<p>Scientific collaboration among universities is traditionally studied by counting coauthored papers with different types of partners. Under this approach, the internal mechanisms of collaboration remain hidden and, without additional normalization, the results can be biased. In our study, we apply co-authorship network analysis to examine collaboration structures within university research networks. We use OpenAlex data from 2017 to 2019 for thirty universities of different ages and status. Our approach constructs normalized co-authorship networks by applying two bipartite normalization approaches (Standard and Strict), and then examines two Ps-core variants: (i) a fixed-size core comprising approximately 100 of the most productive and collaborative authors, and (ii) a fixed-strength core consisting of authors whose fractional contribution is for at least one paper. Our results show that the analyzed universities exhibit high levels of intra-organizational collaboration but markedly different collaboration structures. These differences are reflected in authors’ connectedness, the strength of collaboration, and the balance between external and internal affiliations. Established universities display more diverse collaboration structures than younger and catching-up universities. We discuss the potential benefits and risks associated with differing core compositions and propose policy recommendations based on our findings.</p>