Nvidia GTC 2026: HPE Alletra Storage MP X10000 Becomes First Nvidia-Certified Storage Object-Based Platform for Enterprise AI

Key Takeaways

• First object storage with Nvidia Certified Storage Foundation
• Validated for AI workloads up to 128 GPUs
• Enables faster data feeding to GPU clusters
• Scales capacity and performance independently
• Strengthens HPE‑Nvidia AI ecosystem partnership

Summary

HPE announced that its Alletra Storage MP X10000 has earned Nvidia‑Certified Storage Foundation validation, becoming the first object‑based platform to achieve this milestone. The certification confirms the system can sustain AI workloads scaling to 128 GPUs, delivering the throughput and reliability required for massive unstructured data pipelines. Nvidia’s program tests performance, availability and integration with accelerated computing, underscoring the storage layer’s role in modern AI architectures. The partnership highlights HPE’s strategy to position its storage as a core component of enterprise AI infrastructure.

Pulse Analysis

The rapid growth of generative AI and large‑scale model training has shifted the performance bottleneck from compute to data movement. While GPUs provide raw horsepower, the ability to stream petabytes of unstructured data quickly determines overall system efficiency. Nvidia’s Certified Storage program was introduced to benchmark storage solutions against these demanding AI workloads, ensuring they meet strict latency, throughput, and reliability standards. Achieving Foundation‑level certification signals that a storage platform can reliably feed data to up to 128 GPUs without throttling.

HPE’s Alletra Storage MP X10000 leverages a scale‑out object architecture combined with built‑in data intelligence to meet Nvidia’s criteria. The system separates capacity and performance scaling, allowing enterprises to add storage nodes as data volumes grow while preserving low‑latency access for GPU clusters. Benchmarking demonstrated enterprise‑grade availability and the ability to sustain high I/O rates required for training, fine‑tuning, and inference pipelines. Integrated features such as inline data preparation and enrichment further reduce the need for separate preprocessing steps, accelerating the end‑to‑end AI workflow.

For the market, this certification strengthens the HPE‑Nvidia alliance and positions object storage as a first‑class component of AI infrastructure, challenging traditional block‑oriented solutions. Enterprises seeking to operationalize AI at scale can now rely on a validated storage foundation that promises higher GPU utilization and lower total cost of ownership. As more vendors pursue similar certifications, the ecosystem will likely see tighter integration between storage, networking, and accelerated compute, driving faster time‑to‑value for AI initiatives.