At HPE Discover Las Vegas (running through June 18), HPE and NVIDIA are expanding the HPE AI Factory portfolio to support the shift from agentic AI pilots to governed, production-scale deployments.
The updates center on three pillars: a new CPU purpose-built for agent loops, an agentic operating system layer with governance controls, and Confidential Computing extended across every HPE AI Factory solution.
NVIDIA Vera CPU for Agent Workloads
The NVIDIA Vera CPU (part of the Vera Rubin platform) is now available with HPE Private Cloud AI via the HPE ProLiant Compute DL394 Gen12 server. It is described as the first CPU built specifically for agents — optimized for tool calls, orchestration, and real-time data processing across the agent loop, delivering deterministic low-latency performance.
The Vera Rubin platform targets frontier-scale models larger than 1 trillion parameters. Full-stack NVIDIA Confidential Computing ships across every chip in the platform. HPE is also adding the HPE Compute XD700 (built on NVIDIA HGX Rubin NVL8), supporting up to 128 Rubin GPUs per rack.
Early enterprise exploration is underway: the New York Stock Exchange, in collaboration with Redpanda and HPE, is testing Vera CPU on the DL394 Gen12.
The rack-scale NVIDIA Vera Rubin NVL72 system is available from HPE.
NVIDIA Agent Toolkit Integrated into HPE Private Cloud AI
NVIDIA Agent Toolkit — including NVIDIA Nemotron open models, NVIDIA OpenShell secure runtime, and NVIDIA NemoClaw blueprints — is now available with HPE Private Cloud AI.
This combination is positioned as an “agentic AI operating system” for monitoring agent behavior, enforcing governance policies, and safely running autonomous, long-running multi-agent systems.
HPE Private Cloud AI adds secure local agent registration, allowing customers to approve models, skills, and tools against centralized policies before execution. New HPE Zerto Software capabilities can detect rogue agent actions and use continuous data protection to rewind systems to a clean state.
On the data layer, HPE Alletra Storage MP X10000 (which has achieved the foundation level of NVIDIA-Certified Storage) automatically applies metadata and governance policies to unstructured data, improving token throughput for AI pipelines.
Confidential Computing Extended Across the Portfolio
NVIDIA Confidential Computing is now available across HPE AI Factory at Scale, HPE Sovereign AI Factory, and HPE Private Cloud AI through HPE Services.
It provides hardware-based protection for AI workloads and sensitive data assets during execution (via cryptographic attestation and encryption at every stage) while maintaining NVIDIA acceleration performance. The HPE ProLiant Compute DL380a has achieved certification under the NVIDIA-Certified Systems program for Confidential Computing.
NVIDIA BlueField DPUs and NVIDIA DOCA deliver in-silicon zero-trust policy enforcement, runtime threat detection, and network encryption across AI workloads, agents, and data — with no performance tradeoffs noted.
Enhanced Full-Stack NVIDIA Integration
All three HPE AI Factory solutions now include:
NVIDIA RTX PRO 6000 Blackwell Server Edition GPUsNVIDIA Spectrum-X Ethernet networking (the standard across the portfolio)NVIDIA BlueField-3 DPUs andNVIDIA ConnectX-8 SuperNICs
For next-generation factories, every Vera Rubin NVL72 system ships with NVIDIA networking built in (Vera BlueField-4 DPUs, ConnectX-9 SuperNICs, and Spectrum-X Ethernet). NVIDIA Spectrum-6 switching delivers 1.6x higher networking performance for AI communication versus off-the-shelf Ethernet. InfiniBand options (including NVIDIA Quantum-X800 InfiniBand with HPE Cray Supercomputing GX5000) are also available for large-scale and sovereign workloads.
The HPE Unleash AI partner ecosystem is expanding with nearly a dozen new AI software partners.
Actions to Take
Enterprises moving agentic AI into production: Evaluate HPE Private Cloud AI with the integrated NVIDIA Agent Toolkit + HPE Zerto for governed, rollback-capable multi-agent systems. The local policy approval and rogue-action detection features directly address governance and security gaps that have blocked many deployments.
Organizations with sensitive data or sovereign requirements: Prioritize configurations with full NVIDIA Confidential Computing (now standard across HPE AI Factory solutions). The cryptographic attestation and encryption-at-every-stage model, combined with BlueField/DOCA zero-trust enforcement, provides hardware-rooted protection without noted performance penalties.
Teams running frontier-scale models (>1T parameters): Assess the Vera Rubin NVL72 rack-scale systems and HPE Compute XD700 (up to 128 Rubin GPUs/rack) now available through HPE. The Vera CPU addition targets the orchestration and real-time data processing layers that pure GPU clusters often bottleneck.
Data platform and storage teams: Test HPE Alletra Storage MP X10000 (NVIDIA-Certified) for automated metadata/governance application on unstructured data. The documented improvement in token throughput is a direct infrastructure efficiency lever for RAG and agent pipelines.
Adtech, finance, and regulated verticals: The NYSE/Redpanda exploration of Vera CPU on HPE ProLiant DL394 Gen12 offers an early signal for low-latency, deterministic agent workloads in high-stakes environments.
These updates make the HPE AI Factory with NVIDIA one of the more complete turnkey offerings for enterprises that need both raw scale (Rubin GPUs, high-performance networking) and the operational controls required to run agents safely in production.