This study examines the network of supply and use of significant innovations across industries in Sweden, 1970-2013. It is found that 30% of innovation patterns can be predicted by network stimulus from backward and forward linkages. The network is hierarchical, characterized by hubs that connect diverse industries in closely knitted communities. To explain the network structure, a preferential weight assignment process is proposed as an adaptation of the classical preferential attachment process to weighted directed networks. The network structure is strongly predicted by this process where historical technological linkages and proximities matter, while human capital flows and economic input-output flows have conflicting effects on link formation. The results are consistent with the idea that innovations emerge in closely connected communities, but suggest that the transformation of technological systems are shaped by technological requirements, imbalances and opportunities that are not straightforwardly related to other proximities.