cd /news/ai-infrastructure/everpure-expands-enterprise-ai-push-… · home › topics › ai-infrastructure › article
[ARTICLE · art-142858] src=sdxcentral.com ↗ pub= topic=ai-infrastructure verified=true sentiment=↑ positive

Everpure expands enterprise AI push with unstructured data, inference updates

Everpure, formerly Pure Storage, launched File Edition for its Data Intelligence platform at its London Accelerate event, mapping an entire data estate via file metadata to flag redundant, obsolete or sensitive information before it reaches an AI system. The vendor also detailed DeepReduce improvements for its FlashBlade lines that field CTO for EMEA Fred Lherault said deliver two-to-one data reduction on top of existing compression, plus PureKVA, a key-value accelerator that shares AI inference context across GPU memory, CPU memory and external storage. Lherault said Everpure estimates its upcoming "intelligent token optimization" reference architecture, built on open-weight on-premises models, can cut AI inference costs by 50 to 75%.

read2 min views1 publishedSep 30, 2026
Everpure expands enterprise AI push with unstructured data, inference updates
Image: Sdxcentral (auto-discovered)

LONDON – Everpure (formerly known as Pure Storage) continued to push out of its storage comfort zone with updates to its Data Intelligence platform.

The vendor this week launched File Edition, which maps an entire data estate using file metadata, like last access or user access patterns, to identify redundant, obsolete, or even potentially sensitive information a business might not want to expose to an AI system. It allows users to ensure data is kept in the right place to better comply with corporate requirements.

The update, which was announced at the vendor's London Accelerate event, leans on learnings from its 1Touch acquisition and builds on the initial Data Intelligence launch in June.

Everpure touted the Data Intelligence update in line with meeting increased sovereignty demands, with users able to apply improved understanding of their data agnostic to where it resides, be it unstructured data in a virtual private cloud or a cloud storage container.

“Think about this as simple as a query like, 'please show me all my European Union customers' data and where it sits, and let me know if it's sitting in a non-EU environment,'" Ashish Gupta, 1Touch CEO turned Everpure GM for data management, explained. "And this model context protocol (MCP) server will give you that specific knowledge right away without writing any code of your own … it's just making the whole environment that much easier.”

Everpure also showcased its “universal data plane," which is essentially a multitude of wider enhancements aimed at spicing support for data across workloads, temperatures, protocols, and use cases.

Among those were improvements to DeepReduce for its FlashBlade lines that use seriality-based compression to find additional opportunities to reduce duplicated data, even unearthing instances where traditional deduplication may typically miss. Fred Lherault, Everpure’s field CTO for the Europe, Middle East, and Africa (EMEA) region, said this provides a two-to-one data reduction on top of existing compression.

There’s also PureKVA, a key-value accelerator that expands AI inference caching across GPU memory, CPU memory, and external storage. This lets multiple nodes share context so they don’t have to recompute the same information, a move that improves inference performance and reduce duplicated work.

“You don't want [multiple GPU nodes] to be redoing the same thing all the time just because someone asks the same question to two different nodes,” Lherault said. “So this allows sharing of context and not only improves the performance of AI inference but also avoids duplication of work.”

Everpure also wants to tackle growing token costs with what execs called “intelligent token optimization,” an upcoming reference architecture that brings together all of the updates on display in London.

“Based on something that we're building for ourselves, we're estimating that we'll be able to save our AI inference costs or reduce them by 50 to 75% by actually using open-weight models on-premises for all those AI inference needs,” Lherault added.

── more in #ai-infrastructure 4 stories · sorted by recency
── more on @everpure 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/everpure-expands-ent…] indexed:0 read:2min 2026-09-30 · —