Firms increasingly look to collaboration with alliance partners in their quest for breakthrough innovation. But how does the position of a firm in its alliance network weighted by the centrality of its partners—a concept which we term “partner‐weighted alliance centrality”—and the heterogeneities in the types of partners that it cooperates with—in terms of its private‐public collaboration—influence this quest? Using longitudinal data from the U.S. pharmaceutical industry, we build alliance networks in the period 1985–2001 to investigate these questions. We show that, for breakthrough innovation, collaborating with more partners that are more central in alliance networks the better, but only to a point. Beyond that point, we find that the likelihood of achieving breakthrough innovation drops. Furthermore, and looking at the kinds of knowledge provided by the partners in each firm's alliances, we report that firms with a greater share of private partners, relative to public partners, suffer less from the diminishing benefits of collaboration with central partners when developing breakthrough innovation. Taken together, we make novel contributions about how to organize for breakthrough innovation, and provide actionable managerial advice in terms of selecting collaborative partners in alliance networks.