The enterprise AI race is no longer just about GPUs or models — it’s about the data that fuels them. Enterprises face two pressing challenges: making data AI-ready and keeping it resilient in an era of escalating cyber threats. Pure Storage is addressing both, with a slate of announcements that tie together performance, platform intelligence, and ecosystem partnerships.

Prakash Darji, GM of the Digital Experience Business Unit at Pure Storage, told me the company is enhancing its Enterprise Data Cloud to tackle fragmentation. Previously, it was siloed and complex, with automation disconnected from observability, he explained.” By unifying the control plane and adopting the Model Context Protocol (MCP), Pure is connecting storage, cyber detection, and AI intelligence across ecosystems.

Fixing the KV Cache Bottleneck with NVIDIA

One highlight is the Pure Key Value Accelerator built with NVIDIA Dynamo. Large language models often struggle with KV cache inefficiency, but this partnership delivers dramatic gains. “We’ve seen 20x improvement on cache performance,” Darji said, noting it also lowers GPU costs by eliminating recompute cycles. With AI efficiency now a top enterprise priority, Pure sees this as a major leap forward.

From Reactive to Predictive with AI Copilot

Pure also launched the Pure1 AI Copilot for Portworx, designed to transform Kubernetes operations. By training on Portworx data and exposing insights via MCP, the Copilot shifts troubleshooting from reactive to predictive. “Interoperability costs go down drastically,” Darji explained, since MCP provides a looser, more resilient interface than REST APIs.

Simplifying Hybrid and Multi-Cloud Data

For enterprises balancing VMware workloads with modern Kubernetes apps, the company announced Pure Storage Cloud Azure Native and tighter integration between Portworx and Pure Fusion. This gives customers a single control plane to manage both traditional and cloud-native data. Darji framed it as optionality: some customers will stay on VMware, others will move to Kubernetes. Either way, Pure aims to provide consistent control and resiliency.

Uniting Cyber Detection and Recovery

Cyber resilience remains central. Pure is working with partners like CrowdStrike and Veeam to build layered, connected detection and response. “Every vendor has detection, so customers are left with a fragmented landscape,” Darji said. By associating signals across multiple detectors and tying them to recovery zones, Pure enables faster, more dynamic responses to attacks.

The Long-Term Vision: Truly AI-Ready Data

Darji defined AI-ready data as enabling both training and inference across workloads, supported by strong ingest, observability, and data quality pipelines. Partners like NVIDIA are key to this roadmap, but so are emerging standards like OpenTelemetry.

Pure’s Enterprise Data Cloud is designed to unify this journey — from accelerating LLMs, to securing hybrid workloads, to making data reliable and efficient across the AI lifecycle. As enterprises scale AI adoption, this blend of performance and protection may prove decisive.