STEM PhDs are a critical source of human capital in the economy, contributing to commercial as well as academic science. We examine whether STEM PhD students become new inventors (file their first patent) during their doctoral training at the top 25 U.S. universities (by patenting). We find that 4% of PhDs become new inventors. However, among PhDs of faculty who are themselves top (prolific) inventors, this figure rises to 23%. These faculty train 44% of all the new inventor PhDs by copatenting with their advisees. We also explore whether new inventor PhDs are equally distributed by gender. In our university sample, the female share of new inventors is 9% points (pp) lower than the female share of PhDs. Several channels contribute to this: First, female PhDs are less likely to be trained by top inventor advisors (TIs) than male PhDs. Second, they are less likely to be trained by (the larger number of) male top inventors: The estimated gap in the female % of PhDs between female and male TIs is 7 to 9 pp. Third, female PhDs (supervised by top inventors and especially by other faculty) have a lower probability of becoming new inventors relative to their male counterparts. Notably, we find that male and female top inventors have similar rates of transforming their female advisees into new inventors at 4 to 8 pp lower (17 to 26% lower rate) than for male advisees. The gap remains at 4 pp comparing students of the same advisor and controlling for thesis topic.