) was constructed for prognostic prediction of STAD. Sixteen survival-related DEMRGs were significantly related to the overall survival of STAD and the immune landscape in the tumor microenvironment. Univariate and multiple Cox regression analyses and the nomogram proved that a metabolism-based prognostic risk score (MPRS) could be an independent risk factor. More importantly, the results were mutually verified using TCGA and GEO data. This study provided a metabolism-related gene signature for prognostic prediction of STAD and explored the association between metabolism and the immune microenvironment for future research, thereby furthering the understanding of the crosstalk between different molecular mechanisms in human STAD. Some prognosis-related metabolic pathways have been revealed, and the survival of STAD patients could be predicted by a risk model based on these pathways, which could serve as prognostic markers in clinical practice.