Abstract Although smoking is a major modifiable risk factor for many types of cancer, evidence for colorectal cancer is equivocal in Asian populations. Recent Western studies have proposed that the association between smoking and colorectal cancer is restricted to specific tumor molecular subtypes. However, no studies have evaluated the association according to tumor molecular subtypes in Asian populations. In a Japanese prospective population-based cohort study of 18 773 participants, we collected tumor tissues from incident colorectal cancer cases and evaluated KRAS (Kirsten rat sarcoma viral oncogene homolog) and BRAF (v-raf murine sarcoma viral oncogene homolog B) mutation status using target sequencing. Multivariable-adjusted Cox proportional hazard model was used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) for associations of smoking with the risk of overall colorectal cancer and its subtypes defined by KRAS and BRAF mutation status. Among 339 cases, KRAS and BRAF mutations were identified in 164 (48.4%) and 16 (4.7%) cases, respectively. The multivariable-adjusted HR for ever smoking compared with never smoking was 1.24 [95% CI: 0.93–1.66], 1.75 [1.14–2.68], 0.87 [0.59–1.29], 1.24 [0.93–1.67] and 1.22 [0.38–3.93] for overall, KRAS wild-type, KRAS-mutated, BRAF wild-type and BRAF-mutated colorectal cancer, respectively. The statistically significant heterogeneity was indicated between KRAS mutation status (Pheterogeneity = 0.01) but not between BRAF mutation status. This study is the first to demonstrate that smokers have an approximately 2-fold higher risk of KRAS wild-type colorectal cancer than never smokers in an Asian population. Our findings support that smoking is a risk factor for colorectal cancer, especially for its subtype without KRAS mutations, in Asian populations.