Pure Storage Unveils Comprehensive AI-Ready Data Platform Enhancements

Pure Storage claims 20x faster AI inference - but the real story is who controls AI-ready data

Pure Storage unveils an extensive set of product updates at its September Accelerate event in New York this week, headlined by claims of 20x faster Artificial Intelligence (AI) inference through its new Key Value Accelerator. The announcements build on the company's Enterprise Data Cloud strategy launched in June, but reinforce something we explored at its main user conference in Vegas earlier this year: Pure is positioning itself as the layer that decides which enterprise data is ready for AI consumption. For CIOs navigating the chaos of AI deployment, this creates an interesting proposition that is worth unpacking further. Following my conversations with Pure executives at the June event, where the company outlined its ambition to move from storage vendor to "data trust broker," these product announcements provide further evidence of how that strategy is making its way into new product features. The headline announcement this week is Pure's Key Value Accelerator integration with NVIDIA's Dynamo platform. Patrick Smith, Field Chief Technology Officer (CTO) for Europe, Middle East and Africa (EMEA) at Pure Storage, explains the performance gains: As much as 20 times faster inference, which when everybody talks about inference in the modern world of AI, is a huge step forward. That's a compelling pitch for enterprises where Graphics Processing Unit (GPU) costs are spiraling and inference speed determines whether AI applications are viable (particularly in a world where Return on Investment (ROI) from AI projects is questionable). But the more interesting development is how Pure achieves this speed - through intelligent caching that requires the platform to make decisions about which data should be accelerated and how it should be classified. This builds on Pure's broader data intelligence capability, which Smith describes as: The ability for us to be able to understand what workload a customer runs on our platform in their data center. When Pure's system suggests that certain Structured Query Language (SQL) workloads are "tier-two workloads, and therefore right for you to move to Kubevirt," it's making architectural decisions that traditionally belonged to database administrators and enterprise architects. Pure's timing with automated VMware migration capabilities is shrewd, given ongoing concerns about Broadcom's pricing model. But the three-phase process Smith outlined - discover, execute, validate - reveals how comprehensively Pure is embedding decision-making into its platform. The discovery phase uses the enhanced AI Copilot with Model Context Protocol (MCP) capabilities. Smith explains: The ability to call out from AI Copilot to MCP server running on VMware or vSphere, to understand the configuration of the existing footprint on VMware, and then call out to MCP on OpenShift to understand what my cluster configurations are. This is significant because Pure isn't just automating migration - it's determining workload criticality and making placement decisions. The execution phase adds new workflow templates and Pure-built connectors for Fusion and Pure1, allowing "the workflow to reach into Fusion, reach into Pure1, understand the environment, and as part of the execution, drive the change." The validation phase completes the loop, with Portworx integration into AI Copilot enabling the system to assess the outcomes. Smith comments: Are the Kubernetes clusters still working as expected? Have I introduced performance bottlenecks? Have I introduced any potential risks? While Pure emphasizes its platform strategy, the company hasn't abandoned its hardware roots. The new FlashArray//XL 190 delivers what Smith describes as "bigger, better, and faster" - specifically 25 percent greater capacity than the XL 170, 50 percent lower latency, and 100 percent performance increase. The FlashArray//ST, which Pure has been shipping since early this year, now claims 18 million Input/Output Operations Per Second (IOPS), up from the 10 million announced in Las Vegas. Meanwhile, the refreshed FlashArray//X R5 promises 30 percent performance improvement at the three-year Evergreen Forever renewal, while the FlashArray//C R5 delivers 40 percent improvement in capacity density. Smith emphasizes the value of Evergreen to buyers: Customers buy a FlashArray//X, and they buy it in the knowledge that when they come to their three-year renewal of Evergreen Forever, they're going to get a 30 percent bump in performance improvement. They can put more workloads on that same array. It doesn't cost them anymore. This hardware cadence - updating the C, X, XL, and ST all in 2025 - suggests Pure sees continued differentiation in raw performance even as it pursues higher-level platform capabilities. Pure also says that it is the first storage vendor with an Azure Native service. The company is rebranding Cloud Block Store to Pure Storage Cloud Dedicated and launching Pure Storage Cloud for Azure Native with Azure VMware Solution. Smith acknowledges the complexity of the changing name but emphasizes the significance: It's really a volume-as-a-service capability. It allows customers to reduce their cost in the public cloud, increase their data protection capabilities, and it's a fully managed solution. Unlike Cloud Block Store where customers "still have to manage an array in the public cloud, albeit a software-defined array," the Azure Native service means "they don't even have to manage the array. They just consume volumes that come from the Pure Storage Cloud environment." This extends Pure's management layer into public cloud environments, with Microsoft's Aung Oo, Vice President of Azure Storage, noting: Together, we're making it easier for customers to migrate data-heavy VMware workloads into Azure and quickly leverage the latest cloud innovations in AI and analytics. Pure also makes a host of security announcements, which outline how the company is building governance capabilities directly into the storage layer. The new Bring Your Own Key encryption, which Smith calls "really interesting, and I think important," enables organizations to "encrypt your data with your own key in a secure multi-tenant environment." More telling is how Pure is aggregating security signals from partners. Smith notes: Rubrik is specifically integrating so that they can tag snapshots with an indicator that that snapshot is compromised or not based on the Rubrik insight into the dataset. Similarly, the CrowdStrike Falcon Next-Gen Security Information and Event Management (SIEM) integration promises "real-time visibility and automated response - instantly updating policies, isolating systems, and securing replication." The announcement of Pure Protect Recovery Zones - what Smith calls "isolated recovery environments, or effectively clean rooms" - takes this further. These aren't just backup locations; they're validation environments where Pure's platform determines whether recovered data is trustworthy enough to return to production. Smith also highlights new capabilities including "zero-trust environment with enterprise-grade identity and access management" and a second-generation AI Copilot security assessment that can identify "what vulnerabilities I've got in my environment, what Common Vulnerabilities and Exposures (CVEs) I need to address." Pure's decision to open its intelligent control plane to partners is also worth pointing out. Smith notes that "both Rubrik and Veeam are making use of the intelligent control plane to add functionality into a Pure environment." From the conversation with Smith, it's becoming clear that this isn't just integration - it's about Pure becoming the enforcement point for multiple vendors' assessments. When Superna provides "automated threat detection for file" or when CrowdStrike's Real-Time Threat Graph feeds into Pure's platform, these signals converge at the storage layer where Pure makes enforcement decisions. The new Veeam partnership for "Cyber Resilience delivered as a Service" particularly stands out, with Veeam's John Jester noting it replaces "fragmented, manual processes with a unified, policy-driven automation approach." For enterprise buyers, Pure's announcements signal where the vendor is heading. The company is addressing real problems - VMware migration headaches, AI inference costs, ransomware recovery challenges. But buyers need to realize that these announcements represent a fundamental shift in how data decisions get made. When Pure's AI Copilot determines workload tiers, when its Key Value Accelerator decides caching priorities, when its Recovery Zones validate data trustworthiness - these are governance decisions being made at the infrastructure layer rather than by human administrators or higher-level applications. The FlashBlade//EXA for AI training environments and the enhanced Portworx integration for containerized workloads show Pure is serious about supporting modern architectures. But they also extend Pure's decision-making footprint across more of the enterprise infrastructure. These aren't just product features - they're Pure staking a claim to make governance decisions traditionally made higher up the stack or by human administrators. The company is betting that enterprises will trade some autonomy for the simplicity of having these decisions automated at the infrastructure layer. And given data trust, security and governance are going to be key to the success of AI projects, it's a savvy move.

