University spin-offs (USOs) have been the subject of considerable interest in the literature and practice. This stems from their importance in addressing entrepreneurial opportunities and enriching universities, the business landscape and societies. Nevertheless, an important issue remains and that is to understand the heterogeneity of USOs. In this article, we argue that an analysis of grant proposals using topic modeling can remedy this weakness for USOs, and how a focus on self-perceived competences brings into focus important development issues that are neglected in prior research. Employing established unsupervised machine learning technique on 108 USO grant proposals, we develop a typology of USOs based on their self- perceived first-order technological and second-order R&D competences, and their first-order customer and second-order marketing competences. Our analysis suggests eight types of USOs. The findings have a number of implications for research and practice, enabling to identify USO types and relate them to entrepreneurial processes and expected outcomes in academic entrepreneurship as well as supporting policy makers’ target-oriented design of funding instruments for academic entrepreneurship.