Widespread obesity in the U.S. is a relatively recent phenomenon, reaching epidemic proportions only in the last 15 years. However, existing research shows that while calorie expenditure through physical activity has not changed appreciably since 1980, calorie consumption has risen dramatically. Consequently, any explanation of obesity must address the reason why consumers tend to overeat in spite of somewhat obvious future health implications. This study tests for an addiction to food nutrients as a potential explanation for the obesity epidemic. Specifically, we use a random coefficients (mixed) logit model applied to household scanner data to test a multivariate version of the rational addiction model of Becker and Murphy and Chaloupka. We find evidence of a rational addiction to all nutrients - protein, fat and carbohydrates - as well as to sodium, but particularly strong evidence of a forward-looking addiction to carbohydrates. The implication of this finding is that price-based policies - sin taxes or produce subsidies that change the expected future costs and benefits of consuming carbohydrate-intensive foods - may be effective in controlling excessive nutrient intake.