BACKGROUND: Breast cancer is one of the most prevalent malignant tumors in women. Mitochondria, essential for cellular function, have altered metabolic activity in cancer cells, influencing tumor regulation and clinical outcomes. The connection between mitochondrial metabolism-related genes and breast cancer prognosis remains underexplored. This study aims to investigate the role of these genes in breast cancer by constructing risk models. METHODS: Breast cancer transcriptome data were obtained from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO), and mitochondrial gene data were sourced from the MitoCarta3.01 database. Clustering analysis was conducted using the "ConsensusClusterPlus" package, followed by Gene Set Enrichment Analysis (GSEA), Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. A prognostic model was built using Cox regression and Least Absolute Shrinkage and Selection Operator (LASSO) algorithms. Immune cell infiltration levels were assessed via CIBERSORT and MCPcounter algorithms. Validation of key gene expression was performed on breast cancer tissue specimens and cell models to explore their biological functions in breast cancer cells. RESULTS: The LASSO regression analysis of the TCGA BRCA dataset identified four prognosis-related mitochondrial metabolism genes: MYH11, LTF, FADS2, and PSPHP1. Validation using the GEO dataset confirmed that patients with high-risk scores (based on these four genes) had shorter overall survival compared to those with lower risk scores. Immunological analysis revealed that high-risk patients were less responsive to immunotherapy but more sensitive to conventional chemotherapies. This suggests that combining chemotherapy with immunotherapy might enhance T cell-based treatments. Univariate and multivariate Cox regression confirmed that the mitochondrial gene model was an independent predictor of overall survival, and a nomogram was developed to predict patient prognosis. Tissue validation showed consistent expression patterns with bioinformatic predictions. Functional assays confirmed that FADS2 was highly expressed in breast cancer cells, and its knockout significantly reduced cell invasion, migration, and colony formation. CONCLUSION: This study reveals that mitochondrial metabolism-related genes are closely associated with breast cancer progression, clinical outcomes, and genetic alterations. The findings may offer new avenues for treatment strategies, early intervention, and prognosis prediction in breast cancer.