Pure Storage & Cisco launch unified platform for scaling AI
Pure Storage and Cisco have announced a partnership to provide enterprises with a unified platform designed to help scale generative AI initiatives from pilot stages to full production deployments.
The collaboration centres on a newly validated FlashStack solution incorporating Cisco and Pure Storage technologies, aimed at supporting demanding AI workloads such as retrieval-augmented generation (RAG), agentic AI, and semantic search. As part of the partnership, Pure Storage has also introduced Data Stream, leveraging the NVIDIA AI Data Platform to automate and accelerate the conversion of unstructured, raw data into AI-ready intelligence.
According to the companies, the FlashStack Cisco Validated Design (CVD) creates an integrated, production-grade environment where compute, storage, networking, and software are unified. This eliminates complexities that often hinder enterprise AI projects from advancing beyond initial pilot phases.
Maciej Kranz, General Manager, Enterprise at Pure Storage, said:
"Too often, enterprises focus on GPUs and compute power, but without reliable data, the model never reaches its potential. Our collaboration with Cisco and NVIDIA removes those data barriers, giving customers the performance, simplicity, and efficiency they need to operationalise AI."
Many organisations have spent significant time developing proof-of-concepts and testing large language models but encounter delays due to data readiness issues. The underlying concern is that the model often becomes production-ready before the supporting data pipeline is adequately prepared.
Jeremy Foster, SVP & GM, Cisco Compute, explained the new approach:
"With this new FlashStack CVD, we're not just validating hardware, we're orchestrating all the elements of RAG into an AI-ready infrastructure, removing complexity and reducing risk so customers can focus on turning data into insights that drive strategic outcomes."
AI adoption at the enterprise level depends on reliable, streamlined data pipelines. Fragmented architectures and operational challenges can slow the transition from testing to production environments, limiting the impact of AI investments. The FlashStack solution addresses these barriers by making both structured and unstructured data available to AI workflows in a manner that is performance-orientated and secure.
Focus on data architecture
The solution uses Pure Storage's Enterprise Data Cloud architecture, providing high-performance access for AI workloads and ensuring concurrency and energy efficiency with FlashBlade//S. For AI environments built on Kubernetes, Portworx by Pure Storage enables data mobility, persistence, and protection according to the requirements of modern deployments.
This approach is intended to allow AI teams to devote resources toward development and innovation instead of infrastructure management. The companies describe the FlashStack design as an end-to-end blueprint, integrating elements including Pure Storage FlashBlade//S, Cisco UCS C845a servers equipped with NVIDIA RTX PRO 6000 GPUs, and the NVIDIA AI Enterprise software suite.
Integrating and scaling AI
The partnership aims to deliver a platform that converts raw data and electricity inputs into the usable outputs-or tokens-required by generative AI systems. The groundwork relies on a customer base established with more than 5,000 existing FlashStack installations, now enhanced with new components to address scalability for AI.
The flash-backed, GPU-accelerated infrastructure ensures performance and reliability for data-intensive workloads, supporting enterprise and regulated environments with governance controls. The coordinated inclusion of Cisco's Nexus 9000 series networking provides a low-latency, congestion-aware data fabric, enabling efficient data sharing between distributed GPUs, storage, and compute layers.
Through Cisco's NX-OS and Nexus Dashboard, IT operations teams obtain a unified interface to oversee networking, maintain security standards, and address the specific demands of AI-native tasks. These design choices are presented as solutions for scaling AI workloads while minimising operational complexity and risks related to data fragmentation.
Making the solution available
The FlashStack Cisco Validated Design will be made available through a network of established channel partners, with customers able to access early programme engagement and receive solution briefings from Pure Storage, Cisco, or NVIDIA representatives. The companies emphasise the combined abilities of FlashBlade//S, Cisco UCS C845a servers, and NVIDIA AI Enterprise software in powering a range of enterprise AI use cases.
The partnership highlights efforts to address challenges that arise when enterprises seek to operationalise AI at scale, focusing not only on computation and modelling but on the foundational data infrastructure required for sustained, production-ready deployments.
