Little research has been done to examine the temporal relationship between serum ferritin and the development of nonalcoholic fatty liver disease. The aim of this study was to examine whether serum ferritin levels predict incident fatty liver in non-diabetic men. The study cohort comprised 2410 healthy Korean male who were aged 30 to 59years old with no evidence of ultrasonographically detectable fatty liver (USFL) at baseline. Alcohol intake was assessed with a self-reported questionnaire. At each visit, biochemical and anthropometric measurements and abdominal ultrasonography were done. Cox proportional hazard models were used to calculate the adjusted hazard ratios in separate models for USFL. During 7545.9 person-years of follow-up, 586 participants developed USFL. The hazard ratio (95% confidence interval) for incident USFL comparing the highest quartile of serum ferritin level to the lowest quartile was 1.54 (1.21–1.94) after adjusting for age, body mass index, smoking, alcohol intake, and exercise. That association remained significant after further adjustment for cardiovascular risk factors and in time-dependent models. The association between serum ferritin and incident USFL was still significant in the non-overweight group or the no current smoker group. Serum ferritin level was an independent risk factor of incident fatty liver detected by ultrasonography even in non-obese, healthy Korean men. Increased serum ferritin levels appear to be an early predictor for incident fatty liver.