There is limited information on the relation between sleep duration and incident atrial fibrillation. We aimed to investigate this association in a Chinese population using cohort data from a study in Kailuan. The analysis included 87,693 participants (age range, 18-98 years) free of atrial fibrillation at the baseline survey. Participants were divided into three categories according to self-reported sleep duration: ≤6.0 hours, 7 hours (ref), ≥8.0 hours. Atrial fibrillation diagnosis was made on a standard 12-lead electrocardiogram and via self-reported history. Cox proportional hazards models were used to calculate hazard ratio (HR) and confidence interval (CI) for atrial fibrillation, according to sleep duration. During median follow-up of 7.89 (range, 6.36-8.57) years, 322 cases of atrial fibrillation had occurred. Using 7 hours of sleep as the reference group, multivariable adjusted HRs (95% CI) for atrial fibrillation were 1.07 (0.75-1.53), 1.0 (ref), and 1.50 (1.07-2.10), from lowest to highest category of sleep duration. Secondary analysis showed no evidence of interactions between sleep duration and sex and snoring on the risk of incident atrial fibrillation (p = 0.75/0.25). We conclude long sleep duration may be a potential predictor/marker for incident atrial fibrillation.