HPE expanded its AI cloud platform with enhanced NVIDIA GPU integration and capabilities.

HPE is upgrading its Private Cloud AI stack with Nvidia technology ahead of its Discover event in Barcelona. The latest RTX PRO 6000 Blackwell Server Edition GPUs will be available across all of HPE's AI-focused private cloud platforms, along with STIG-hardened NIMs. STIG refers to Security Technical Implementation Guides published by the Defense Information Systems Agency (DISA), and NIMs are Nvidia's inferencing microservices for deploying AI models at scale. HPE is also adding support for GPU fractionalization—virtualization for Nvidia GPUs aimed at optimizing utilization and lowering costs—to those private cloud SKUs. Additionally, HPE is bringing in Datacenter Ops Agents to simplify datacenter management and enable operations across agentic AI and hybrid cloud environments.

Following the acquisition of Juniper Networks, which closed in July, HPE is integrating Juniper technology with its AI and private cloud services. The first fruit of this integration is Edge on-ramp using the MX family of routers to link a private cloud with users and devices, while Datacenter interconnect (DCI) employs PTX routers to connect AI clusters operating across long distances or across multiple clouds. For storage, HPE announced Alletra Storage MP X10000 Data Intelligence Nodes with built-in capabilities to prepare data for AI processing.

Chief technology officer Fidelma Russo emphasized that data preparation, not GPU capacity, is the primary bottleneck for enterprises pursuing AI: "Most enterprises are discovering that their bottleneck to AI is not GPU capacity. It's preparing the data for GPUs." She explained that the Alletra platform solves this by enriching and structuring data inline as it enters the system, performing metadata tagging, embedded vector generation, and formatting automatically. "What this means to a customer is they don't need a plethora of separate data prep tools before touching an LLM. And this is all built on our disaggregated Alletra Storage MP architecture, which means that you can scale capacity and performance independently. And what is the result? It's a faster pipeline, it's higher GPU utilization," Russo claimed.

HPE and Nvidia plan to open an AI Factory Lab in Grenoble, France, due to open in Q2 2026, for customers to try out and refine their workloads. The facility is "dedicated to advancing AI factories, gigafactories and sovereign initiatives in the region," Russo said, and will be equipped with the latest HPE and Nvidia AI factory technology. "We believe that this will be a large accelerator to helping our customers get their workloads ready for production into their environments," she added. In the UK, Carbon3.ai is launching a similar initiative with its Private AI Lab, built on HPE's Private Cloud AI platform and intended to boost enterprise adoption by providing a workspace to take AI projects from pilot to production. Carbon3.ai says it is building a sovereign AI infrastructure network for UK customers powered by renewable energy, with chief Tom Humphreys stating: "UK enterprises want secure, sovereign, and sustainable AI infrastructure they can trust. We're accelerating enterprise adoption and helping the UK convert its AI potential into economic impact."