In an era of explosive growth in AI computing power demand, the underlying network architecture supporting all of this is facing a once-in-a-decade major transformation. At this year's MWC 2026, Huawei not only proposed the concept of "AI-centric all-optical target network" at a forum hosted by the European analyst firm IDATE, but also simultaneously launched its "next-generation WAN target network architecture" and a series of innovative all-optical network products. Huawei's core strategy focuses on the dual-drive of "AI for Networks" and "Networks for AI," assisting global telecom operators in building the ultimate network infrastructure capable of supporting the "Internet of Agents."
AI-centric all-optical target network: Four major upgrades in bandwidth, reliability, coverage, and latency.
At the Green All-Optical Network Forum, Jin Zhiguo, Vice President of Huawei's Optical Product Line, elaborated on four key initiatives for building an "AI-centric all-optical target network":
• A revolution in home networking experience:Based on a Wi-Fi 7 tri-band FTTR (Fiber to the Room) solution, it achieves a stable 4Gbps transmission rate and seamless roaming with 10ms latency throughout the house. Combined with AI anti-interference technology, it can even boost transmission speeds by 20% in environments with severe signal interference.
• End-to-end ultra-low latency:By extending all-optical switching (OXC) to metropolitan area networks and utilizing next-generation PON slicing technology to establish dedicated connection channels for VIP applications, Huawei aims to create an ultra-low latency circle of "5ms for national backbone, 3ms for regional networks, and 1ms for metropolitan areas" to meet the extreme demand of AI services for "millisecond-level computing".
• Eliminate backbone bandwidth bottlenecks:Continuing to advance towards 400G/800G specifications, through 3D-Mesh architecture and Super C+L technology, it efficiently supports the data center interconnect (DCI) and cross-border massive traffic transmission needs.
• "Zero interruption" high reliability:By combining algorithms and eOTDR technology, network vulnerabilities can be accurately predicted, and innovative liquid crystal materials can be used to achieve ultra-fast switching of the light layer in as little as 50 milliseconds, ensuring that AI services can continue to operate online.
The Two-Way Race Between AI and the Internet: Innovation in All-Optical Networks
Chen Banghua, president of Huawei's optical product line, further pointed out that the key to upgrading the next generation of optical networks lies in "AI for Networks" and "Networks for AI".
Huawei leverages AI models to achieve intelligent sensing of optical fibers (fault location accuracy within 10 meters), improved performance evaluation accuracy (20% increase in transmission distance), and deep green energy saving (full hibernation when there is no traffic, with average energy savings of up to 40%). On the operation side, it has also launched the Home Broadband Operation and Maintenance Intelligent Agent (FANSpirit) and the Optical Transmission Intelligent Agent (OTNSpirit), which can assist engineers in quickly troubleshooting more than 60 types of faults through natural language dialogue, significantly reducing maintenance costs.
Next-Generation WAN Target Network: Multi-Dimensional Perception and Intrinsic Security
In addition to all-optical networks, Wang Lei, President of Huawei's Data Communications Product Line, also presented the "Next-Generation WAN Target Network Architecture," emphasizing a comprehensive upgrade of the IP bearer network across three dimensions:
• Multidimensional sensing (accurate identification and lossless transmission):To address the challenges of ensuring a smooth user experience due to the surge in encrypted network traffic, Huawei has launched the "StarNet Recognition Engine," claiming to improve the accuracy of encrypted stream recognition by over 95%. On the enterprise side, the StarNet lossless algorithm ensures "lossless" transmission of computing data over long distances of hundreds of kilometers, thereby driving the commercialization of integrated computing and network capabilities.
• Security resilience (endogenous and quantum security):Introducing a smart intrinsic security board that can block abnormal behavior and prevent route hijacking in real time, and more forward-lookingly launching the "intrinsic QKD board" to achieve "two-network integration" of quantum key distribution (QKD) and communication network, thereby resisting future quantum hacking threats.
• Network Autonomy (Netmaster Agent):By importing Netmaster network intelligence as the "brain," a comprehensive upgrade of the network is achieved, moving from passive response to proactive operation and maintenance.
