Knowledge graph - Scientific news
NVIDIA has introduced "cuGraph--LLM," a GPU--accelerated library designed to bridge the gap between massive--scale knowledge graphs and Large Language Models. Released on May 21, 2026, this technology enables real--time graph traversal and subgraph sampling for LLM context windows, supporting graphs with trillions of edges. For enterprises, this provides the computational power necessary to run complex "GraphRAG" workflows on global--scale datasets, such as fraud detection networks or global supply chains, reducing latency by up to 100x compared to CPU--based graph processing.