HPE and NVIDIA expand their AI Factory to bring agents to production with Vera, more security and deeper stack integration

HPE and NVIDIA have taken advantage HPE Discover Las Vegas to announce an expansion of their joint proposal HPE AI Factory with NVIDIAa platform with which both companies want to accompany the transition of agentic artificial intelligence from proofs of concept to real production environments.

Within that movement, NVIDIA highlights several new elements. The first is the arrival of NVIDIA Vera to HPE Private Cloud AI, the turnkey solution developed between both companies. As NVIDIA explains, Vera is its first CPU designed specifically for agents, with an approach aimed at tool calls, orchestration and real-time data processing within the so-called “agent loop”. Availability is in 2027 through the HPE ProLiant Compute DL394 Gen12 server.

The release adds that the New York Stock Exchange, in collaboration with Redpanda and HPE, is already listed as an early enterprise customer exploring the use of Vera in that environment. In addition, NVIDIA remembers that Vera is part of the platform Vera Rubindesigned for frontier models with more than a billion parameters, and which will also be deployed using Vera Rubin NVL72 systems available through HPE. In parallel, HPE will incorporate the HPE Compute XD700 system, based on NVIDIA HGX Rubin NVL8, with support for up to 128 Rubin GPUs per rack.

HPE Private Cloud AI adds agent tools and new governance layers

Another notable announcement is the incorporation of NVIDIA Agent Toolkit to HPE Private Cloud AI. NVIDIA presents it as an enterprise agentic AI operating system that combines Nemotron open models, the OpenShell secure environment and NemoClaw blueprints. The idea is to give organizations tools to monitor agent behavior, apply governance policies, and deploy long-lived, autonomous multi-agent systems with more control.

HPE also adds its own management functions, such as secure local agent registrationwhich allows models, tools and skills to be approved based on centralized security policies before launching them. The release also mentions new capabilities in HPE Zerto Software to detect anomalous agent actions and roll back to a clean state through continuous data protection.

At the storage layer, HPE highlights that Alletra Storage MP X10000 has reached the foundation level within NVIDIA-Certified Storage and will be able to automatically apply metadata and governance policies on unstructured data to prepare it for AI pipelines. The goal, as explained, is to improve token processing performance and make it easier to feed agent and model workflows.

Confidential Computing and NVIDIA networking as a common foundation across the portfolio

The third major block of the announcement is security. NVIDIA communicates that NVIDIA Confidential Computing extends to the entire HPE AI Factory portfolio through HPE Services, including HPE AI Factory at Scale, HPE Sovereign AI Factory, and HPE Private Cloud AI. The company positions this layer as essential to protect sensitive models and data during execution, especially in on-premise and sovereign deployments, through encryption and cryptographic attestation throughout the entire process.

Along the same lines, the server HPE ProLiant Compute DL380a has obtained certification within the NVIDIA-Certified Systems for NVIDIA Confidential Computing program. In addition, NVIDIA highlights that BlueField DPUs and DOCA provide zero-trust control in silicon, real-time threat detection and network encryption without sacrificing performance.

The infrastructure part is completed with a deeper integration of the entire NVIDIA stack within the HPE portfolio. Likewise, NVIDIA adds that Spectrum-X becomes the networking standard for HPE AI Factory, while for large-scale and sovereign loads, HPE will also add InfiniBand options, such as Quantum-X800 in combination with the HPE Cray Supercomputing GX5000. The suite is supported by NVIDIA reference architectures, NVIDIA AI Enterprise software and the HPE Unleash AI ecosystem, which is also expanded with new software partners.