Everpure expands AI storage, adds Data Stream beta

Everpure has expanded its Evergreen//One consumption offering for AI workloads to include FlashBlade//EXA and outlined a forthcoming beta for a new data-pipeline product, Everpure Data Stream.

It positioned both moves as ways to reduce operational friction that can slow enterprise AI deployments. Evergreen//One for AI now covers FlashBlade//EXA, which Everpure markets as a scale-out storage platform for large training and inference environments. Data Stream is scheduled to enter beta later in 2026.

Evergreen//One is Everpure's pay-as-you-go model for storage. Extending it to FlashBlade//EXA gives organisations another option to consume storage as an operating expense rather than buy systems upfront. Everpure also framed the model as a way to manage shifting demand as AI projects move from experiments to repeatable production workflows.

Kaycee Lai, Vice President, AI, Everpure, said: "Most AI projects fail to reach production for enterprises because many treat AI as 'just another workload.' We are helping customers break down siloed data and move AI initiatives from pilot to production with infrastructure that delivers guaranteed performance, flexibility, and growth. Whether organizations are preparing data or running large-scale inference, we ensure they have the tools to succeed."

Consumption model

Options Technology, which provides technology services to capital markets firms, said Evergreen//One has changed how it plans capacity. Andrea Moccia, VP, AI and Data, Options Technology, said: "Evergreen//One completely solved our capacity planning challenges. We can now deploy storage anywhere in the world, consume it on a pay-as-you-go basis, and scale on demand-bringing down the barriers to global growth and flexing to meet the demands of rapidly evolving AI workloads."

Everpure also highlighted a customer using FlashBlade//EXA at scale. STN said it significantly expanded a FlashBlade//EXA deployment without seeing the performance drop-offs that can emerge in distributed storage environments as clusters grow.

Sabur Mian, CEO and Founder, STN, said: "Everpure's technology allows us to deliver data storage performance at unprecedented consistency for even the most demanding AI workloads. In a typical storage infrastructure, researchers might start training a model on four nodes and get good performance-but as soon as they start scaling up, that performance collapses. With FlashBlade//EXA, we've scaled up to 192 nodes so far, and we've yet to find the limit."

Benchmark claims

Everpure pointed to industry benchmarks as evidence of FlashBlade//EXA performance in AI-related storage tests. It said FlashBlade//EXA set a high score in the SPEC Storage AI_Image benchmark, citing 6,300 simultaneous AI jobs in that test. It also referenced MLPerf-related measurements, saying the platform sustained more than 90% GPU utilisation across large NVIDIA Hopper clusters, while noting the figures were based on internal component measurements rather than an official MLPerf submission.

Everpure said it is aligning FlashBlade//EXA with the modular NVIDIA STX reference architecture and described that work in the context of Nvidia's Vera Rubin platform. It also referenced BlueField-enabled storage controllers and context memory architectures as part of its configuration approach. Everpure said the combination addresses storage access requirements for large-scale inference workloads that rely on rapid data retrieval.

Everpure also said it plans to extend NVIDIA-Certified Storage validation to FlashBlade//EXA. It described this as a route to a higher NVIDIA certification tier it referred to as the "NCP" level, linking that positioning to Nvidia Cloud Partner reference architectures.

Data pipelines

Alongside the storage and consumption updates, Everpure described Everpure Data Stream as an automated pipeline that links ingestion with inference. It said the product reduces manual data movement and administrative effort, with a beta due later in 2026.

Everpure said Data Stream is co-engineered with Supermicro and built on the NVIDIA AI Data Platform reference design. It described the design as a compact option for organisations that want a packaged way to prepare and serve data into AI environments, combining Supermicro hardware with Everpure's software-defined storage.

On hardware support, Everpure said it supports accelerated platforms including the NVIDIA RTX PRO 6000 Blackwell Server Edition. It also said it will expand support to the NVIDIA RTX PRO 4500 Blackwell Server Edition GPU.

Everpure said the combination of FlashBlade//EXA, Evergreen//One, and Data Stream reflects an emphasis on repeatable data preparation and performance validation as AI use expands across organisations.