Huawei is urging telecom operators to shape network investment around AI data centers, saying the location of computing demand is changing traffic flows and opening new ways to sell services.
Huawei Optical Business Product Line President Bob Chen presented the proposal at the Vienna congress on Wednesday, Oct. 14, an event hosted by the World Broadband Association. Huawei outlined three priorities: placing network infrastructure near AI data centers, selling enterprise connections based on bandwidth and performance, and using AI in network operations and maintenance.
Huawei said demand is increasingly shaped by where computing power is installed rather than by population distribution. It estimated token use was nearly 50 times its level a year earlier and said large AI models approach 10 trillion parameters. The company also said the number of AI data centers had increased nearly 10-fold. Huawei compared the shift with earlier demand cycles for voice, internet and video services, which it said helped drive network development.
To illustrate the change, Huawei cited Zhongwei, China. The company said the city once accounted for about 0.1% of national traffic, while intelligent-computing traffic there now exceeds 10% of China’s total bandwidth. The figures use different measures for the past and present.
For businesses, Huawei proposed combining network access with computing and applications, with services reflecting latency and reliability as well as bandwidth. It cited an unnamed Chinese industrial-design firm that needs latency below 2 milliseconds for real-time AI rendering. For households, Huawei pointed to whole-home Wi-Fi and smart-home services. It presented these as opportunities for operators and did not report an operator adopting Chen’s proposal. Huawei’s FAQ separately cites China Telecom’s AI Smart Home service as an existing example.
Huawei said AI-based Wi-Fi tuning can raise speeds by about 20% when interference affects a connection. In a separate claim, it said traffic-based power adjustments for optical line terminal (OLT) chassis and line cards can reduce total device energy use by more than 20%.
