In this paper we use a newly constructed dataset following 30,000 Italian individuals from high school to labor market and we analyze whether the gender composition of peers in high school affected their choice of college major, their academic performance and their labor market income. We leverage the fact that the composition of high school classmates (peers), within school-cohort and teacher-group, was not chosen by the students and it was as good as random. We find that male students graduating from classes with at least 80% of male peers were more likely to choose "prevalently male" (PM) college majors (Economics, Business and Engineering). However, this higher propensity to enroll in PM majors faded away during college (through transfers and attrition) so that men from classes with at least 80% of male peers in high school did not have higher probability of graduating in PM majors. They had instead worse college performance and did not exhibit any difference in income or labor market outcomes after college. We do not find significant effects on women.