Most of the studies about industrial clusters and innovation stress the importance of firms’ geographical proximity and their embeddedness in local business networks (BNs) as factors that positively affect their learning and innovation processes. More recently, scholars have started to claim that firm-specific characteristics should be considered to be central in the process of learning and innovation in clusters. This article contributes to this latter direction of research. It applies social network analysis to explore the structural properties of knowledge networks in three wine clusters in Italy and Chile. The results show that in spite of firms’ geographical proximity and the pervasiveness of local BNs, innovation-related knowledge is diffused in clusters in a highly selective and uneven way. This pattern is found to be related to the heterogeneous and asymmetric distribution of firm knowledge bases in the clusters.