Huawei Proposes AI-Native Architecture for Future Agentic Networks

Traditional mobile networks, built for voice and mobile internet, lack the intelligence to manage autonomous agents. To bridge this gap, Huawei envisions a core network that embeds AI directly into its interfaces, functions, and procedures. This transition shifts the network from a passive transport layer to an active environment capable of interpreting intent and scheduling computing resources in real time.

To facilitate this, the company proposes three distinct service frameworks: Agent Over New Radio to minimize task latency for portable devices, Wide-Area Computing Networks to standardize resource distribution, and Agent Computing Networks for secure, multi-agent collaboration across ecosystems. These models rely on a new Data & AI Plane that operates alongside existing control and user planes to align network resources with specific task requirements.

Execution of this vision involves two distinct evolutionary stages. The first focuses on overlaying AI capabilities onto existing connectivity, while the second establishes a fully native architecture that moves beyond communication to provide integrated computing and agent-coordination services. Success depends on deep synergy between hardware, chips, and network infrastructure to support the shift from transporting bits to enabling task-oriented tokens.