Detailed molecular phenotyping gives insight into disease mechanisms and can individualize medical practice for improved risk assessment and treatment. We show in an epidemiological study (n = 4309) that the multi-metabolic profiles obtained by serum NMR metabonomics inherently associate with the extent of atherosclerosis already in preclinical stages. Data-driven analysis of the spectral profiles of healthy, young adults revealed three distinct metabolic phenotypes associated with high carotid intima-media thickness (IMT), a surrogate marker of cardiovascular disease. The phenotypes were characterized by varying combinations of well-known metabolic disturbances like elevated VLDL and LDL and low HDL levels. Low IMT was also associated with distinct metabolic phenotypes with lipoprotein as well as other biochemical characteristics partly opposing those found for the high IMT phenotypes. Profiles of low-molecular-weight metabolites quantified from the experimentation were also characteristic for the metabolic phenotypes and substantiate developments toward the use of multi-metabolic risk phenotypes. The methodology can be taken as a direct extension for the routine analytics used for the risk assessment of atherosclerosis; quantification of metabolites will complement and might even replace conventional lipid measurements. Serum NMR metabonomics is therefore anticipated as a rational option for comprehensive cardiovascular risk assessment.