Huawei, in collaboration with the China Academy of Information and Communications Technology, Thailand's KBTG, and Nigeria's Fidelity, jointly released a research report on Smart Computing and Networking (AI Fabric 2.0)

[China, Shanghai, September 18, 2025] During the Huawei Global Connect 2025 event, the first Data Center Innovation Summit themed "Leading AI DC Innovation, Co-Creating a Smart Future" saw Huawei jointly release the "Intelligent Computing and Networking (AI Fabric 2.0) Research Report" (hereinafter referred to as the "Report") with China Academy of Information and Communications Technology, Thailand's KBTG, and Nigeria's Fidelity. The report outlines the evolution trends and challenges of data center networks in the AI era, defines key technological directions for generational advancements in data center networks through a three-tier architecture of AI brain, AI connection, and AI network elements, and demonstrates Huawei's technological leadership and influence in intelligent computing and networking solutions for the AI era.

The report indicates that with the rapid advancement of AI technology, data centers are evolving from business hubs to computing power centers, while networks are transitioning from traditional connectivity to the era of intelligent computing and networking. AI innovation is driving data center networks toward three key dimensions: ultra-large-scale intensive deployment, single-architecture support for exponential computing power demands, and the integration of general and intelligent computing architectures. Meanwhile, the continuous growth in computing power demands is propelling data center networks toward newer-generation architectures with greater scale, enhanced performance, and higher reliability.

To this end, Huawei has introduced an intelligent computing and networking solution for the AI era. Based on a three-tier network architecture comprising the AI brain, AI connections, and AI network elements, this solution builds a robust network infrastructure foundation for the AI era, enabling enterprises to achieve full computing capacity, sustained business operations, and accelerate their digital and intelligent transformation.

The AI brain restructures the network management model through three-tier automation: "business, network, and operations." At the business layer, the drag-and-drop orchestration tool and over 10 pre-configured scenario APIs reduce the deployment cycle to a weekly level. At the network layer, the integrated network and security architecture, combined with AI policy algorithms, enables automated generation of network and security configurations, achieving 100% accuracy in changes. At the operations layer, a unified application-network system achieves minute-level troubleshooting. Additionally, by integrating knowledge graphs and AI algorithms, it proactively alerts for over 40 high-frequency risks across five major categories, transforming network management from "reactive response" to "proactive prediction.".

AI networking breaks the limitations of traditional passive transmission, reshaping the path of computational power flow with "end-to-network collaboration, full network reliability, and intelligent scheduling." Through end-to-network collaboration and in-network computing technologies, it achieves deep integration between the end and the network. Relying on the iReliable technology of the磐石 high-reliability architecture, it enables millisecond-level fault switching, improving network reliability by 10 times. Meanwhile, the NSLB (network-level load balancing) series algorithms achieve full network load balancing, enhancing training/inference performance by over 10%.

The AI network element addresses the differentiated requirements of AI training (high bandwidth and burstiness) and inference (high concurrency and low latency), achieving over 95% throughput efficiency and sub-400ns latency through FLB balancing and Token balancing algorithms. In terms of security, it integrates inherent security technologies and packet isolation capabilities to comprehensively enhance network security performance.

Looking ahead, Huawei will continue to collaborate with partners and customers to strengthen research and innovation in data center networking, continuously optimize network architectures, and provide enterprises with more efficient and intelligent data center networking solutions, helping industries across all sectors achieve digital and intelligent transformation.