Huawei Debuts Stellar AI Network to Secure Agentic Infrastructure

The updated Stellar AI Fabric 2.0 introduces the SF9300 series switches, which utilize a two-layer architecture to eliminate the core layer, effectively cutting deployment costs by 30%. By unifying protocols, the system reduces end-to-end latency to 11 μs. Furthermore, the integration of Link-Layer Retransmission ensures zero packet loss during link flaps, potentially saving 100 hours of annual cluster downtime. Hardware innovations, such as the CloudEngine XH9300-EN series, utilize near-packaged architecture to lower optical-electrical conversion latency by 180 ns.

To address the challenges of cross-region computing, the Stellar AI WAN solution employs the Starnet lossless algorithm, which maintains computing efficiency above 95% over distances exceeding 1,000 km. This setup allows for significant cost optimization, reducing initial branch xPU expenses fourfold while cutting idle bandwidth by 80%. Security is bolstered by built-in Quantum Key Distribution (QKD), which extends transmission security to 80 km without the need for additional hardware.

Finally, the Stellar AI Campus Network addresses the rise of enterprise agent adoption by automating maintenance. Through Deep Causal Pruning algorithms, Huawei has improved troubleshooting accuracy to 95%, slashing mean time to recovery (MTTR) to under five minutes. The Asset Security 2.0 framework further hardens these environments by embedding AI micro-models directly into switches, enabling the detection of malicious agent commands and terminal threats in under 10 seconds.