REF: https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2026/m08/cisco-secure-ai-factory-nvidia-rack-scale.html
AI Infrastructure
Cisco Expands Secure AI Factory with NVIDIA to Power Rack-Scale AI Infrastructure
On August 25, 2026, Cisco announced an expansion of its Cisco Secure AI Factory with NVIDIA solution to meet the rapidly growing demand for AI infrastructure, partnering with Supermicro to bring high-performance AI servers — both liquid-cooled and air-cooled — into Cisco's full-stack AI architecture.
The expansion comes as organizations worldwide accelerate their investment in AI infrastructure and shift from experimenting with AI to running it in production, particularly for workloads that demand large amounts of GPU and compute power, such as training large AI models, inference, and agentic AI.
Cisco says that modern AI infrastructure shouldn't compete on compute power alone — it also needs to handle networking, cooling, security, data privacy, operational efficiency, and scalability all at once.
Cisco Brings AI Compute and Networking Together in a Full-Stack Architecture
At the core of this announcement is the integration of Supermicro's high-density AI compute systems into the Cisco Secure AI Factory with NVIDIA architecture.
The solution is designed to let organizations build AI infrastructure that works cohesively from compute through networking to management, without having to design each layer separately.
Supermicro will offer server systems that support both air cooling and liquid cooling, enabling higher-density AI clusters than before.
This capability matters a great deal for next-generation AI workloads: newer GPUs deliver more performance, but also demand more power and generate more heat. Cisco sees rack-scale infrastructure as playing a key role in building the AI factories of the future.
What Is Rack-Scale AI, and Why Does It Matter?
Rack-scale AI is an infrastructure design approach that treats the entire rack as a single system, rather than looking at individual servers or GPUs on their own. A system like this needs the following components working together:
AI compute and GPUs
High-speed networking
Power infrastructure
Cooling systems
Storage
Security
Monitoring and management
As AI workloads grow larger, simply adding more GPUs may not deliver proportional gains in performance, since data traveling between GPUs and compute systems has to move across the network at high speed. That's why network architecture has become just as critical to AI infrastructure as the GPUs themselves.
Cisco is bringing its networking expertise together with AI compute to build infrastructure that can support large AI clusters in a systematic way.
Supporting AI From Trillion-Parameter Training to Inference
Cisco says the newly expanded solution will support a wide range of AI use cases, from training very large AI models to deploying them in real-world applications.
One key use case is trillion-parameter model training, which requires large numbers of GPUs along with high-bandwidth, low-latency networking. It also supports high-throughput inference for organizations that need to process large volumes of AI requests simultaneously.
The system can be applied to a variety of AI infrastructure use cases, including:
- Generative AI
- Large language models (LLMs)
- AI training
- AI inference
- Agentic AI
- Enterprise AI
- Edge AI
- Sovereign AI infrastructure
- Neocloud infrastructure
This lets organizations plan AI infrastructure that scales from the enterprise level up to large-scale AI systems.
NVIDIA Cloud Partner Compliant Architecture
Another key point in this announcement is that Cisco will offer architecture compliant with NVIDIA Cloud Partner (NCP), a framework that helps cloud and AI infrastructure providers build systems properly optimized for NVIDIA AI infrastructure.
Cisco says the newly expanded Secure AI Factory with NVIDIA solution is built to serve the Neocloud and Sovereign Cloud segments.
Neocloud refers to cloud providers focused on specialized workloads, particularly AI and GPU computing, while Sovereign Cloud emphasizes data control, regulatory compliance, and keeping data within a specific country or jurisdiction.
For organizations looking to apply AI to sensitive data, the ability to control data and security is becoming increasingly important.
Cisco Silicon One and NVIDIA Spectrum-X Working Together
On the networking side, Cisco is combining multiple technologies to build its AI architecture. The front-end network runs on switches powered by Cisco Silicon One, while the back-end network uses Cisco switches based on NVIDIA Spectrum-X technology.
Both sides are managed and connected through Cisco Nexus One to create a unified networking architecture. This approach makes it easier for organizations to manage infrastructure from compute through to networking, and improves overall system visibility.
Cisco also notes that it is the only NVIDIA technology partner using its own networking switches and network operating system within this NCP-compliant architecture.
