China Unicom's Wang Guangquan: AI Reshapes Broadband Access Evolution, U-CAN Architecture Builds a 10-Gigabit Foundation

C114, September 18 — The 27th China International Optoelectronic Exposition (CIOE) was successfully held at the Shenzhen World Exhibition & Convention Center last week. At the concurrent "10G Optical Access: AI+ Intelligent Upgrade and Application Forum," Wang Guangquan, chief expert at China Unicom Research Institute, delivered a keynote titled "AI Empowerment: Evolution of China Unicom Broadband Network Technology and Exploration of Service Innovation."

In his keynote, Wang Guangquan offered an in-depth analysis of the changes and challenges facing access networks in the AI era. He detailed the core technology system, innovation pathway and deployment scenarios of China Unicom's self-developed ultra-converged all-optical access network U‑CAN, and outlined the evolution of next-generation access networks driven by deep synergy among 10G optical networks, AI and computing power.

Wang Guangquan noted that the access network, with its dual core roles as the network edge and the service entry point, is a key vehicle for carriers' digital infrastructure. Traditional access networks mainly connect people to people, with relatively relaxed requirements for latency, jitter and reliability. But with the spread of 10-gigabit optical networks, machine interconnection, embodied intelligence and large-scale deployment of AI Agents, massive numbers of machine-type connections that are high-frequency, low-latency, highly reliable and concurrent have become the mainstream, and the original network architecture, O&M model and service capabilities can no longer meet new service demands. At the same time, the access network has long suffered from industry pain points such as difficult fault location, high O&M costs, inefficient network energy saving and uneven home network experience. The deep penetration of AI technology provides a core breakthrough for intelligent access network upgrades, better experience and lower O&M costs.

Responding to these new service shifts and industry pain points, China Unicom has formally proposed the U‑CAN hyper-converged all-optical access network architecture to meet the demands of ubiquitous 10G coverage, computing-network convergence and multi-service integration. The aim is to build a new-generation intelligent access system featuring ubiquitous 10G, whole-home intelligent networking, device-edge computing-network coordination and end-to-end reliability, precisely matching fragmented, personalized and high-quality converged service scenarios. Wang Guangquan emphasized that the core implementation logic of the U‑CAN architecture is the deep integration of the access network with BNC capabilities, using controllers to automate the scheduling of bandwidth resources, computing resources and slicing capabilities. This delivers true end-to-end 10G connectivity, intelligent slicing enablement and automated storage-computing coordination, addressing the core challenges of dynamic resource allocation and multi-service concurrency assurance in the 50G high-bandwidth era.

At the core technology level, U‑CAN has built five supporting technology systems — millimeter wave, AI agents, integrated sensing and communications, device-cloud-edge collaboration, and a broadband core network — comprehensively addressing the shortcomings of traditional access networks.

To address heavy Wi-Fi interference, weak indoor sensing capabilities and limited 6G spectrum, China Unicom is focusing on indoor short-range millimeter-wave technology development. Rather than pursuing wide-area cellular millimeter-wave, it is prioritizing high-precision, low-interference, indoor-only coverage solutions, using FTTR optical networking for seamless roaming and compatibility with future Wi-Fi evolution. Related standardization work is now progressing in an orderly manner.

At the same time, the architecture deeply integrates AI Agent technology to reshape access network operations and service systems. Leveraging large model capabilities, it enables precise fault identification, rapid localization and intelligent handling, effectively addressing pain points such as redundant alarms, inefficient localization and difficult experience optimization in traditional operations and maintenance, while comprehensively improving home network determinism and data security.

In terms of integrated sensing and communication, China Unicom has connected multi-dimensional sensing systems spanning devices, ODN and Wi-Fi. Combined with the low-power, low-cost characteristics of millimeter-wave technology, this significantly strengthens integrated network sensing capabilities, laying the foundation for scenario-based intelligent services.

On compute-network convergence, U-CAN establishes a new end-cloud-edge collaboration mechanism. Leveraging distributed edge computing in home gateways and OLTs, it enables AI training and inference tasks to run close to the terminal side, reducing cloud data transmission pressure and service latency to meet new compute scheduling and intelligent service carrying needs. In addition, by optimizing the broadband core network architecture and retaining existing network functions, it activates core BNC capabilities to enable agile provisioning and intelligent scheduling of bandwidth, latency and SLAs, delivering all-optical access that converges broadband, intelligence and agility.

According to Wang Guangquan, China Unicom's U-CAN technology has completed full-chain product deployment from chips and modules to terminals, and achieved breakthroughs in commercial pilots across multiple scenarios. In Malanshan, Changsha, a lossless-compression ultra-high-definition video transmission solution has been deployed to efficiently support high-bandwidth audiovisual services such as 8K and XR. In the smart home sector, the company is building out smart health and elderly care scenarios using terminal devices such as millimeter-wave sensing and smart mattresses. Combined with AI agents and edge computing, this enables integrated analysis of home data and low-latency intelligent interaction, while linking with home robots to significantly speed up terminal intelligent service responses.