Childhood obesity is one of the biggest public health threats facing the UK, and video game use is considered a risk behavior for obesity among children. However, few studies have explored the prospective association between video game use and body mass index (BMI) or the potential mediators of this association. To investigate whether a longer-term association exists between video game use at a young age and BMI SD score in later years, independent of television use, and to ascertain whether this association is mediated by other energy-balance behaviors. This cohort study is a secondary analysis of data from the Millennium Cohort Study, a nationally representative sample of children who were born in the UK between September 1, 2000, and January 31, 2002, that focused on data collected when the children were aged 5, 7, 11, and 14 years. Data for all variables, except BMI, were provided by parental or caregiver reporting if the children were younger than 14 years of age. At age 14 years, the children self-reported their own behavior. Initial data analysis was conducted between September 18, 2018, and September 28, 2018, with supplementary analyses conducted from October 7, 2019, to November 22, 2019. The main outcome variable was BMI SD scores, with video game use as the exposure variable of interest. Physical activity, bedtime regularity, sugar-sweetened beverage consumption, and high-calorie food consumption were included as potential mediating behaviors. The full sample comprised 16 376 children and had a nearly equal number of boys (8393 [51.3%]) and girls (7983 [48.7%]). Every 1 SD increase in the number of hours of video game use at age 5 years was associated with a β = 0.018 higher BMI SD score at age 14 years (95% CI, 0.004-0.032). A small partial mediation of this association was found (direct association: β = 0.017 [95% CI, 0.003-0.031]; indirect association: β = 0.0011 [95% CI, 0.0003-0.0019]), suggesting that irregular bedtimes and higher consumption of sugar-sweetened beverages were mediators. The mediation model accounted for 36.7% (95% CI, 35.5-37.8) of the variance of the BMI SD score at age 14 years. Results of this study suggest a small (and not clinically meaningful) association between video game use in early childhood and higher BMI in later years, which may be mediated by irregular bedtimes and higher consumption of sugar-sweetened beverages. Future interventions to prevent childhood obesity should incorporate health promotion in mainstream video games to target children most at risk because of their high level of video game use.