Everpure announces Evergreen//One for AI and upcoming Data Stream Beta

Everpure has announced Evergreen//One for FlashBlade//EXA and the upcoming beta of Everpure Data Stream to help organizations reduce cost and complexity barriers that stall enterprise AI projects.

Evergreen//One (EG1) for AI now extends across FlashBlade//EXA, providing the massive performance, scalability, and throughput required for large-scale training and inference. Complementing this, the Everpure Data Stream Beta–launching later in 2026–accelerates time-to-result by eliminating the friction of manual data movement with a direct, automated pipeline from data ingestion to inference.

“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,” said Kaycee Lai, Vice President, AI, Everpure.

AI deployments succeed when infrastructure keeps GPUs running at peak capacity. Everpure is aligning FlashBlade//EXA with modular NVIDIA STX reference architecture to support the next generation of AI factories powered by the Vera Rubin platform. By combining EXA’s performance and scalability with STX components such as BlueField-enabled storage controllers and context memory architectures, Everpure optimizes the AI pipeline—from data preparation to long-context inference. This architecture specifically addresses the high-performance context memory requirements of giga-scale inference, providing the low-latency data access necessary to sustain agentic workflows and multi-step reasoning systems at scale.

Recent industry benchmarks from SPECstorage Solution 2020 and MLPerf validate FlashBlade//EXA’s ability to deliver consistent, repeatable performance needed to turn raw data into trained models at scale.

• Record-breaking SPEC benchmarks: FlashBlade//EXA achieved the highest score ever recorded for the SPEC Storage AI_Image benchmark. By successfully powering 6,300 simultaneous AI jobs, Everpure demonstrated FlashBlade//EXA’s ability to sustain more concurrent training tasks at full speed than any other solution on the market today.
• Redefining AI economics: FlashBlade//EXA moves data twice as fast as its closest competitor while occupying less than half a rack of storage. By sustaining over 90% GPU utilization across large NVIDIA Hopper clusters—validated by MLPerf^[1] model-driven workloads—EXA ensures compute resources never sit idle. As workloads grow, EXA scales linearly to dramatically reduce idle time and ensure your compute resources are never waiting around for their next task.

Additionally, extending NVIDIA-Certified Storage (NVCS) validation to FlashBlade//EXA provides the foundation for full-stack confidence. This integration creates a definitive path toward the NVCS ‘NCP’ certification level, purposefully aligned with NVIDIA Cloud Partner (NCP) reference architectures.

To fully operationalize these performance gains, Everpure Data Stream simplifies curation and orchestration–ensuring high quality, AI ready data flows seamlessly into AI infrastructure. This accelerates time to insight and shortens the path from pilot to production, ensuring models are always working on the freshest data without administrative overhead.

Lowering the barrier to entry with a compact AI design is the Everpure Data Stream, co-engineered with Supermicro, and built on the NVIDIA AI Data Platform reference design. By combining Supermicro’s hardware with Everpure’s software-defined storage, enterprises can rapidly unlock the true value of their data.

As enterprise AI Factories require an AI Data Platform to prepare and deliver AI-ready data, Everpure also supports accelerated platforms including the NVIDIA RTX PRO 6000 Blackwell Server Edition and will expand support to the NVIDIA RTX PRO™ 4500 Blackwell Server Edition GPU.

Success in the AI-driven landscape requires a strategy rooted in continuous data optimization–not a one-time infrastructure upgrade. Everpure’s platform provides this essential foundation, treating AI readiness not as a single milestone, but as an ongoing journey of preparation and performance validation.