This study examined the longitudinal association between changes in sugar-sweetened and/or caffeinated beverage consumption and smoking/vaping behaviour among Canadian adolescents. Using longitudinal data from the COMPASS study (2015/16 to 2017/18), four models were developed to investigate whether beverage consumption explained variability in smoking and vaping behaviour in adolescence: (1) smoking initiation, (2) vaping initiation, (3) current smoking status, and (4) current vaping status. Models were adjusted for demographic factors. Multinomial logit models were used for model 1, 2, and 3. A binary logistic regression model was used for model 4. An association between change in frequency of beverage consumption and smoking/vaping behaviour was identified in all models. A one-day increase in beverage consumption was associated with smoking initiation (OR = 1.38, 95% CI: 1.25, 1.51), vaping initiation (OR = 1.23, 95% CI: 1.14, 1.32), identifying as a current smoker (OR = 1.17, 95% CI: 1.01, 1.35), and currently vaping (OR = 1.08, 95% CI: 1.04, 1.11). Change in high-energy drink consumption was the best predictor of smoking behaviours and vaping initiation but not current vaping status. Given the health consequences of smoking and vaping and their association with high-energy drink and coffee consumption, policy initiatives to prevent smoking/vaping initiation, and to limit youth access to these beverages, warrant consideration.