Background: Observational studies have reported an association between childhood obesity and a higher risk of multiple sclerosis (MS). However, the difficulties to fully account for confounding and long recall periods make causal inference from these studies challenging. The objective of this study was to assess the contribution of childhood obesity to the development of MS through Mendelian randomization, which uses genetic associations to minimize the risk of confounding. Methods: We selected 23 independent genetic variants strongly associated with childhood body mass index (BMI) in a genome-wide association study (GWAS) which included 47,541 children. The corresponding effects of these variants on risk of MS were obtained from a GWAS of 14,802 MS cases and 26,703 controls. Standard two-sample Mendelian randomization methods were performed, with additional sensitivity analyses to assess the likelihood of bias from genetic pleiotropy. Results: The inverse-variance weighted MR analysis revealed that one standard deviation increase in childhood BMI increased odds of MS by 26% (odds ratio=1.26, 95% confidence interval 1.10-1.45, p=0.001). There was no significant heterogeneity across the individual estimates. Sensitivity analyses were consistent with the main findings and provided no evidence of pleiotropy. Conclusions: This study provides genetic support of a role for increased childhood BMI in the development of MS.