Particle.news
Download on the App Store

Cisco and Nvidia Expand AI Infrastructure With N9100 Switches and Hyperfabric AI Nearing Launch

The tie-up targets faster AI cluster rollouts across neocloud, sovereign cloud, plus telecom deployments.

Image
Image

Overview

  • Cisco introduced the N9100 Series data center switches built on Nvidia Spectrum-X silicon, delivering 51.2 Tbps with a choice of NX-OS or SONiC and an NCP-compliant reference architecture, with orders expected before year-end and general availability slated for early 2026.
  • Cisco said Nexus Hyperfabric AI will be orderable in November 2025 to automate fabric design, generate cabling plans, and guide installation, cutting AI cluster deployment from months to weeks.
  • A new Cisco Cloud Reference Architecture uses Silicon One switches with an add-on license enabling Nvidia Spectrum-X capabilities on Cisco platforms, pairing adaptive routing and congestion-aware load balancing with SuperNIC-based congestion management to improve AI job completion times.
  • Under the Secure AI Factory with Nvidia, Cisco highlighted UCS C880A M8 and UCS X-Series systems with HGX B300, Spectrum-X SuperNICs, and BlueField-3 DPUs as the compute foundation for Cisco AI PODs supporting training, fine-tuning and inference.
  • Cisco, Nvidia and partners also unveiled what they describe as an AI-native wireless stack for 6G, positioning an American AI-RAN approach for future mobile and pre-6G applications.