Objective: unify compute, storage, networking, and software on an enterprise-grade platform that accelerates the leap from pilot projects to production in generative AI initiatives (RAG, agents, semantic search, video analytics, code generation).
Pure Storage (NYSE: PSTG) and Cisco (NASDAQ: CSCO) announced a new Cisco Validated Design (CVD) for FlashStack® that joins the AI PODs of the Cisco Secure AI Factory with NVIDIA. The solution integrates Cisco UCS servers, FlashBlade//S™ storage, Nexus 9000 switch, Portworx® for Kubernetes data management, and NVIDIA AI Enterprise within a referencing architecture to deploy AI at scale with reduced complexity and risk.
“Without reliable data, a model can’t reach its full potential. With Cisco and NVIDIA, we eliminate data barriers to operationalize AI,” stated Maciej Kranz, GM Enterprise, Pure Storage.
“We’re not just validating hardware; we orchestrate complete RAG workflows on an ‘AI-ready’ infrastructure so the customer can turn data into insights,” added Jeremy Foster, SVP & GM, Cisco Compute.
The missing link in enterprise AI: data
Organizations have spent months conducting pilots with LLMs, RAG, and domain-specific applications, but deployment stalls when data architecture is fragmented and operations are complex. The CVD addresses this challenge:
- Ready-to-use structured and unstructured data for AI via FlashBlade//S™ (high performance, concurrency, energy efficiency) within Pure’s Enterprise Data Cloud architecture.
- Persistent, portable, and protected data in Kubernetes environments with Portworx® (snapshots, replicas, cloud-to-cloud mobility).
- Low-latency, highly available access for training and inference pipelines.
Result: AI teams create and iterate; the platform manage and deliver data.
The definitive ‘AI Factory’: key components
- Compute: Cisco UCS C845a with NVIDIA RTX PRO 6000 GPUs to accelerate inference/training and vectorization/embedding workloads.
- Storage: Pure Storage FlashBlade//S™ as a unified data platform for high-performance files/objects.
- Network: Cisco Nexus 9000 with telemetry, adaptive routing, and per-packet load balancing to create a low-latency fabric connecting GPUs, storage, and compute.
- Software: NVIDIA AI Enterprise (production stack to build and operate AI at scale).
- Data in K8s: Portworx® for stateful sets, persistent storage, and cross-cloud data protection.
- Operations: Cisco NX-OS and Nexus Dashboard for unified visibility and control, with enterprise-grade security and multi-tenancy.
Built upon a base of 5,000+ FlashStack customers, the new CVD provides a ‘production-ready’ path that integrates systems seamlessly without the need for disjointed architectures.
Use cases and benefits
Use cases: RAG with vector stores, agents (agentic AI), semantic search, video analytics, code generation, and other data-intensive AI workloads.
Key benefits
- Reduced complexity and risk: validated design by Cisco (CVD) accelerates time-to-value.
- Performance and concurrency: high-volume read/write for AI pipelines with minimal network latency.
- Governance and reliability: consistent, portable, and protected data suitable for regulated industries.
- Scalability: expand in compute, network, and storage as integrated blocks without re-architecting.
Availability and go-to-market
The FlashStack CVD from Cisco and Pure will be generally available in Q1-2026 through partners such as AHEAD, ePlus, SHI International, and World Wide Technology (WWT). From today, customers can join early access programs and participate in briefings with teams from Pure, Cisco, or NVIDIA.
More info: demonstrations and blueprints for FlashBlade//S™, Cisco UCS C845a (NVIDIA GPUs), and NVIDIA AI Enterprise at NVIDIA GTC (October 27-29, Washington D.C.) or at FlashStack.com.
Quick overview for decision-making
- Problem solved: moving from pilots to production by aligning data, compute, and network within a validated block.
- Why now: AI is no longer experimental; the bottleneck is now in trustworthy data pipelines.
- Includes: FlashStack CVD (UCS + Nexus + FlashBlade//S + Portworx + NVIDIA AI Enterprise) with telemetry and unified control.
- Target audience: enterprises with RAG/LLM/Agents requiring performance, governance, and SLA in multi-cloud and regulated environments.
via: pure storage