Liquid Cooling Plays a Key Role in AI Infrastructure
One of the biggest challenges for AI-era data centers is power and cooling. Newer GPUs deliver higher compute performance, but they also consume more power and generate more heat. Relying solely on traditional cooling can limit how densely GPUs can be packed into a rack.
Cisco and Supermicro are addressing this with systems that support liquid cooling. Cisco also offers a rack-to-fabric liquid cooling approach that spans everything from the AI server to AI networking.
This approach lets data centers support higher-density AI clusters without being limited purely by cooling constraints.
Three Key Benefits of Cisco Secure AI Factory with NVIDIA
Cisco highlights three major benefits customers gain from this architecture.
1. Lower Investment Risk
AI infrastructure is a high-value investment made up of many components — GPUs, servers, networking, storage, and cooling. Cisco and Supermicro bring deep supply chain and infrastructure expertise, helping reduce risk around sourcing and deployment. Architecture based on NVIDIA Reference Architectures also gives customers greater confidence that the system is properly designed for their AI workloads.
2. Faster Time to Production
After infrastructure is delivered, Cisco helps customers validate and certify their systems through Cisco Validated Infrastructure Services (CVIS), designed to align with NVIDIA Infrastructure Services (NVIS) and confirm the infrastructure was built to the specified architecture. This approach reduces the time organizations need to spend testing and validating systems before moving into production.
3. Simplified Operations
As AI infrastructure scales up, managing it becomes more complex. Cisco addresses this with NVIDIA AI Enterprise software and AgenticOps via Cisco Cloud Control, which help teams manage the system. Organizations can view the relationship between job health and resources such as compute, NICs, optics, and network performance in one place, helping IT teams troubleshoot AI workload issues from compute all the way through the network.
AI Infrastructure Is Shifting From Servers to Factories
Cisco's announcement reflects a significant shift in the AI infrastructure market. In the past, organizations might have thought of AI infrastructure simply as buying GPU servers or AI servers.
Today, AI workloads are larger and more complex, requiring infrastructure to be viewed as a holistic system. The AI factory concept has emerged from this shift, framing AI infrastructure as a factory that uses compute, data, and networking to produce intelligence that can be turned into business value.
As a result, an AI factory's efficiency doesn't depend on GPU count alone — it depends on how well the entire infrastructure works together.
Cisco Targets the Enterprise, Neocloud, and Sovereign Cloud Markets
This expansion of Secure AI Factory with NVIDIA isn't limited to enterprise organizations — it also targets the Neocloud and Sovereign Cloud markets, each with distinct AI infrastructure needs.
| Market Segment | Primary Requirement |
|---|---|
| Enterprise | Systems with strong security, reliability, and ease of management |
| Neocloud | Infrastructure that supports large GPU workloads and scales quickly |
| Sovereign Cloud | Data residency, security, and data control |
By bringing compute, networking, cooling, and security together into a full-stack architecture, Cisco is positioned to serve a wider range of customer needs.
When Will the Solution Be Available?
Cisco says it will begin offering Supermicro Compute Solutions as part of Secure AI Factory with NVIDIA starting:
October 2026Availability date for Supermicro Compute Solutions
This portfolio expansion marks another significant step for Cisco in the AI infrastructure market, coming at a time when organizations worldwide are increasing their investment in AI and data centers.
As AI models continue to grow, demand for compute, networking, and cooling will grow along with them. Infrastructure that can scale at the rack and cluster level is likely to become a core component of the modern data center.
Summary
The expansion of Cisco Secure AI Factory with NVIDIA through the partnership with Supermicro reflects a broader shift in AI infrastructure — away from thinking purely about GPUs or servers, and toward building full-stack, rack-scale systems.
Bringing together AI compute, high-speed networking, liquid cooling, security, and infrastructure management lets organizations build AI infrastructure that supports everything from enterprise AI to trillion-parameter model training. At the same time, support for NVIDIA Cloud Partner-compliant architecture helps serve the growing Neocloud and Sovereign Cloud markets.
For organizations planning AI infrastructure investments, the key question isn't just "which GPU should we use," but whether compute, networking, cooling, security, and management can all work together as a single system. In the era of rack-scale AI, performance no longer depends on compute alone — it depends on how well the entire infrastructure works together.
