Large Language Models (LLMs) have emerged as versatile tools capable of addressing a wide range of text-related tasks and extending to various domains through task-to-text mappings. Simultaneously, Graph Neural Networks (GNNs) have unlocked unprecedented advancements by leveraging the inherent structure of graph-encoded data to tackle complex tasks, from protein folding (AlphaFold) to solving NP-hard problems in operations research. Graphs are omnipresent in domains like biology, chemistry, infrastructure, and knowledge systems, making them a key modality for AI applications. While LLMs have already integrated multiple modalities, such as images and videos, incorporating graph-based reasoning into foundation models remains unexplored. This convergence offers immense potential to redefine both LLM capabilities and state-of-the-art graph machine learning techniques, presenting a unique opportunity to lead in this nascent area of research and its industrial applications.