Under the Hood: Scaling AI infrastructure
34:45
Microsoft’s Deepak Bansal and Cisco’s Will Eatherton discuss the shift to AI-first infrastructure. They explore using DPUs and Sonic to improve network efficiency and cost-per-token, while analyzing the competition between hyperscalers and Neoclouds.
Related Videos
In AI Networking
-
Play video Cisco Nexus: Unified AI Networking for Cloud and On-Premises
Cisco Nexus: Unified AI Networking for Cloud and On-Premises
Cisco Nexus unifies Silicon One, systems, optics, and software for AI. Intelligent packet flow delivers 38% better performance, and agentic operations provide 20% more accurate reasoning. Keep GPUs productive, finish jobs faster, eliminate idle time.
3:53
-
Play video Cisco N9000: 51.2T Switching Built for Distributed AI Workloads
Cisco N9000: 51.2T Switching Built for Distributed AI Workloads
Distributed AI workloads demand next-generation infrastructure. Cisco N9000 switches powered by Cisco Silicon One P200 deliver 51.2T capacity, high-density 800G performance, and security on every port—all unified under Cisco Nexus One.
2:00
-
Play video Building Scalable AI Infrastructure: Neoclouds, GPU Clusters & Operational Excellence
Building Scalable AI Infrastructure: Neoclouds, GPU Clusters & Operational Excellence
Explore how Neoclouds are scaling AI infrastructure with GMI Cloud and Cisco. Learn about managing GPU clusters, simplifying operations with unified visibility, and deploying NCP-compliant architectures for predictable, high-performance AI services.
30:23
-
Play video The Future of AI Networking: Intelligence, Performance, and Security
The Future of AI Networking: Intelligence, Performance, and Security
Transform fragmented data centers into unified AI infrastructure. Cisco Nexus One offers breakthrough performance for massive clusters, intelligent packet flow, agentic operations, and security woven into the fabric. When networks match AI ambitions.
2:21
-
Play video Cisco Optics: 1.6T and 800G LPO for Power-Efficient AI Networks
Cisco Optics: 1.6T and 800G LPO for Power-Efficient AI Networks
Cisco expands its optics portfolio with 1.6T OSFP and 800G Linear Pluggable Optics. Rigorous qualification ensures GPU cluster reliability while LPO cuts optics power by 50% and system power by 30%. Trusted performance and resilience at scale.
2:55
-
Play video Cisco Silicon One G300: Intelligent AI Networking at 102.4 Terabits per Second
Cisco Silicon One G300: Intelligent AI Networking at 102.4 Terabits per Second
Meet Cisco Silicon One G300—purpose-built for demanding AI workloads with 102.4 Tbps connectivity. Features industry-leading shared buffer, fabric-aware load balancing and programmable architecture.
1:48
