PARTNER CONTENT: At the Mobile World Congress (MWC) Doha 2025, the GSMA Mobile AI Community Roundtable was held at the Doha Exhibition & Convention Center in Qatar. Mobile AI Community Group was established in March of this year by leading global operators, vendors, and AI industry partners. It aims to explore mobile AI applications, incubate next-generation intelligent networks and fully unlock the industry’s potential.
During the conference, Mr. Xu Hui, vice president of Huawei’s Wireless 5G-A Domain, delivered a keynote speech on Mobile AI application advancement and correspondent strategies for cellular network evolution. He stated that the development of Mobile AI has gone beyond public imagination, and there will be agents for people, vehicles, and embodied AI to reshape everyone’s life, catalyzing an intelligent revolution across all industries. Meanwhile, the wireless network will evolve towards agent-centric, building a dynamic, layered channel to bring modal-level assurance for AI applications.
Mobile AI Developments Beyond Imagination, Demanding Connectivity Upgrade
By 2035, the network will be able to accommodate approximately 900 billion intelligent agents for 9 billion people. This means a hundred-fold increase in communications network capacity.
——HUAWEI <Intelligent World 2035>
Personal Agent: In the consumer market, the public has seen a rapid growth of AI-native terminals from Flagship smartphones to more affordable models since 2024. During this period, AI wearables such as smart glasses, earphones, and watches have emerged one after another. The AI applications on these devices have quickly evolved into personal intelligent agents, unlocking individual potential, transforming them from simple tools into intelligent companions. Beyond voice commands, multimodal interactions such as video chat have been commercialized. These advancements demand faster connectivity and a shift from downlink-centric networks to networks with balanced uplink and downlink capacity.
Vehicle Agent: In travel scenarios, Vehicle agents are also elevating consumer experience. In intelligent cockpits, passengers could enjoy a variety of high-quality and content-rich entertainment enabled by AI agent assistants and built-in 5G connectivity. The driving experience itself is transformed by Level 2 and above autonomous driving systems. With continuous uploading and training of the road data, drivers could gradually move towards a “ZERO takeover” driving experience.
Embodied Agent: Embodied intelligence has become the new focal point of the AI industry. Agents are expected to unlock billions of robots and robo-dogs as “silicon labor” across all scenarios. In 2025 alone, humanoid robots have demonstrated capabilities in a wide range of scenarios including home, work, retail, and delivery. In November, the first real-world demo of 5G-A-powered embodied intelligence robots on an actual power grid have been released. According to CAICT (China Academy of Information and Communications Technology), robots will be deployed at scale in industrial scenarios where multi-agent cooperation is enabled by a shared cloud-based embodied foundation model. 5G-A-inside has become the new standard configuration, allowing robots to operate without remote controls. Moreover, 5G-A’s ultra-low latency transmission enables robots to utilize cloud-based foundation models in real time, which not only upgrades the model size from 10 to 273 billion parameters, but also extends battery endurance.
Mobile AI Catalyzing New 5G-A Service Models, Calling for Agent-Centric Solutions
The rise of Mobile AI applications is changing how 5G-A networks deliver services. Multimodal interactions have shifted the traffic patterns to continuous data pulses, which requires the network to be more flexible and responsive. The frequent use of agents has shifted the data flow model to become more uplink-centric. The increasing demand for data upload and real-time response necessitates larger uplink bandwidth and lower latency from 5G-A. In addition, think tank reports indicate that device density can exceed one million connections per square kilometer, while the rapid growth of the AI agents demands much higher concurrency and longer online durations.
In response to these changes, the telecom industry should collaborate to build an Agents over NR (AoNR) network. This network would support the agent-based services by providing versatile and ultra-uplink broadband. Furthermore, as an end-to-end solution, AoNR should function as a dynamic, layered channel, offering modal-level assurance to ensure each type of multimodal content can efficiently and reliably reach cloud-based foundation model in parallel.
Realizing these benefits calls for a joint effort. Xu stressed that a superior mobile AI experience will depend on collaboration between equipment vendors, carriers, and AI companies. “We need to work together to build an agent-centric ecosystem and incubate the AoNR network foundation,” Xu said. “Only in this way can we achieve mutual success in the mobile AI era.”