Cisco Secure AI Factory with NVIDIA is a full-stack architecture for building and operating enterprise AI infrastructure. It combines Cisco networking, security and management with NVIDIA accelerated computing, Ethernet technology and AI software, plus certified storage partners. Its scope has expanded from data centers to edge sites, and an August 2026 rack-scale expansion adds Supermicro GPU systems. Cisco says those systems are expected to be offered in October 2026 through authorized channel partners.
What is Cisco Secure AI Factory with NVIDIA?
It is an architecture rather than a single server or software product. Cisco and NVIDIA describe it as an integrated way to assemble the infrastructure for production AI workloads, from GPU compute and networking to storage, operations and security. Buyers can use reference designs built around Cisco UCS systems and Cisco networking, or consider the newer dense-GPU systems developed with Supermicro.
The initial announcement, made by Cisco and NVIDIA on March 18, 2025, centered on Cisco Nexus Hyperfabric AI or Nexus 9000 networking, Cisco UCS servers using NVIDIA HGX or MGX platforms, NVIDIA Spectrum-X Ethernet, NVIDIA AI Enterprise software and storage from certified partners. Cisco said at the time that the solutions were expected to be purchasable before the end of 2025, although many individual components were already available.
The architecture has since broadened in two directions: smaller deployments closer to where data is created, and high-density rack-scale systems for large training and inference workloads.
#1 Best Overall
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
What hardware and software does it include?
| Layer | Components named by Cisco | Role in the architecture |
|---|---|---|
| Compute | Cisco UCS AI servers based on NVIDIA HGX and MGX; later Supermicro GPU servers and NVIDIA NVL72 and HGX Rubin NVL8 platforms | Runs AI training and inference workloads. |
| Networking | Cisco Nexus Hyperfabric AI and Nexus 9000; Cisco Silicon One front-end fabrics; NVIDIA Spectrum-X Ethernet back-end fabrics | Connects compute, storage and external networks. The rack-scale design assigns different network roles to Cisco and NVIDIA silicon. |
| Storage | Certified partners Pure Storage, Hitachi Vantara, NetApp and VAST Data | Provides storage options within the validated architecture; the announcement does not specify a single required storage product. |
| AI software | NVIDIA AI Enterprise | Supports production AI workloads. |
| Management and observability | Cisco Cloud Control, Nexus One and Intersight in the expanded design | Provides management and observability across infrastructure. Cisco describes Nexus One as unifying the architecture. |
| Security | Cisco Hybrid Mesh Firewall, Hypershield and AI Defense | Addresses network policy, workload protection, and AI model and application risks. |
| Validation | Cisco Validated Infrastructure Services (CVIS) | Validates full-stack designs against the reference architecture and aligns with NVIDIA Infrastructure Services methodology. |
How does Cisco secure an NVIDIA AI factory?
The security approach uses different controls for different parts of the stack rather than treating AI security as a single firewall feature.
- Hybrid Mesh Firewall provides unified policy management.
- Hypershield provides workload segmentation and runtime enforcement.
- AI Defense focuses on models and AI applications, including risks such as prompt injection, data privacy exposure and unsafe behavior. Cisco says it can validate and protect models and applications across development and runtime.
In its 2025 announcement, Cisco characterized security as embedded from the application through workloads to infrastructure. In March 2026, Cisco also announced AI Defense controls for NVIDIA’s OpenShell agent platform. These are vendor-described capabilities; the announcements do not establish an independent security assessment of a complete deployment.
Rank #2
- GPU-Modell: Gefoce RTX 3080
- Memory Type: GDDR6X Memory Capacity: 20GB Memory Bus Width: 320bit Output Interfaces: 3*DP + HDMI Core Clock: 1710MHz Memory Clock: 19Gbps Power Interface: 8+8pin Recommended Power Supply: 850W or higher
What changes at the edge?
In its March 16, 2026 expansion, Cisco extended the architecture beyond central data centers to local sites such as hospitals, warehouses and moving vehicles. The release added support for NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs across Cisco UCS and Unified Edge, and introduced a Cisco AI Grid reference design for service providers.
The practical distinction is deployment location: edge infrastructure can place AI compute nearer to data and the decisions that depend on it, while the data-center and rack-scale designs address centralized infrastructure. The announcement establishes support and reference designs, but does not give a common performance, capacity or cost comparison between edge and central deployments.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsRank #3
- No Processor Installed; Supports 2x AMD EPYC 9004 Series Processors
- No Memory Installed; Supports 24x DDR5 4400/4800 Regsitered Memory Modules
- 8x 3.5" Trays; (Bring Your Own SATA/NVMe Drives)
- 4x H200 NVL Tensor Core 141GB HBM3e PCI Express 5.0 x16 GPU Accelerator Card
- In Original Packaging; Includes Rails and ASUS GPU Cables
What is the difference between Cisco Silicon One and NVIDIA Spectrum-X?
In the announced rack-scale design, they serve different parts of the network. Cisco Silicon One is used for front-end switching, while NVIDIA Spectrum-X silicon is used for the back-end switching fabric. Cisco says Nexus One provides unified operations across the design.
