The purpose was to examine the association between exercise during pregnancy and intelligence score in offspring. We analyzed data from 4008 women from the Aarhus Birth Cohort, Aarhus University Hospital, Denmark, recruited during pregnancy from July 1989 to November 1991 and their sons who were registered at conscription at 17-20 years of age. The women gave information by self-administered questionnaires during the first trimester including leisure-time physical activity and weekly hours of sport. This information was linked to the sons’ measures of intelligence by Børge Priens test scores from the Danish Conscription Registry. Only sons were included since very few women register at conscription in Denmark. The main outcome measure was the Børge Priens test score as a continuous variable and with a low score defined as <10% of the population score. Analyses were adjusted for maternal body mass index, years in school, and smoking. Sons of women with light and moderate to heavy leisure-time physical activity had lower risk of having a low intelligence score compared with sons of women with sedentary activity: adjusted odds ratio (aOR) 0.66 (95% CI 0.49;0.88) and 0.46 (95% CI 0.23;0.93), respectively. Furthermore, sons of women engaged in sports had lower risk of a low intelligence score: aOR 0.50 (95% CI 0.30; 0.83) for 1-2 h/wk and 0.62 (95% CI 0.35; 1.10) for ≥3 h/wk compared with no weekly sports activity. In conclusion, a higher level of physical activity during pregnancy was associated with a lower risk of low intelligence score in early adulthood in sons.