Objective To investigates the metabolic changes in serum between patients with normal cognition of cerebral small vessel disease (CSVD) and those with cognitive impairment of CSVD. It aims to identify distinct metabolic pathways of CSVD-related cognitive impairment, which can provide new research directions for the diagnosis and treatment of this disease. Methods Serum samples from CSVD patients diagnosed in the Department of Neurology, West China Hospital of Sichuan University between July 2021 and December 2023 were used in this study. According to the patients’ Montreal Cognitive Assessment scores, they were divided into cognitively unimpaired CSVD group and cognitively impaired CSVD group. Untargeted metabolomic detection was performed using ultra-high performance liquid chromatography-tandem mass spectrometry. Quality control of the metabolomic data was conducted through correlation analysis of quality control samples. This study constructed an orthogonal partial least squares-discriminant analysis model to examine the relationship between metabolites and sample groups. Different metabolites were selected based on the criteria of P<0.05 and variable importance in projection >1, which were then subjected to metabolic pathway enrichment analysis. Results A total of 157 CSVD patients were included, including 51 cognitively unimpaired CSVD patients and 106 cognitively impaired CSVD patients. Untargeted metabolomics analysis, conducted in both positive and negative ion modes, identified a total of 68 significantly different metabolites between the cognitively unimpaired and cognitively impaired CSVD groups. These metabolites primarily consisted of lipids and lipid-like molecules, amino acids and their metabolites, and steroid hormones. Among these, the serum levels of 21 metabolites were increased in patients with CSVD-related cognitive impairment, while the levels of 47 metabolites were decreased. Further enrichment analysis revealed that these differential metabolites were predominantly enriched in 11 metabolic pathways, which included signaling pathways such as sphingolipid metabolism, protein digestion and absorption, and amino acid biosynthesis. Conclusions Compared with cognitively unimpaired CSVD patients, those with cognitive impairment showed increased levels of endogenous sphingolipids, such as phytosphingosine, and decreased levels of essential amino acids, including valine and leucine, in their serum. This suggests that lipid metabolism reprogramming and energy metabolism disturbances may be the main metabolic features in CSVD-related cognitive impairment. These different metabolites not only serve as promising biomarker candidates for CSVD-related cognitive impairment, but also offer new directions for investigating its pathological mechanisms.