HPE, Nvidia expand AI partnership

HPE and Nvidia have boosted their partnership, adding a new server blade, GPU support, enhancements to HPE’s turnkey private AI package, and services targeting enterprise customers with growing AI workloads. A considerable portion of the HPE-related news coming from Nvidia’s GTC event is aimed at the high end of the AI workload spectrum and targets service providers and neocloud operators. For example, HPE announced its Nvidia Vera Rubin NVL72 rack-scale system (pictured) that it says is capable of supporting in excess of 1 trillion AI parameters. In addition, the companies announced the new GX240 liquid-cooled compute blade for HPE’s Cray Supercomputer GX5000. For enterprise customers, HPE enhanced its Private Cloud AI package, which integrates Nvidia GPUs, networks, and software with HPE’s AI memory, computing and GreenLake cloud support. Also, HPE has extended its network expansion racks beyond the current level of 64 GPUs to scale up to 128 GPUs, allowing customers to run larger, more demanding AI workloads. HPE Private Cloud AI delivers a preconfigured hardware and software stack featuring the latest Nvidia AI Enterprise software and blueprints. It now includes the updated Nvidia AI‑Q blueprint for AI agents and the just-released Nvidia Omniverse blueprint for digital twins. The most recent Nvidia AI-Q blueprint enables developers to build customizable AI agents that they own, inspect and control. Additional new HPE Private Cloud AI features include: * An air-gapped capable configuration for isolated or sovereign deployments * HPE ProLiant Compute DL380a Gen12 servers and HPE Private Cloud AI systems based on the DL380a are being certified for Fortanix Confidential AI, a joint offering using Nvidia Confidential Computing that enable secure on-premises deployments for AI models and processing of sensitive data without exposure. * The latest HPE ProLiant servers and HPE AI factories now support the latest Nvidia Nemotron open models to simplify deployment of secure, on‑prem and sovereign infrastructure. * Support for the Nvidia RTX Pro 6000 Blackwell server GPUs are now offered as part of the package. Nvidia’s RTX Pro 6000 Blackwell Server Edition GPUs will now be standardized across HPE AI factory configurations, and RTX Pro 4500 Blackwell Server Edition GPUs will be available in other ProLiant server models aimed at edge deployments, small language models and more. At the high end, HPE rolled out one of the first Nvidia Vera CPU systems, the Nvidia Vera Rubin NVL72 rack-scale system. This flagship AI system is engineered for frontier‑scale models in excess of 1 trillion parameters. It features 36 Nvidia Vera CPUs, 72 Nvidia Rubin GPUs, sixth-generation NVLink support for scale-up networking, as well as ConnectX‑9 SuperNIC, and BlueField‑4 DPUs along with HPE’s liquid cooling integration. In addition, the company announced the HPE Cray Supercomputing GX240 liquid-cooled compute blade for its GX5000 platform. The GX240 starts with 16 Nvidia Vera CPUs per blade and scales to 40 blades per rack, supporting up to 640 Nvidia Vera CPUs and 56,320 ARM cores per rack. In addition, HPE said new network connectivity—Nvidia Quantum-X800 InfiniBand—optimized for large-scale system connectivity is now available with HPE Cray Supercomputing GX5000. The Quantum-X800 InfiniBand switches provide 144 ports of 800 Gb/s connectivity per port with power efficiency features, the vendor stated. The vendor also rolled out the HPE Compute XD700, an AI server built on Nvidia HGX Rubin NVL8. The system is designed to deliver higher GPU density per rack and reduce space, power, and cooling costs while increasing AI training and inference throughput. Each rack of XD700 servers supports up to 128 Rubin GPUs, providing double the GPU density compared to the previous generation, according to HPE. During his GTC opening keynote, Nvidia CEO Jensen Huang said: “Vera is arriving at a turning point for AI. As intelligence becomes agentic—capable of reasoning and acting—the importance of the systems orchestrating that work is elevated. The CPU is no longer simply supporting the model; it’s driving it. With breakthrough performance and energy efficiency, Vera unlocks AI systems that think faster and scale further.”