Cisco Nexus Hyperfabric’s Full Stack AI Infrastructure Data Sheet (Formerly Nexus Hyperfabric AI)
Available Languages
Bias-Free Language
The documentation set for this product strives to use bias-free language. For the purposes of this documentation set, bias-free is defined as language that does not imply discrimination based on age, disability, gender, racial identity, ethnic identity, sexual orientation, socioeconomic status, and intersectionality. Exceptions may be present in the documentation due to language that is hardcoded in the user interfaces of the product software, language used based on RFP documentation, or language that is used by a referenced third-party product. Learn more about how Cisco is using Inclusive Language.
To build on the robust and scalable AI infrastructure provided by Cisco Secure AI Factory with NVIDIA, the Cisco Nexus Hyperfabric™’s full stack AI infrastructure option offers a vertically integrated approach that unifies on-premises network, compute, GPUs, and storage components. This integration streamlines deployment and management, delivering optimized performance and agility tailored to enterprise AI workloads, enabling organizations to accelerate innovation with a cohesive, end-to-end AI platform.
Cisco Nexus Hyperfabric’s full stack AI infrastructure option is a cloud-managed, on-premises solution designed to accelerate AI lifecycle management from design to scale. It delivers high throughput, low latency, and scalable AI infrastructure optimized for AI workloads.
The deployment model for the Cisco Nexus Hyperfabric’s full stack AI infrastructure option is turnkey, vertically integrated, and NVIDIA ERA–compliant, consisting of networking, servers, GPUs, and storage. It is cloud-managed to simplify the entire lifecycle, from design to operation, for demanding AI enterprise workloads. This option optimizes for time to first token for dedicated AI clusters from concept to production.
Cisco Nexus Hyperfabric’s full stack AI infrastructure option
Cisco Nexus Hyperfabric’s full stack AI infrastructure option
Cisco Nexus Hyperfabric’s full stack AI infrastructure option is a turnkey, on-premises, AI-cluster option managed by a cloud-hosted controller, designed to simplify and accelerate AI deployments for enterprises and neocloud. It simplifies AI infrastructure deployment with a comprehensive, integrated, cloud-managed option compliant with NVIDIA Enterprise Reference Architecture (ERA). This cloud-managed vertical stack option consists of purpose-built hardware, software, a Cisco Nexus Hyperfabric cloud controller, day-2 operations, automation, and Cisco support that eliminates complexity. IT, application, and/or DevOps teams manage the full lifecycle of designing, ordering, deploying, validating, monitoring, and scaling fabrics without requiring deep networking or operational expertise.
It integrates Cisco’s high-performance networking hardware, including Cisco® Silicon One® switches and Cisco UCS® C885A M8 Rack Servers with NVIDIA HGX, with a cloud-managed operational model. This option offers preconfigured templates for AI clusters, enabling rapid design, ordering, deployment, and scaling of AI infrastructure with zero-touch provisioning and continuous monitoring. It is a unified stack that includes networking, compute, GPUs, AI software, and storage components, all managed through a cloud controller to provide full visibility and control.
Cisco Nexus Hyperfabric’s full stack AI infrastructure option is built to deliver a simplified, scalable, and secure AI infrastructure that aligns with NVIDIA’s Enterprise Reference Architecture, facilitating efficient AI workloads and innovation across industries.
Key components:
● Cloud-hosted Cisco Nexus Hyperfabric controller that manages network fabric configuration and lifecycle (design, ordering, deployment, management, and scaling of AI compute, GPUs, storage, and network infrastructure)
● AI cluster templates for small to extra-large deployments
● Cisco Nexus® 6000 Series Switches with Silicon One technology, including HF6100 variants with Cisco N9000 Series planned for the future
● Cisco UCS C885A M8 Rack servers equipped with NVIDIA HGX H200 GPUs and BlueField-3 DPUs
● Optional VAST data-storage platform integrated with Cisco UCS servers
● Cisco optics and cables
● Optional AI software stack including NVIDIA AI Enterprise (NVAIE)
● Onsite installation services by Cisco Customer Experience (CX)
Performance and features:
● NVIDIA ERA–compliant for AI workloads, ensuring validated performance and scalability
● High-throughput fabric with 800 GB/s bandwidth and ultra-low latency
● Support for RDMA, NVMe Over Fabrics (NVMeoF), and advanced telemetry
● Cloud-managed lifecycle with zero-touch provisioning and continuous software and firmware updates
● End-to-end assertion-based monitoring from network fabric to compute NICs and storage
● Preconfigured leaf-and-spine pods for plug-and-play deployment
Scalability:
● Modular design supporting small, medium, and large AI clusters
● Pre-configured designs for the entire infrastructure component suite – frontend, backend, rails, storage, and management
● Independent scaling of network, compute, GPUs, and storage resources
● Fully integrated turnkey option with VAST storage or support for bring your own (BYO) storage
Use cases and benefits:
● Large-scale enterprise AI model training, fine tuning, and inference
● Dedicated AI/ML clusters with optimized efficiency and low latency
● Faster time to value with replicable deployment models
● Validated turnkey AI stack through Cisco, NVIDIA, and storage vendor partnership, ensuring optimized hardware and software integration
How Cisco Nexus Hyperfabric is built
Architecture and cloud management
Cisco Nexus Hyperfabric offers a cloud-managed full-stack AI infrastructure that simplifies the entire lifecycle management process, including design, ordering, deployment, management, and scaling of AI compute, GPUs, storage, and network infrastructure. This option features a cloud controller that enables comprehensive design, planning, ordering, operation, upgrade, and monitoring of the AI fabric down to the NIC level, with continuous monitoring of loss and latency. Switches, GPU clusters, and storage servers are bootstrapped and fully visible and controlled from the cloud, ensuring seamless management.
