Acute myeloid leukemia (AML) is a life-threatening hematological disease. Novel diagnostic and prognostic markers will be essential for new therapeutics and for significantly improving the disease prognosis. To characterize the metabolic features associated with AML and search for potential diagnostic and prognostic methods, here we analyzed the phenotypic characteristics of serum metabolite composition (metabonome) in a cohort of 183 patients with de novo acute myeloid leukemia together with 232 age- and gender-matched healthy controls using (1)H NMR spectroscopy in conjunction with multivariate data analysis. We observed significant serum metabonomic differences between AML patients and healthy controls and between AML patients with favorable and intermediate cytogenetic risks. Such differences were highlighted by systems differentiations in multiple metabolic pathways including glycolysis/gluconeogenesis, TCA cycle, biosynthesis of proteins and lipoproteins, and metabolism of fatty acids and cell membrane components, especially choline and its phosphorylated derivatives. This demonstrated the NMR-based metabonomics as a rapid and less invasive method for potential AML diagnosis and prognosis. The serum metabolic phenotypes observed here indicated that integration of metabonomics with other techniques will be useful for better understanding the biochemistry of pathogenesis and progression of leukemia.