Huawei is proposing a new method to evaluate service quality for AI applications called AI MOS, modeled after the Mean Opinion Score used to measure voice service quality
In sum – what to know:
Traffic shift ahead – Huawei expects AI agent applications to generate far more uplink traffic than traditional mobile services, forcing networks designed around downlink-heavy usage to evolve.
Connection surge – Huawei estimates AI agent connections could reach 100 billion, potentially creating major new infrastructure demands across mobile networks, cloud platforms and edge computing systems.
New AI metrics – The company is proposing AI MOS as a way to evaluate AI service quality and A2A-T protocols to support agent-to-agent communication across telecom networks.
Huawei told MWC last week that the rapid emergence of AI agents will require significant changes to how mobile networks are designed, particularly in areas such as uplink capacity, service evaluation, and communication between agents.
Speaking during a media and analyst roundtable, Zhao Dong, vice president and chief marketing officer of the company’s wireless network product, said the telecom industry is entering what the “agent era”, where AI agents interact with users, devices, and other agents across different environments.
“We already stand at the dawn of the agent era,” he said. “Agents go to your mobile phone, go to cars, go to robots.”
AI agents will significantly increase the number of connections on telecom networks, he suggested. Huawei estimates that intelligent agent connections could reach 100 billion in the future, creating new infrastructure requirements for operators.
One of the most significant changes highlighted by Zhao concerns traffic patterns. He said traditional mobile networks were designed mainly to handle downlink traffic, since most digital services involved downloading content. “The traditional capabilities of networks focus more on downlink,” Zhao said. “This time agents need more uplink.”
He illustrated the shift with an example from an AI-powered museum guide trial in Shanghai. In the trial, visitors used mobile AI agents to analyze artworks by sending photos and video to the system. In that scenario, Zhao said uplink traffic accounted for 63% of total traffic, compared with traditional mobile networks where downlink traffic typically represents about 90%.
“This very big change means the network needs to evolve,” Zhao said.
Huawei has previously discussed the use of AI to improve network operations. Zhao said the focus is now shifting to designing networks that directly support AI applications. “Simply put, last year we talked about AI for wireless networks, and this year we talk about networks for AI,” Zhao said.
Huawei describes this concept as an “agent-oriented network”, which aims to support large-scale AI agent interactions and services. Huawei is proposing a new method to evaluate service quality for AI applications called AI MOS, modeled after the Mean Opinion Score used to measure voice service quality.
“AI MOS is just like the MOS for voice service,” Zhao said. “It’s an evaluation criterion for service experience.” Huawei is discussing the concept with industry partners and standards organizations including the International Telecommunication Union.
The company is also proposing a telecom-focused extension of agent-to-agent communication protocols called A2A-T. “A2A is the protocol for communications between agents,” Zhao said. “In the telecom industry we defined A2A-T.”
Zhao also said the development of AI agent services will involve collaboration across the broader technology ecosystem. “Being the terminal provider, chip maker, vendor or service provider, they are all making efforts to facilitate the prosperity of the ecosystem,” Zhao said.
