Cluster policies have been recently called into question in the aftermath of several empirical evidences. Disentangling how market and network failures arguments play together in cluster policy design, we look for more robust micro foundations of network structuring in clusters. Our aim is to show that, in spite of this growing skepticism, new opportunities for cluster policy exist. They require moving their focus from the “connecting people” one best way that gets through the whole of cluster policy guidelines, to more surgical incentives for R & D collaborations, which favor suited structural properties of local knowledge networks along the life cycle of clusters.