Cisco Nvidia AI factory expands to 200kW liquid-cooled rack scale

The expansion into rack-scale architecture addresses the critical deployment gap between acquiring high-end GPUs and maintaining production-ready AI environments. By integrating liquid cooling and unified management planes, Cisco and Nvidia are providing the high-density infrastructure required for reasoning and agentic AI that standard data centers struggle to support. For the streaming ecosystem, this signifies a shift toward localized, high-performance AI factories capable of processing massive datasets for real-time video personalization and autonomous content operations. Watch for how enterprise adoption of these 200kW+ liquid-cooled racks influences future data center power and cooling standards.

The rack-scale AI factory market is intensifying as multiple vendors race to deliver integrated compute, networking, and cooling solutions for enterprise AI workloads. In August 2026, Blue Planet and Telefónica Deutschland completed a joint proof of concept using agentic AI to power 5G network slicing services, integrating Blue Planet AI Studio with Telefónica Germany's multi-domain service orchestration environment. The PoC demonstrated that tasks such as defining slice specifications and generating standards-compliant service payloads were completed in minutes instead of weeks, illustrating the operational demands that rack-scale AI infrastructure like the Cisco Nvidia AI factory is designed to support at scale.

On the business and competitive front, Cisco's expansion into rack-scale AI factories comes as rivals face their own structural pressures. Ericsson announced the cutting of approximately 100 technical and support jobs in Canada as part of a broader global cost-cutting strategy, consolidating managed services into global hubs amid a slowdown in 5G spending across North America. Meanwhile, BCE's Bell AI Fabric initiative secured a USD $220 million sovereign AI GPU agreement supplying compute capacity for Cohere, signaling that telecom operators are positioning their data center and fiber assets as foundational AI infrastructure rather than pure connectivity utilities. These moves underscore a broader industry trend where network operators and infrastructure vendors are competing to capture AI compute revenue streams.

From a technical perspective, the challenges of coordinating AI workloads with real-time communication are driving research into new scheduling frameworks. A team demonstrated HAFS, a coordinated networking framework for on-device agent-augmented real-time communication, achieving 1.5x higher video quality while reducing agent response time by 31%, built atop WebRTC and llama.cpp across devices including MacBook Pro, Samsung Galaxy S25, and NVIDIA Jetson platforms. The framework addresses contention between concurrent traffic flows generated by humans for live video streaming and agents for context analysis, a problem that becomes more acute as rack-scale AI factories like those from Cisco and Nvidia push agentic AI workloads closer to production environments handling real-time video and communication tasks.