The infrastructure includes fit-for-purpose fabrics that are preconfigured specifically for AI workloads, providing a low-latency, lossless backend network alongside high-performance frontend and storage fabrics. Software and monitoring capabilities are integrated through a Nexus Hyperfabric container that delivers server network telemetry, unified administration of frontend, backend, storage, and management fabrics, as well as assertion-based fabric monitoring and real-time health checks.
Scalability and flexibility are key features of this architecture, supporting independent scaling of network, compute, and storage resources. Its modular design allows customers to bring their own storage or opt for fully integrated turnkey options, enabling tailored deployments that meet diverse AI workload requirements. This comprehensive approach facilitates fast, reliable AI cluster deployment with a replicable model that eases AI cluster management and future scalability.
Deployment experience
The guided installation process for Cisco Nexus Hyperfabric’s full stack AI infrastructure option simplifies AI cluster deployments through real-time verification, cloud-managed lifecycle, continuous software updates, and a replicable deployment model. On day 0, users access the centralized cloud-managed controller portal at hyperfabric.cisco.com to plan and design the infrastructure using the designer tool, finalize the bill of materials including optics and cables, and begin installation planning based on the cabling plan. On day 1, upon equipment arrival, Cisco 6000 Series Switches (Cisco N9000 Series Switches are coming soon), Cisco UCS servers with GPUs, and optional Cisco UCS servers with third-party storage software are registered and connected to the cloud controller. The mobile-friendly portal provides step-by-step deployment guidance and real-time topology validation to ensure that the physical fabric matches the design. From day 2 onward, lifecycle management is handled remotely through the cloud controller, enabling monitoring, upgrades, and collaboration. This process automates provisioning, guides on-site tasks, and allows remote management to be accessible even to non-expert staff, thereby streamlining deployment and operational management of AI infrastructure.
A subscription entitlement is needed for every Cisco 6000 Series Switch that is deployed and used. Subscription entitlements may be initially purchased for three, five, or seven years and may be renewed. The subscription-feature entitlement tiers are based on fabric use cases. The “Premier” package applies to Cisco Nexus Hyperfabric’s full stack AI infrastructure option. All the switches in a fabric must use the same entitlement tier; however, an organization may concurrently manage multiple fabrics that use different entitlement tiers.
Table 1. Cisco Nexus Hyperfabric Essentials and Premier feature tiers
| Feature |
Premier (Cisco Nexus Hyperfabric’s |
| Cisco support 8x5xNBD |
Yes |
| Cloud controller |
Yes |
| Designer (no purchase required) |
Yes |
| Cloud-driven software upgrades |
Yes |
| BOM generation |
Yes |
| On-site deployment assistance |
Yes |
| Plug-and-play deployment |
Yes |
| Spine-and-leaf topologies |
Yes |
| Mesh (spineless) topologies |
Yes |
| EVPN/VXLAN underlay (opaque) |
Yes |
| Static and BGP routing |
Yes |
| MLAG |
Yes |
| Supports RDMA over Converged Ethernet v2 |
Yes |
| Real-time cloud-accessed telemetry |
Yes |
| IPv4 and IPv6 |
Yes |
| Assertion-based monitoring |
Yes |
| Survivable data and local management plane |
Yes |
| Hardware-based attestation and security |
Yes |
| API for headless provisioning and monitoring |
Yes |
| AI use-case support |
Yes |
| Aligned to NVIDIA Enterprise Reference Architecture |
Yes |
| NVIDIA Adaptive Routing on backend switches |
Yes |
| Deploy AI-validated blueprints built into workflow |
Yes |
| Automatically provisions lossless backend AI and storage networks |
Yes |
| Internal performance monitoring between switches and servers |
Yes |
| Options for AI servers |
Bundled Cisco UCS |
Information about Cisco’s environmental, social, and governance (ESG) initiatives and performance is provided in Cisco’s CSR and sustainability reporting.
Table 2. Cisco environmental sustainability information
| Sustainability topic |
Reference |
| General: Information on product-material-content laws and regulations |
|
| Information on electronic waste laws and regulations, including products, batteries, and packaging |
|
| Information on product takeback and reuse program |
|
| Sustainability inquiries |
Contact: csr_inquiries@cisco.com |
| Material: Product packaging weight and materials |
Contact: environment@cisco.com |
Cisco and partner services offer a wide range of services to help accelerate your success in connecting Cisco 6000 Series Switches and Cisco N9000 Series Switches to the Cisco Nexus Hyperfabric’s full stack AI infrastructure option cloud controller. Our innovative service offerings are delivered through a unique combination of people, processes, tools, and partners and are focused on helping you increase operational efficiency and improve your network control. Cisco Nexus Hyperfabric’s full stack AI infrastructure option provides proactive support with the Cisco SMARTnet® service to help you resolve mission-critical problems with direct access at any time to Cisco network experts and award-winning resources. Spanning the entire network lifecycle, our service offerings help increase investment protection, optimize network operations, support migration operations, and strengthen your IT expertise.
For more information, please visit www.cisco.com/go/services.
Flexible payment solutions to help you achieve your objectives
Cisco Capital® makes it easier to get the right technology to achieve your objectives, enable business transformation and help you stay competitive. We can help you reduce the total cost of ownership, conserve capital, and accelerate growth. In more than 100 countries, our flexible payment solutions can help you acquire hardware, software, services, and complementary third-party equipment in easy, predictable payments. Learn more.
Try before you buy
Anyone with a Cisco ID or CCO ID may log into the Cisco Nexus Hyperfabric’s full stack AI infrastructure option cloud controller at hyperfabric.cisco.com to request an organization identifier; they can then begin building network fabric blueprints for free.
For more information, visit: cisco.com/go/nexus-hyperfabric-ai.