BACKGROUND High cumulative tobacco consumption is associated with short telomeres and with increased all-cause mortality. We tested the hypothesis that high tobacco consumption is causally associated with short telomeres and with increased all-cause mortality. METHODS We studied 55,568 individuals including 32,823 ever smokers from the Danish general population, of whom 3430 died during 10 years of follow-up. All had telomere length measured, detailed information on smoking history, and CHRNA3 rs1051730 genotype, which is associated with tobacco consumption, determined. In a Mendelian randomization study, we conducted observational, genetic, and mediation analyses. RESULTS First, tobacco consumption was 21.1 pack-years in non-carriers, 22.8 in heterozygotes and 24.8 in homozygotes (P-trend<0.001). Second, the observational multivariable adjusted hazard ratio for all-cause mortality was 1.12 [95% confidence interval (CI): 1.09, 1.15] per doubling in tobacco consumption. In Mendelian randomization analysis, the hazard ratio was 1.08 (1.02, 1.14) per minor CHRNA3 allele in ever smokers. Third, in observational analysis telomeres shortened with -13 base pairs (-18, -8) per doubling in tobacco consumption. In Mendelian randomization analysis, the estimate was +3 base pairs (-10, +15) per minor CHRNA3 allele. Finally, individuals with the shortest vs longest telomeres had a multivariable adjusted hazard ratio of 1.30 (1.13, 1.50) for all-cause mortality; however, in mediation analysis short telomeres explained only +0.4% (-3.5%, +4.3%) of the association between high tobacco consumption and increased all-cause mortality. CONCLUSIONS High tobacco consumption is causally associated with increased all-cause mortality. High cumulative tobacco consumption is associated with short telomeres observationally, but there is no clear genetic association.