Jeetu Patel of Cisco discusses distributed inference and hybrid AI infrastructure in a live interview at Advanced Micro Devices Advancing artificial intelligence 2026, recorded by theCUBE Research with host Dave Vellante of theCUBE Research. Patel outlines Cisco's strategy for managing the transition from chatbot-driven bursts of inference to sustained agent-driven workloads. They address device-side compute, hybrid inference, Cisco Cloud Control, model routing with AMD compute, agent monitoring, tokenomics and enterprise security considerations for scaling AI. The conversation highlights growing inference demand and a structural shift to continuous agent activity. Patel emphasizes the need for hybrid inference deployments, unified management for security and cost control and telemetry to monitor agent behavior. Vellante notes that intelligent model routing and token cost optimization are critical for enterprises planning scalable controlled AI adoption.