Objective: To develop a lipid metabolism-associated gene signature to stratify sepsis patients for prognostic risk and evaluate their immune function. Methods: Based on the sepsis dataset GSE65682, lipid metabolism-related genes were screened. Univariate Cox regression, LASSO regression, and multivariate Cox regression analyses were integrated to identify hub genes. Patients were divided into high- and low-risk groups based on the median risk score. Kaplan-Meier and ROC curve analyses were used to evaluate predictive performance. Additionally, an independent external dataset, GSE95233, was introduced to further validate the robustness of the model. Immune function differences were analyzed using ssGSEA, CIBERSORT, and correlation network analysis. Results: A 9-gene prognostic signature was constructed, consisting of AHRR, CLN8, FASN, LSS, MED29, PAFAH1B1, PIP5K1C, TRIB3, and UGCG. Patients in the high-risk group showed poorer survival (KM test P=6.75×10??, area under the ROC curve AUC =0.951), with enrichment of chemokine receptor signaling and para-inflammatory responses. The low-risk group exhibited higher levels of tumor-infiltrating lymphocytes, type II interferon response, regulatory T cells, and macrophage infiltration. Immune network analysis revealed synergy between activated NK cells and M1 macrophages (r=0.44) and antagonism with resting NK cells (r=-0.62). CD86 and TNFSF4 were highly expressed in the low-risk group, while the enrichment score of CD200R1 was significantly reduced (P<0.05). The model maintained robust predictive performance in the independent external validation dataset GSE95233 (AUC=0.757). Conclusion: The established lipid metabolism-based 9-gene signature effectively predicts sepsis outcomes, revealing significant immune differences between risk groups. It provides a potential tool for risk stratification and personalized clinical intervention.