“The industry is exploring the concept of "inferencing grids" that would dynamically route AI inference requests based on real-time factors like power costs and available compute capacity.”