Shift from AI training to real-time inference requires distributed, CPU-heavy infrastructure to handle agentic workloads.