Explore the cutting-edge research on foundation models for knowledge graph reasoning in this comprehensive conference talk. Delve into the challenges of designing transferable representations for knowledge graphs and learn about ULTRA, a novel approach for learning universal and transferable graph representations. Discover how ULTRA builds relational representations conditioned on interactions, enabling pre-trained models to generalize inductively to unseen knowledge graphs with arbitrary entity and relation vocabularies. Examine the results of link prediction experiments conducted on 57 different knowledge graphs, showcasing ULTRA's zero-shot inductive inference performance. Gain insights into the potential of fine-tuning to further enhance performance. The talk covers background information, introduces foundation models for graph reasoning, explains the ULTRA approach in detail, presents experimental results, and concludes with a Q&A session.