Abstract Aims We performed a Mendelian randomization (MR) study to elucidate the associations of ever smoking, lifelong smoking duration, and smoking cessation with heart failure (HF) risk. Methods and results We extracted genetic variants associated with smoking initiation, age at initiation of regular smoking, cigarettes per day, and smoking cessation from the genome-wide association study and Sequencing Consortium of Alcohol and Nicotine use (1.2 million individuals), as well as a composite lifetime smoking index from the UK Biobank (462 690 individuals). The associations between smoking phenotypes and HF were explored in the Heart Failure Molecular Epidemiology for Therapeutic Targets Consortium (47 309 cases; 930 014 controls) employing inverse variance-weighted meta-analysis and multivariable MR. The mediation effects of coronary artery disease and atrial fibrillation on smoking–HF risk were explored using mediation analysis. The odds ratios (ORs) for HF were 1.28 [95% confidence interval (CI), 1.22–1.36; P = 1.5 × 10−18] for ever regular smokers compared with never smokers and 1.25 (95% CI, 1.09–1.44; P = 1.6 × 10−3) for current smokers vs. former smokers. Genetic liability to smoking more cigarettes per day (OR, 1.37; 95% CI, 1.20–1.58; P = 6.4 × 10−6) and a higher composite lifetime smoking index (OR, 1.49; 95% CI, 1.31–1.70; P = 2.5 × 10−9) were associated with a higher risk of HF. The results were robust and consistent in all sensitivity analyses and multivariable MR after adjusting for HF risk factors, and their associations were independent of coronary artery disease and atrial fibrillation. Conclusions Genetic liability to ever smoking and a higher lifetime smoking burden are associated with a higher risk of HF.