Genetics, environment, and their interactions impact autism spectrum disorder etiology. Smoking is a suspected autism spectrum disorder risk factor due to biological plausibility and high prevalence. Using two large epidemiological samples, we examined whether autism spectrum disorder was associated with prenatal paternal smoking in a Discovery sample (N = 10,245) and an independent Replication sample (N = 29,773). Paternal smoking was retrospectively assessed with questionnaires. Likelihood of having autism spectrum disorder was estimated with the Autism Spectrum Screening Questionnaire at three levels: low (<10), intermediate (10-14), and high ([greater than or equal to]15). Ordinal regression was used to examine the relationship between prenatal paternal smoking and likelihood of having autism spectrum disorder, adjusting for confounders. A total of 36.5% of Discovery sample fathers and 63.3% of Replication sample fathers smoked during the pregnancy period; 7% of the Replication sample smoker fathers smoked during the pre-conception period but quit during pregnancy period. Discovery sample prenatal paternal smoking significantly increased the likelihood of having autism spectrum disorder in their offspring (adjusted odds ratio=1.27). This was confirmed in the Replication sample with adjusted odds ratio of 1.15 among smoking pre-conception period + pregnancy period fathers; 14.4% and 11.1% increased high likelihood of autism spectrum disorder was attributable to prenatal paternal smoking in Discovery sample and Replication sample, respectively. Smoking prevention, especially in pregnancy planning, may decrease autism spectrum disorder risk in offspring.