Building knowledge graphs for 20 years: Should we keep doing this?

Over the past two decades, this community has developed ontology engineering methodologies and produced tools for constructing, querying and exploiting knowledge graphs, which are nowadays core infrastructure in both enterprise and web-scale systems. But as new paradigms like foundation models, vector databases, and large-scale machine learning gain dominance, the role of symbolic, structured representations is being reassessed. In this keynote, I aim to reflect on 20 years of  building, using, and evangelizing knowledge graphs. Have they lived up to their promise? Where have they delivered real-world value, and where have they fallen short? I will explore the changing landscape of data, semantics, and AI, and consider whether knowledge graphs are still relevant—or if we need to radically rethink our approach. Should we continue building them, evolve them, or abandon them altogether? This talk aims to spark a critical conversation about the next generation of semantic infrastructure in the age of intelligent systems.