In many models of endogenous growth, differentiated product lines are produced using improvable technologies. These technology clusters are central to innovation-led growth, but there is a lack of data-driven research on technology cluster dynamics. I introduce a new method of identifying technology clusters in the data based on a match between patent text and Wikipedia articles. A patent belongs to a technology cluster defined by the category of the Wikipedia article that is most similar to the text of the patent. I build a network of relations for technology clusters using links between Wikipedia categories. I apply these new data to the study of technology cluster network structure, entry and evolution since 1980. I show that young and small firms are more likely to discover new technology clusters by undertaking breakthrough innovations. However, it is mostly big and experienced firms that contribute to the growth of existing technology clusters. This growth is a result of reallocation of technology clusters from less productive to more productive firms. While these results hold for any given year, I find that the role of small and young firms as a source of technology cluster entry has declined in the last decades, in line with other measures of firm dynamics and competition documented in the literature. Reallocation of technology clusters among firms has also slowed down which explains the pervasive decline in the growth of technology clusters over time.