We document that since 1997, the rate of startup formation has precipitously declined for firms operated by US PhD recipients in science and engineering. We explore how increasing knowledge complexity can be associated with fewer science-based startups. The decline in startup formation is accompanied by an earnings decline, increasing work complexity in R&D, and more administrative work for science-based founders. Founding a startup appears to have become increasingly harder over the past 20 years, while established firms are becoming more attractive workplaces for PhDs.