HNS 2026 | Huawei Launches Xinghe AI Fabric Solution to Power Ultra-Resilient, Ultra-Efficient Infrastructure in Southern Africa

[Johannesburg, South Africa, July 23, 2026] At the Huawei Network Summit 2026 (HNS 2026) in Johannesburg, Huawei launched its next-generation data center network solution and products tailored for general-purpose computing and AI computing scenarios. Designed to establish an ultra-resilient, ultra-efficient computing foundation for local enterprises, the solution aims to accelerate regional digital industrial growth.

Arthur Wang, Vice President of Huawei Data Communication Product Line

Arthur Wang, Vice President of Huawei Data Communication Product Line, delivered a keynote address outlining current data center networking trends and Huawei's latest innovations.

General-Purpose Computing: AI-Driven Multi-DC Management and Resilient Architecture

As digital-intelligent transformation accelerates across industries, over 90% of enterprises are projected to migrate to multi-DC architectures by 2027, with 80% adopting multi-vendor strategies. Managing these heterogeneous environments presents O&M challenges. For instance, cross-DC and cross-vendor fault boundaries traditionally take over three hours to isolate, threatening business continuity.

To solve these bottlenecks, Huawei integrated advanced AI capabilities into its StarryWing Digital Map and Rock-Solid Architecture:

• 1-minute fault boundary identification: xFlow intelligent full-flow analysis helps isolate faults across heterogeneous networks within 1 minute.

• Real-time traffic visibility: The AI Eagle-Eye Engine visualizes the quality of up to 200,000 service flows simultaneously and completes root-cause analyses in minutes, significantly improving network resilience.

AI Computing: Bridging the Bandwidth Gap to Maximize Token Output

Global token consumption has surged over the past two years, growing 300-fold daily and transforming DCs into full-scale "token factories." During multi-turn LLM inference, demand for Key-Value (KV) cache storage bandwidth increases 5 to 8 times. However, existing network architectures suffer from an 80% bandwidth deficit, creating a bottleneck where computing capacity stalls while waiting for storage transfers. This delay severely limits Tokens Per Second (TPS).

At the same time, inference workloads demand strict network reliability. A network disruption exceeding seven seconds forces compute cards to swap and tasks to restart, generating downtime costs up to 100 times higher than in traditional general-purpose computing. Maximizing token throughput while maintaining continuous, cost-effective inference operations has become a critical industry challenge.。

To address these demands, Huawei upgraded both the performance and reliability of the Xinghe AI Fabric:

• Compute-network-storage synergy: The Hyper-Converged Fabric (HCF) Architecture boosts network-wide throughput above 98%, driving a 20% increase in inference TPS.

• 5-second fast reboot: CloudEngine AI switches feature iFlashboot 2.0, enabling 5-second reboots to keep inference workloads running continuously without interruption.

• Rapid diagnostics: StarryLink optical modules automatically detect contaminated or loose fiber connections, resolving faults in under a minute and reducing inference interruptions caused by network faults by 90%.

Huawei also introduced the high-density 100T CloudEngine XH9330-128EO AI switch, equipped with 128 × 800GE ports. Built for high bandwidth and performance, it received the Best of Interop 2026 Award for AI Infrastructure.

Customer Perspective: Automated Infrastructure at Scale

During the summit, Thaodi Mapodile, Acting Chief Information Security Officer of PRASA (Passenger Rail Agency of South Africa), shared operational insights from their recent deployment:

"We implemented a two-layer spine-leaf architecture in our data center network for flexible scaling, alongside high-performance Huawei CloudEngine switches equipped with iReliable technology to improve reliability. Using the iMaster NCE platform, PRASA now unifies management, deployment, and O&M across multiple DCs, driving full lifecycle network automation."

Huawei remains committed to deep engagement across Southern Africa, partnering within open ecosystems to advance data center network technology and accelerate enterprise digital-intelligent transformation across the region.