Objective: We aimed to investigate whether liver steatosis severity affects the risk of developing diabetes in a large cohort study. Methods: We prospectively examined the association in 41,650 Chinese adults with negative hepatitis-B surface antigen who were free of alcohol consumption, diabetes, and liver cirrhosis at baseline. Cox proportional models were used to estimate the risk of diabetes after a mean of 3.6 years of follow-up. Nonalcoholic fatty liver disease NAFLD) was assessed with hepatic ultrasonography. Elevated alanine transaminase (ALT) was defined as ALT concentrations >19 and >30 U/L in females and males, respectively. Diabetes was defined as a fasting glucose (3)7.0 mmol/L or treatment with hypoglycemic medication. Results: Liver steatosis severity was significantly associated with higher risks of developing diabetes (adjusted hazard ratio [HR] for severe vs. without NAFLD = 2.66, 95% confidence interval [CI]: 2.17-3.25, P-trend<.001) and impaired fasting glucose (fasting glucose between 5.6 and 6.9 mmol/L, adjusted HR = 1.36, 95% CI: 1.16-1.59, P-trend<.001), as well as a faster increase rate of fasting glucose concentrations (P-trend<.001), during 3.6 years of follow-up. Elevated ALT was also associated with incident diabetes (HR = 1.12, 95% CI: 1.02-1.22), adjusting for NAFLD and other covariates. Conclusion: We observed a dose-response relationship between liver steatosis severity and increased diabetes risk, and ALT may predict incident diabetes independently of NAFLD.