We aimed to explore the associations between serum retinol and all-cause mortality among people with prediabetes and diabetes. The study included 2582 participants with prediabetes and 1654 with diabetes aged ≥40 years from the National Health and Nutrition Examination Survey 2001-2006. Serum retinol was collected from laboratory tests and categorized into five groups, including <50, 50-60, 60-70, 70-80, and ≥80 μg/dL. Deaths were obtained by linkage to National Death Index up to December 31, 2019. Cox proportional hazards models were used to estimate the associations between serum retinol and all-cause mortality. During the follow-up, 993 participants with prediabetes died and 874 participants with diabetes died. There were U-shaped associations between serum retinol and mortality among participants with prediabetes and diabetes, separately. Among participants with prediabetes, compared to serum retinol levels of 50-60 μg/dL, the hazard ratio (HR) (95% confidence interval [CI]) of mortality was 1.40 (95% CI 1.11 to 1.76) and 1.26 (95% CI 1.00 to 1.57) for serum retinol <50 or ≥80 μg/dL, respectively. Among participants with diabetes, compared to serum retinol levels of 50-60 μg/dL, the hazard ratio (HR) (95% confidence interval [CI]) of mortality was 1.25 (95% CI 0.96 to 1.62) and 1.21 (95% CI 0.91 to 1.62) for serum retinol <50 or ≥80 μg/dL, respectively. The U-shaped associations between serum retinol and mortality still existed among participants aged ≥60 years with prediabetes or diabetes but were not statistically significant among those aged 40-59 years with prediabetes or diabetes. In conclusion, both low and excessive serum retinol tended to be with higher mortality risk among people with abnormal blood glucose.