Pure Storage, Inc. Announces Enhancements to the Pure Storage Platform

Pure Storage, Inc. announced enhancements to the Pure Storage Platform that make it easier for organizations to make full use of what AI can offer, streamlining workflows across on-premises and cloud environments. As AI is increasingly key to business, data is the critical asset for deriving insight, automation, and competitive advantage. Pure Storage's Enterprise Data Cloud architecture provides a unified architecture and operating model that delivers true end-to-end data control, automation, and cyber resilience across any environment. And the company are now offering a major expansion of the Enterprise Data Cloud into the public cloud, with Pure Storage Cloud. For IT leaders managing a hybrid cloud storage infrastructure, Pure Storage Cloud unifies a data landscape so data is always exactly where it's needed, powering new innovation and fully accessible in the cloud so AI tools can leverage and train on the data. NEW: Pure Storage Cloud Azure Native (Generally available now.) Through the Azure Native Integrations program, Pure Storage has delivered a first-of-its-kind service built for Azure VMware Solution. As a result, customers can reduce overhead, seamlessly migrate without refactoring, and dec couple storage from compute. Offered as a fully managed Azure native service with enterprise-grade resiliency and efficiency, Pure Storage Cloud is available within the Azure portal now. "Organizations looking to embrace AI and next-gen cloud services with Microsoft must first move data to Microsoft Azure. This extends powerful fleet management capabilities to modern containerized cloud-native applications and Kubevirt based VMs across any environment - on-premises or within a hybrid cloud. Pure1 AI Copilot Expansion Announced as Generally Available at Pure//Accelerate 2025, AI Copilot combines the power of dashboards with the simplicity of conversation to make storage systems more accessible and easier to manage. The company continue to expand the ways users can manage file services using natural language. NEW: Portworx Pure1 AI Copilot. (Generally available now.) With the first AI-powered platform engineering assistant for Portworx customers, users can query Portworx clusters in the same way they interact with FlashArray systems. They can now monitor their Kubevirt and Portworx clusters at scale via instant interaction with an AI agent on the Pure1 Copilot user interface. Pure1 AI Copilot Integration with Model Context Protocol (MCP) Servers (Generally available Fourth Quarter fiscal26.) IT teams can now troubleshoot, provision, or optimize without deep technical expertise. Pure Storage AI Copilot operates as both an MCP server and client, enabling seamless integration with internal systems for hardware performance, subscriptions, and security data, as well as external tools like analytics engines and application monitors. U Updated: Next-Gen FlashArray Portfolio The Pure Storage architecture is uniquely designed so organizations can run enterprise apps and modern AI/ML pipelines on the same, scalable foundation; the FlashArray family supports both latency-sensitive applications and high-bandwidth, high-concurrency workloads. The expansion of next-generation storage products unveiled at Pure//Accelerate 25 continues with FlashArray 190 (generally available Fourth Quarter fiscal26), and FlashArray//X R5 and FlashArray//C R5 (both generally available now). Pure Storage delivers data storage and management at scale--unifying and governing data across every environment. With an Enterprise Data Cloud, block, file, and object are managed and governed by policy; every system and site contributes capacity and performance through a shared virtual layer; and all data is controlled through a single console. Additional Resources. Pure Storage Introduces the Enterprise Data Cloud so Customers can Manage their Data, Not Their Storage.