Huawei Urges Operators to Rebuild Networks Around AI Data Centers

Three shifts for the AI era

The World Broadband Association (WBBA) hosted the Broadband Development Congress (BDC) during Network X 2026 in Vienna. Bob Chen, President of Huawei Optical Business Product Line, gave a keynote titled "Reshaping Network Value for New Growth in the AI Era."

Chen argued that every service boom of the past three decades, from voice to the Internet to video, has depended on network support. According to Huawei, large AI models now feature nearly 10 trillion parameters, and token consumption has surged nearly 50-fold over the past year. He said the objects, frequency and quality of network connections have changed dramatically as a result.

His proposal rests on three "R"s:

• Rebuild network architecture, moving from networks centered on administrative regions to networks centered on AIDCs.
• Reshape the business model, from selling bandwidth alone to selling richer services.
• Redefine network O&M, from semi-automated operations toward full-lifecycle intelligence.

From administrative regions to AIDC-centered networks

Chen noted that networks have traditionally been organized around administrative regions, split into national backbone, provincial backbone and metro layers. He said the number of AIDCs has grown nearly 10-fold, and traffic is now determined by computing power rather than population.

As an example he cited Zhongwei, China. Its traffic once made up only about 0.1% of the country's total. Today Zhongwei is one of China's eight national computing hubs, and its intelligent computing traffic accounts for over 10% of national bandwidth, making it a core node in the AIDC interconnect network. According to Chen, similar cases are appearing worldwide.

Huawei's recommendation is to organize networks in layers around Core DCs, Regional DCs and Edge DCs. Direct one-hop mesh connections between data centers, it says, enable deterministic 1-5-20 ms latency circles and rapid time-to-market.

New business models for enterprise and home

For enterprise customers, Chen described an AI industrial design firm in China that uses AI rendering to shorten production cycles. Real-time AI rendering requires latency below 2 milliseconds. Operators, he said, can upgrade private lines from bandwidth alone to "Bandwidth + Latency + Reliability," supported by OXC, ASON and Quantum Encryption. They can go further and sell "Network + Computing + Model/Application" as integrated computing-network services.

For home broadband, Huawei proposes a "One Fiber, One Network, One Home" strategy. One Network means high-quality whole-home Wi-Fi using FTTR or Wi-Fi Mesh. One Home means smart applications powered by home AI terminals and AI agents. Chen added that home AI agents let users interact in real time and control home networks with simple voice commands.

AI-driven operations and maintenance

On O&M, Chen said end-to-end intelligence across planning, construction, maintenance, optimization and operation should be delivered through network element intelligence, network intelligence and service intelligence. He gave three examples, all figures from Huawei's presentation:

• AI Wi-Fi tuning: it identifies interference sources and types and adjusts transmit power and other parameters in real time, boosting Wi-Fi speeds by approximately 20% in interference environments.
• AI energy saving: intelligent traffic analysis dynamically adjusts power for OLT chassis and line cards, reducing total device energy consumption by over 20%.
• Fault self-healing: home broadband issues are split into three tiers. Air-interface problems such as roaming failures are fixed automatically by O&M AI agents. Configuration problems are handled remotely at the NOC with AI agent assistance. Physical fiber issues require on-site visits, with faults localized using O&M Copilot.

What it means for operators

The talk reflects Huawei's view that operators should plan networks around computing hubs and sell latency and reliability alongside bandwidth. In his closing remarks, Chen said: "Through these three 'R's, we can reshape UBB network value and jointly move towards new growth in the AI era!" The source gives no deployment timelines or commercial terms, so the proposals remain a strategic framework rather than a product announcement.