AI researchers are leaving academia for industry at an accelerating rate, creating a compounding talent drain that threatens the open source research pipeline. A Journal of Finance study documented 211 AI faculty departures from North American universities between 2004 and 2018, with the pace increasing as deep learning matured. Beyond individual losses, departing senior researchers take with them the ability to train the next generation of students. Compounding the problem is a 'compute divide' — most academic researchers lack access to the GPU resources needed for frontier AI work, which concentrates capability in a handful of large companies. Industry now produces over 90% of notable AI models, most under restrictive licenses, inverting the open-commons model that produced foundational projects like Spark and Ray. The author argues this shift risks moving AI innovation behind closed doors, leaving governance to institutions with the least impartiality.
