When Pure Storage rebranded to Everpure in February 2026, shedding a 16-year-old name after its first billion-dollar quarter. The company that built its reputation on flash storage was declaring that storage is no longer the product. Data is.
At Accelerate 2026, Everpure backed that declaration with a set of announcements that carry direct implications for how Indian enterprise CIOs think about their AI readiness. In an exclusive interaction with CIO&Leader, Matthew Oostveen, CTO and Vice President for Asia Pacific and Japan at Everpure, walked through what was being launched, why now, and what he believes most Indian organisations are getting structurally wrong.
The problem most CIOs have not solved yet
“If you’re a CIO at a large Indian organisation, you have a bunch of servers, storage, network equipment, laptops, PCs, mobile phones. More than likely, you don’t have a complete inventory of where all of that is, which systems are still running, what’s connected to the network. Take this one step further — it means you probably don’t have a really good view of where all your data is. You don’t know what the security implications of that are going to be, and you now have challenges around finding and cataloguing that data,” said Oostveen.
The industry has spent years moving from application-centric to data-centric architecture, placing data at the centre and building applications aroundit. Oostveen’s argument is that with that shift came a set of obligations that most enterprises have not yet met. Knowing where data lives. Who has access to it. Whether it is sensitive. Whether it can legally be used in an AI pipeline. Whether it should ever leave Indian shores. “We’ve moved from application centricity to data centricity,” he said. “When we do that, we need to solve a lot of challenges. That’s before we even get to AI.”
A commissioned IDC Global AI Readiness Survey reinforces this: 94 percent of IT leaders identify data quality as the determining factor in AI success. The implication is stark — most AI projects are constrained not by model capability but by the state of the data underneath them.
What Everpure announced— and what changed from before
The three-part announcement at Accelerate 2026 — Everpure Data Intelligence, Everpure Data Stream, and an updated Intelligent Control Plane together form what the company describes as an end-to-end AI data pipeline.
Data Intelligence is built on the acquisition of 1touch, a data discovery and classification company that Everpure acquired in February 2026 alongside the rebrand. The prior capability, basic metadata management and storage monitoring has been significantly extended. The new system does not merely catalogue metadata tags. It peers inside the data itself.
Legacy data management tools indexed what existed and where. Everpure Data Intelligence reads what the content is, links it to a semantic knowledge graph, and maps it to the business processes that use it. The result is classification that can tell a compliance officer not just that a file exists in a development environment, but that the file contains citizen health data and is therefore subject to DPDP’s restrictions on cross-border transfer.
Critically, the system works across environments that the company does not own — cloud storage, on-premises arrays from competing vendors, mainframe systems running on FCON and ESCON channels, and SaaS environments. Here, a CIO does not need to be an Everpure customer to deploy this as a software tool to discover what exists across their estate.
Data Stream is the newer capability generally available as of June 17. Its predecessor was a manual, people-intensive data preparation process. Building a data pipeline for enterprise AI the old way took months, involved significant engineering lift, and required copying data out of production environments into external vector stores. Data Stream compresses that to days and removes the need to exfiltrate data at all.
The mechanism is GPU-accelerated ingestion that takes unstructured enterprise data like documents, PDFs, transactional records and converts it into vector format for use in AI models, without requiring the data to leave the corporate network. The integration is built on NVIDIA’s AI Data Platform reference design, with FlashBlade providing the storage layer beneath it.
What Indian CIOs need to do differently
Beyond the product announcements, the interaction surfaced a perspective on Indian enterprise IT.
On the question of data silos, which Everpure’s announcements are partly designed to address, Oostveen said “You’ve been trying and failing at unifying your silos for decades because you’re trying to do the same thing. You’re adding more complexity to the environment. All I can think when I hear how organisations are planning their path forward is: you are building tomorrow’s silos today.”
As organisations rush to deploy AI tools with standing up vector databases, building RAG pipelines, integrating LLMs with enterprise data, many are creating new, isolated data structures that replicate the fragmentation they have spent years to solve.
In his prescription for the CIOs dealing with legacy data he says, “ Don’t try to lift and shift everything. Start with knowing where it is. Deploy the intelligence tool, have it scan across your network, and if it finds something that makes you liable from a sovereignty or compliance perspective, bring that back to a safe landing destination — but keep the architecture in place. Over time, build toward a unified data environment. But it’s a huge heavy lift to do all of this at once. Start with the simple stuff: knowing where it is, moving it to a safe location, and then build the strategy.”
The Evergreen One addition
One addition in the Accelerate announcement that carries specific relevance for Indian enterprises evaluating AI infrastructure investment is the expansion of Evergreen One — Everpure’s as-a-service consumption model.
Previously, the model allowed customers to scale capacity on demand, like a cloud consumption model, without capital outlay. The update adds a performance dimension; customers can now burst performance by up to 25 percent above their baseline guarantee when Ai workloads require it, without committing to that performance level permanently.
For Indian enterprise CIOs caught between the pressure to scale AI infrastructure quickly and the regulatory requirement to keep certain data classes on Indian soil, the Evergreen One model can offer an alternative to hyperscaler dependency, one that can scale without forcing a hard architectural commitment upfront.