Abstract Observational studies have suggested that physical activity might lower the risk of lung cancer in former and current smokers, but not in never-smokers. Using genetic instruments for self-reported and accelerometer-measured physical activity traits implemented through two-sample Mendelian randomization (MR), we sought to strengthen the evidence for causality. We used 18 genome-wide significant (P < 5 × 10−8) single-nucleotide polymorphisms (SNP) for self-reported moderate-to-vigorous physical activity and seven SNP for accelerometer-measured (“average acceleration”) physical activity from up to 377,234 UK Biobank participants and evaluated these in relation to risk using 29,266 lung cancer cases (including 11,273 adenocarcinomas, 7,426 squamous cell carcinoma, and 2,664 small-cell carcinoma cases) and 56,450 controls. MR analysis suggested no effect of self-reported physical activity [OR (95% confidence interval (CI)) = 0.67 (0.42–1.05); P = 0.081; Q-value = 0.243] and accelerometer-measured activity [OR (95% CI) = 0.98 (0.93–1.03); P = 0.372; Q-value = 0.562] on lung cancer. There was no evidence for associations of physical activity with histologic types and lung cancer in ever and never smokers. Replication analysis using genetic instruments from a different genome-wide study and sensitivity analysis to address potential pleiotropic effects led to no substantive change in estimates. Collectively, these findings do not support a protective relationship between physical activity and the risk of lung cancer. Significance: A new genetic study provides little evidence that recommending physical activity would help prevent lung cancer.