Explore how modern scale-out architectures are being redesigned to meet the demanding requirements of high-performance computing and artificial intelligence workloads in this 20-minute conference talk from the Hot Interconnects Symposium. Discover the fundamental shifts needed in distributed computing infrastructure as traditional HPC approaches evolve to accommodate AI's unique computational patterns and data movement requirements. Learn about the architectural innovations that enable efficient scaling across thousands of nodes while maintaining the low-latency, high-bandwidth characteristics essential for both scientific computing and machine learning applications. Examine the interconnect technologies, memory hierarchies, and system designs that are reshaping how we approach large-scale parallel computing in the era of AI-driven workloads.