This article investigates whether attending a sequence of racially diverse schools predicts science, technology, engineering, and mathematics (STEM) college outcomes. Such a relationship is important because of the increasingly diverse population of school-aged children who are likely to attend racially segregated K-12 schools and colleges, the benefits for individuals and society of STEM college graduates, and the projected shortage of people trained for future STEM workforce demands. We use a unique panel data set (N = 14,980) of the University of North Carolina graduates. Our main analytical approach is multilevel modeling to examine the relationship between attending a sequence of racially diverse educational institutions and the odds of declaring and/or graduating with a STEM major. We find that students who attended a diverse sequence of schools are more likely to declare and graduate with STEM majors than those who did not. Framing our results with theories of cumulative advantage and intergroup contact theories, we offer science education policy reform recommendations.