Using the historical random assignment of MBA students to peer groups at a top business school in the United States, I study the effect of the gender composition of a student's peers on the gender pay gap at graduation and long-term labor market outcomes. I find that a 10 percentage point increase in the share of male peers leads to a 2.1 percent increase in the relative earnings of female students at graduation, closing the gender gap in earnings at graduation by two-thirds. The effects on women's long-term earnings grow even larger with time. Using novel data on job offers, I find that two different mechanisms drive the effects on short- and long-term earnings. Women with a greater share of male peers take more quantitative coursework in business school and receive job offers at graduation in occupations, industries, and firms associated with higher wages, longer hours, and greater earnings growth. However, the effect of male peers on women's earnings at graduation is primarily driven by female students' increased willingness to accept the maximum salary offered within their offer set. In contrast, peer-induced effects on human capital alone place female students on dramatically different long-term expected earnings paths due to changes in the initial occupation, initial industry, and initial firm accepted at graduation. This change in the characteristics of the first job at graduation largely explains the effect of peer gender composition on long-term outcomes. [Additional financial support from the Brookings Institution's Rubenstein Fellowship.]