Abnormal alanine aminotransferase level (ALT) levels might be associated with type 2 diabetes, but whether higher ALT levels within the normal range predict the risk is unknown. Methods: We followed a community-based cohort of 3446 individuals who were ≥35 years old without diabetes and hepatitis B or C in southern Taiwan for 8 years (1997–2004) to study the risk for type 2 diabetes with different normal ALT levels. Results: Among the 337 incident diabetes cases, 16.0% were from those with ALT levels <10 IU/L, 44.5% with ALT levels 10–19 IU/L, 30.0% with ALT levels 20–39 IU/L, and only 9.5% with ALT levels ≥40 IU/L. A cumulative hazard function test showed that the higher the ALT levels, the greater the cumulative incidence rate of diabetes (p < 0.001, log-rank test). A multiple Cox proportional hazards analysis showed that increasing age, lower educational levels, higher body mass index levels (≥25 vs. <25), and higher ALT levels (vs. the reference group, ALT <10 IU/L), from hazard ratio (HR) = 1.8, for ALT = 10–19, HR = 3.7 for ALT = 20–39, to HR = 4.5 for ALT ≥40, were significant factors for developing diabetes (p < 0.001). The hazard ratios of higher ALT levels in the participants without alcohol consumption were similar to or higher than those in the total cohort. Conclusion: Higher ALT levels, even within the normal range, are strong predictors of type 2 diabetes independently of body mass index levels with a dose–response relationship.