Hewlett Packard Enterprise Co. and Nvidia Corp. today are extending the Nvidia AI computing by HPE initiative they announced last year with new servers, software and storage targeted specifically at artificial intelligence use cases.

They include a new HPE Private Cloud AI developer system using Nvidia accelerated computing and featuring an integrated control node, a full suite of AI development software and 32 terabytes of integrated storage. “We’re extending the private cloud AI family to make the portfolio more accessible with the ability to start even faster,” said Cheri Williams, senior vice president and general manager of private cloud and AI solutions at HPE. “This solution delivers instant AI development capabilities, and it has the same predefined software tools as in our standard Private Cloud AI.”

HPE is also expanding its Private Cloud AI offering with support for the new Nvidia AI Data Platform. Private Cloud AI is a self-service cloud that transforms data through continuous data processing leveraging Nvidia’s accelerated computing, networking, AI software and enterprise storage.

HPE is unveiling a unified data layer for AI that supports structured, unstructured and streaming data across hybrid cloud environments. The Nvidia AI Data Platform is a customizable reference design that integrates Nvidia accelerated computing, networking, and AI software with HPE storage. It will be available in HPE Private Cloud AI and support HPE Alletra Storage MP B10000, HPE Alletra Storage MP X10000 and HPE GreenLake for File Storage.

Alletra Storage MP is a software-defined, scale-out data system combining all-flash object storage with unified management capacity. It will be expanded to support Nvidia Networking. Its integrated HPE Data Fabric supplies AI models with structured, unstructured and streaming data. The fabric software will support HPE Alletra Storage MP X10000 and Apache Iceberg as the data backbone of the HPE Private Cloud AI data lakehouse.

Also, HPE GreenLake for File Storage, the HPE Private Cloud AI file storage service, is now validated for enterprise AI Factories with the new Nvidia-Certified Storage Program.

AI-optimized storage

Enhancements to Alletra Storage MP X10000 will enable organizations to create AI-ready object data with new automated, inline metadata tagging for faster ingestion by downstream AI applications. HPE said it expects to improve the X10000’s performance further through a collaboration with Nvidia that will enable a direct data path for remote direct memory access transfers between GPU memory, system memory and the X10000.

The Alletra Storage MP B10000 multiprotocol storage subsystem is expanding to address a wider variety of workloads via unified file access, enhance ransomware protection and simpler data movement between on-premises data centers and public clouds. HPE said it’s the first vendor to support disaggregated scale-out unified block and file storage on a single operating system and storage architecture. That combination allows management of the entire storage environment to be unified.

New HPE Alletra Block Storage for Azure simplifies data management and workload placement across hybrid cloud with software-defined storage on Microsoft Corp.’s Azure cloud and AIOps automate provisioning and support.

New multi-layered ransomware detection and protection in the B10000 array and Zerto data replication and protection software protect data across storage arrays and applications.

Private Cloud AI expands Nvidia support

HPE Private Cloud AI now also supports rapid deployment of Nvidia blueprints covering agentic and physical AI applications. HPE said that means pre-validated blueprints like the Multimodal PDF Data Extraction Blueprint and Digital Twins Blueprint can be deployed faster.

HPE’s OpsRamp hybrid digital operations management platform has been enhanced with graphic processing unit optimization and training and inference workload observability when running on Nvidia accelerated computing clusters. The capability is available through HPE Private Cloud AI and on a standalone basis for large clusters. It’s also available as a new day 2 operational service HPE Complete Care Service with Nvidia GPU optimization.

New agentic AI initiatives include Zora AI by Deloitte LLP in the HPE Private Cloud AI. HPE will be the first customer to deploy the Zora agentic AI platform for finance, which supports dynamic, on-demand and interactive use cases include financial statement analysis, scenario modeling, competitive and market analysis. It will be available worldwide.

CrewAI Inc., the developer of a popular tool for building artificial intelligence agents, has joined the HPE Unleash AI program to deliver multi-agent automation: Combined with HPE Private Cloud AI, CrewAI can securely deploy and scale special-purpose agent-driven automations.

HPE is also adding professional services for agentic AI that combine Nvidia NIM microservices and Nvidia NeMo with HPE Private Cloud AI.

New Blackwell-based AI servers

New HPE ProLiant Gen12 servers feature an enhanced and dedicated security processor enclave that establishes an unbreakable chain of trust to protect against firmware attacks and creates full line-of-sight from the factory and throughout HPE’s trusted supply chain. HPE iLO 7, an embedded server management technology that provides out-of-band management capabilities for HPE ProLiant servers, will provide post-quantum cryptography that meets the requirements for FIPS 140-3 Level 3 high-level cryptographic security standard.

A new integrated AI environment in a software container called AI Mod POD provides a modular, performance-optimized data center for AI and high-performance computing workloads. The modular data center supports up to 1.5 megawatts of power per module and can be set up quickly. It supports HPE’s AI and HPC servers and HPE Private Cloud AI and uses HPE’s patented Adaptive Cascade Cooling technology, a hybrid system that supports air and 100% and hybrid liquid cooling.

“It will enable three times faster deployment with industry leading power usage effectiveness of under 1.1,” said Trish Damkroger, senior vice president and general manager of high-performance computing and AI. “It is a fully integrated solution in a container.”

Photo: Mark Albertson/SiliconANGLE