The objective was to examine the role of total and beverage-specific alcohol consumption on the incidence of type 2 diabetes mellitus (DM) among elderly men and women. We studied prospectively 4655 participants of the Cardiovascular Health Study who were free of DM at baseline. Alcohol consumption was obtained at baseline and during follow-up examinations. DM was defined using fasting glucose and/or use of hypoglycemic medications. We used Cox proportional hazard models to estimate adjusted relative risks of diabetes across alcohol categories. During a mean follow-up of 6.3 years, 234 incident cases of DM were documented. Compared with never drinkers, hazard ratios [95% confidence interval (CI)] for DM were 0.7 (0.3 to 1.4), 0.5 (0.3 to 0.9), 0.6 (0.4 to 1.1), and 0.8 (0.4 to 1.3) for former drinkers and current drinkers of <1, 1 to 6, and 7+ drinks per week, respectively, for men after adjustment for age, BMI, education, and smoking. Corresponding values for women were 1.2 (0.6 to 2.3), 0.7 (0.4 to 1.1), 0.6 (0.3 to 1.1), and 0.4 (0.2 to 1.0), respectively. A reduced risk of DM was observed with all types of beverage consumed. Similar findings were observed when we repeated the above analyses using simple or weighted cumulative alcohol update and covariates over time. Light to moderate alcohol consumption was associated with a lower incidence of DM among elderly people, irrespective of the type of beverage consumed.