Hewlett Packard Enterprise (HPE) and Nvidia have unveiled new hardware and software solutions for enterprise AI applications at Nvidia GTC 2025.
Beyond powerful servers equipped with Nvidia's latest Blackwell chips – graphics processors optimized specifically for AI computations – HPE introduced a unified data layer for AI and significant improvements to its enterprise storage portfolio.
The new data layer combines structured data (like tables and databases) with unstructured data (such as texts, images, and videos) to accelerate the AI data lifecycle by merging HPE's powerful Data Fabric and enterprise storage with intelligent data capabilities.
Private cloud for AI data
HPE Private Cloud AI – an internal cloud solution that companies can operate on their own hardware – now supports the Nvidia AI Data Platform. This combination allows enterprises to leverage their own data for AI applications without moving it to public clouds like AWS or Google Cloud, which is particularly important for sensitive information.
The Nvidia AI Data Platform serves as a customizable reference design that combines Nvidia's accelerated computing power, networking technology, and AI software with enterprise storage. HPE will support this platform through HPE Private Cloud AI as well as HPE Alletra Storage MP B10000, HPE Alletra Storage MP X10000, and HPE GreenLake for File Storage – all storage solutions designed for large data volumes in enterprise environments.
According to HPE the platform accelerates AI reasoning workloads with specialized AI agents using Nvidia AI Enterprise Software, including Nvidia AI-Q Blueprints, Nvidia NIM Microservices, and new Nvidia Llama Nemotron Reasoning models.
Unified data layer for hybrid clouds
The intelligent, unified data layer aims to accelerate data flow across HPE GreenLake Cloud and multi-vendor, multi-cloud environments. A central component is the HPE Data Fabric Software, which now supports HPE Alletra Storage MP X10000 and Apache Iceberg (an open table format for very large analytics). It forms the data backbone of the HPE Private Cloud AI Data Lakehouse – an architecture combining the advantages of data warehouses and databases – and the unified data layer. This expansion ensures AI models receive continuous feeds of optimized, high-quality structured, unstructured, and streaming data.
Additionally, the latest improvements to HPE Alletra Storage MP X10000 enable businesses to create AI-ready object data with new automated, inline metadata tagging. This allows companies to enrich their object data during storage, which should accelerate intake by downstream AI applications.
New servers with Blackwell chips
HPE is introducing four new server models based on Nvidia's latest Blackwell architecture. The Nvidia GB300 NVL72 by HPE is designed for training very large AI models with trillions of parameters and comes with liquid cooling. For training and fine-tuning large AI models and complex workloads like agent-based AI, HPE offers the ProLiant Compute XD with Nvidia HGX B300.
The HPE ProLiant Compute DL384b Gen12 with Nvidia GB200 Grace Blackwell NVL4 combines Nvidia's Arm-based Grace CPU with Blackwell GPU for scientific calculations and AI applications. For less demanding tasks, the HPE ProLiant Compute DL380a Gen12 with Nvidia RTX PRO 6000 is available – a PCIe-based solution for AI inference and visual computing with lower performance requirements.
Expanded block storage with new features
HPE is extending Alletra Storage MP B10000 to help customers manage various workloads more easily. As the first provider in the industry, HPE supports disaggregated scale-out unified block and file storage on a single operating system and storage architecture. Block storage stores data in fixed blocks (ideal for databases), while file storage manages hierarchically organized files (ideal for shared documents). This unification allows customers to manage their entire environment as a unified fleet, reducing complexity and improving efficiency.
For hybrid cloud environments – combinations of local data centers and cloud services – the new HPE Alletra Block Storage for Azure offers simplified data management with software-defined storage on Microsoft Azure. Comprehensive AI-powered operations automate provisioning and support, establishing what HPE describes as the foundation for autonomous storage management.
To protect against ransomware, customers can now secure their data with multi-layered ransomware detection and protection. The combination of new native ransomware detection and recovery in the B10000 array and Zerto software ensures data protection across storage arrays and applications.
Most of the announced products should be available throughout 2025. HPE Data Fabric with support for HPE Private Cloud AI and HPE Alletra Storage MP X10000 will be available in summer 2025, while the updates for B10000 and X10000 can be ordered starting May 2025.
