Background: Low serum potassium appears to be independently associated with incident type 2 diabetes, and low dietary potassium is more common in African Americans than in whites. Objective: We hypothesized that low serum potassium contributes to the excess risk of diabetes in African Americans. Design: We analyzed data collected from 1987 to 1996 from the Atherosclerosis Risk in Communities (ARIC) Study. At baseline, we identified 2716 African American and 9493 white participants without diabetes. We used multivariate Cox models to estimate the relative hazards (RHs) of incident diabetes related to baseline serum potassium during 9 y of follow-up. Results: Mean serum potassium concentrations were lower in African Americans than in whites at baseline (4.2 compared with 4.5 mEq/L; P < 0.01), and African Americans had a greater incidence of diabetes than did whites (26 compared with 13 cases/1000 person-years). The adjusted RHs (95% CI) of incident diabetes for those with serum potassium concentrations of <4.0, 4.0-4.4, and 4.5-4.9 mEq/L, compared with those with serum potassium concentrations of 5.0-5.5 mEq/L (referent), were 2.28 (1.21, 4.28), 1.97 (1.06, 3.65), and 1.85 (0.99, 3.47) for African Americans and 1.53 (1.14, 2.05), 1.49 (1.19, 1.87), and 1.27 (1.02, 1.58) for whites, respectively. Racial differences in serum potassium appeared to explain 18% of the excess risk of diabetes in African Americans, which is comparable with the percentage of risk explained by racial differences in body mass index (22%). Conclusions: Low serum potassium concentrations in African Americans may contribute to their excess risk of type 2 diabetes relative to whites. Whether interventions to increase serum potassium concentrations in African Americans might reduce their excess risk deserves further study.