Cisco also cited two N9100 configurations in its 2026 announcement: a new model using NVIDIA Spectrum-6 Ethernet silicon with stated throughput of 102.4 Tbps, and a generally available 800G model using NVIDIA Spectrum-4 Ethernet silicon. These are Cisco-stated product figures, not an independent comparison of the two fabrics. The announcement does not provide enough detail to treat the models as interchangeable or to infer a system-wide workload performance result from the switch specifications alone.
Rank #4
- 【Brilliant AI Performance for production】 on-device processing with up to 100 TOPS AI performance with low power and low latency, Due to the high thermal demands of Super mode, only the J30 Series supports upgrading to Super mode via the JetPack 6.2 update
- 【Hand-size edge AI device】 compact size at 130mm x120mm x 58.5mm, includes NVIDIA Jetson Orin NX 16GB production module, a cooling fan with a heatsink, enclosure, and a power adapter. Support desktop, wall mount, fit in anywhere
- 【Expandable with rich I/Os】4x USB 3.2, HDMI 2.1, 2xCSI, 1xRJ45 for GbE, M.2 Key E, M.2 Key M, CAN, and GPIO
- 【Accelerate solution to market】pre-installed Jetpack with NVIDIA JetPack 5.1 on the included 128GB NVMe SSD, Linux OS BSP, 128GB SSD, support Jetson software and leading AI frameworks and software platforms
- 【Comprehensive certificates】FCC, CE, RoHS, UKCA
What does the Supermicro rack-scale expansion add?
Announced on August 25, 2026, the Cisco-Supermicro partnership adds high-density compute systems with liquid- and air-cooled options. The target customers include enterprises, neoclouds and sovereign clouds. The named platforms include NVIDIA NVL72 and HGX Rubin NVL8 systems, intended for large-scale GPU deployments.
Cisco says liquid cooling becomes a system-level requirement for modern rack-scale systems such as NVIDIA NVL72 at power levels above 200 kW per rack. That is Cisco’s characterization of the design requirement, not a universal threshold for every AI rack. Cooling design, facility power and rack capacity therefore belong in procurement planning, not just server selection.
Free tools Windows power users keep installed
One-click scans. No signup required.
Best Value
- Supercomputer performance directly to your desk in a compact, energy-efficient design, enabling enterprise-scale AI and high-performance computing right where you need it.
- The power of Grace Blackwell architecture, delivering up to 1 petaFLOP of AI performance for local model fine-tuning, inference, and analytics, accelerating your time-to-solution.
- Designed from the ground up to build and run AI, delivering seamless integration of the full NVIDIA AI software stack —so you can develop locally and deploy anywhere.
- NVIDIA DGX Spark gives you the freedom to experiment, prototype, and innovate faster by augmenting laptop, desktop, cloud, or data center resources. With more power to learn, prototype, test, and innovate, NVIDIA DGX Spark delivers exceptional ROI for increased productivity.
- Use NVIDIA DGX Spark to unlock new ideas and experiment with large models (up to 200 billion parameters at FP4) directly on your desktop with 128GB of unified memory. Empower rapid testing, validation, and iteration—driving innovation in a secure, high-performance setting.
When can enterprises buy the rack-scale systems?
Cisco’s August 25, 2026 announcement and FAQ state that the rack-scale architecture and Supermicro systems can be ordered from authorized channel partners in October 2026; Cisco says Supermicro compute solutions begin offering in October. Because that is a stated availability window rather than confirmation of stock at every partner, buyers should confirm configuration, orderability and delivery timing with an authorized channel partner.
Availability should also be distinguished from the earlier architecture’s components: Cisco’s 2025 announcement said many of those components were already available, while full solutions were expected before the end of that year. That earlier timing does not establish current availability for every combination or for the later rack-scale systems.
How should buyers evaluate the architecture?
Start with the workload and deployment shape, then validate the whole system rather than comparing GPU specifications in isolation.
- Choose a deployment path. Decide whether a Cisco reference architecture is the right starting point or whether a dense Supermicro rack-scale system better fits the workload. Ask the supplier which components are included and which must be selected separately.
- Map the network roles. Confirm where the front-end Cisco Silicon One fabric and NVIDIA Spectrum-X back-end fabric fit in the proposed design, and what Nexus One management covers.
- Check facility constraints. For high-density racks, confirm power delivery and cooling capacity with the data-center or facility team, including whether the proposed configuration requires liquid cooling.
- Confirm storage and software choices. Select among the named certified storage partners and clarify the software, management and observability components included in the deployment.
- Scope validation and operations. Ask what CVIS validation covers, what evidence is delivered, and how ongoing operations use Cloud Control, Nexus One or Intersight.
- Verify procurement details. For the October 2026 rack-scale offer, confirm the exact system, channel partner, order status and expected delivery rather than assuming every configuration is immediately available.
Cisco says its CVIS automation can reduce validation timelines from months to weeks. Treat that as a vendor claim about validation, not a guaranteed deployment schedule or a measured reduction in total project time.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What is not established by the announcements?
The cited announcements do not provide an independent benchmark of the complete Secure AI Factory, a total-cost-of-ownership study or independently measured return on investment. Cisco’s statements about simpler deployment, efficiency, validation timelines and business value should be read as vendor claims, not as independently verified outcomes. Actual results will depend on the chosen hardware, workload, facility, storage, software and operating model.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
