Huawei has introduced an optical network strategy designed to align fixed infrastructure with the throughput demands of AI clusters.
Unveiled during Network X 2026 in Vienna, Austria, the approach addresses challenges caused by rising token processing and expanded parameter scales across modern computing installations.
Speaking at the Broadband Development Congress organised by the World Broadband Association, Bob Chen, President of Huawei Optical Business Product Line, outlined a framework based on three core concepts: rebuilding network architecture, reshaping business models, and redefining network operations and maintenance.
“Facing the AI era, we must shift our philosophy to ‘Rebuild’ network architecture, ‘Reshape’ business model, and ‘Redefine’ network O&M,” explains Chen. “Through these three ‘R’s, we can reshape UBB network value and jointly move towards new growth in the AI era!”
Network operators face shifting traffic flows as AI models approach 10 trillion parameters. Data presented by Chen indicates that token consumption climbed nearly 50-fold over the past year, while the total tally of dedicated AI data centres expanded nearly ten-fold.
In China, traffic linked to the Zhongwei national computing hub rose from approximately 0.1 percent of national volume to over 10 percent of the country’s total bandwidth in intelligent computing traffic.
To support these volumes, Huawei proposes replacing traditional administrative regional backbones with architectures centred directly on computing nodes. The physical layout organises connections between Core, Regional, and Edge data centres.
Direct one-hop mesh configurations establish deterministic latency rings calibrated to 1ms, 5ms, and 20ms. The sub-two-millisecond threshold addresses the technical threshold required for real-time visual rendering.
The transition alters commercial packaging for telecoms carriers. Chen stated that enterprise offerings must move beyond basic bandwidth delivery to metrics covering latency commitments and uptime reliability.
Providers are combining transport capacity with computing power and software applications into unified enterprise contracts. These enterprise connections deploy Optical Cross-Connect, Automatically Switched Optical Network capabilities, and quantum encryption to maintain transmission isolation.
For domestic subscriptions, Huawei detailed a strategy designated as “One Fibre, One Network, One Home”, incorporating Fibre to the Room equipment alongside Wi-Fi mesh nodes. Integrated intelligence identifies local frequency interference to adjust radio frequency parameters, increasing local transfer rates by roughly 20 percent under congested conditions.
“In addition, home AI agents are reshaping home broadband interaction models—allowing users to interact in real time, control home networks, and manage IoT devices through simple voice commands,” Huawei explained.
Chen outlined an automated operational model divided across network elements, overall transport paths, and customer services. The maintenance system categorises diagnostics into a structured hierarchy:
Self-healing routines: Over-the-air client issues, such as station roaming drops, trigger automated remediation routines.
Remote adjustments: Engineers handle configuration faults from the central network operations
centre using diagnostic assistants.
Directed field dispatches: On-site dispatches for damaged physical fibre rely on operational copilot software to locate cable faults.
The platform also targets facility energy overheads within optical distribution nodes. Automated line adjustments dynamically modify consumption levels across optical line terminal chassis and integrated line cards, reducing total hardware energy requirements by more than 20 percent.
(Copyright: IoT News Huawei outlines AI-ready optical networks for connected infrastructure)
Chinese
Japanese
Korean
