HPE expands Private Cloud AI factory portfolio to support next-gen autonomous agents

Hewlett Packard Enterprise Co. today unveiled a massive expansion of its artificial intelligence infrastructure for private cloud deployments, enhancing its portfolio with Nvidia Corp.’s new Vera chips and tools for developers.

The new offerings are meant to provide enterprises with everything they need to bring autonomous AI agents online within secure, on-premises environments to ensure they can operate with strict compliance in regulated industries.

Announced at HPE Discover in Las Vegas today, the updates include a new, high-performance ProLiant server powered by Nvidia’s Vera central processing units, the Nvidia Agent Toolkit and an extension of Nvidia Confidential Computing, which is now available across HPE’s entire Private Cloud AI server lineup.

The new HPE ProLiant Compute DL394 Gen12 server is expected to be available in early 2027, and will be one of the first to feature Nvidia’s new Vera CPUs, which are purpose-built to support AI agents. The chips are designed to handle the rapid tool calls, complex orchestration and real-time data processing that’s required to support thousands of autonomous agents, enabling deterministic, low-latency performance for on-premises agentic deployments.

The Vera CPUs are part of the broader Nvidia Vera Rubin platform, as the chipmaker expands production of them in its massive NVL72 rack-scale systems. HPE will be one of the first server makers to get its hands on these too. When it launches its upcoming HPE Compute XD700 system built on the Nvidia HGX Rubin NVL8 architecture, it will support up to 128 Rubin graphics processing units per rack.

On the software side, HPE said its Private Cloud AI servers will now ship with the Nvidia Agent Toolkit, which encompasses the chipmaker’s open-source Nemotron models, the OpenShell secure runtime and NemoClaw blueprints. These allow it to serve as a kind of agentic AI operating system, providing everything enterprises need to design, build, run, orchestrate and monitor autonomous multi-agent systems at large scale.

Meanwhile, HPE is providing its own capabilities in the shape of enhanced governance and security tools. These include the addition of secure local agent registration features to HPE Private Cloud AI. Customers will be able to vet and approve AI models and tools against centralized security policies before they’re deployed.

In case an agent decides to go off-the-rails, HPE’s Zerto Software will immediately detect what’s happening and rewind its actions to return it to a clean state. Moreover, HPE Alletra Storage MP X10000 is there to apply metadata policies to unstructured data to prevent misuse and unauthorized access.

The final piece of the puzzle is Nvidia Confidential Computing, which is being rolled out across HPE’s entire Private Cloud AI hardware portfolio. HPE explained that modern applications frequently have to access sensitive data, and so there’s a need to protect that information at the moment it’s unencrypted for processing.

Nvidia’s confidential computing capabilities help with this by establishing a cryptographic “chain of trust” that protects data by ensuring it’s only processed in a secure, isolated environment within the server. As part of this rollout, the HPE ProLiant Compute DL380a is the first to be certified for confidential computing.

The updates are meant to cater to the complex requirements of enterprises that need to run AI agents on-premises for reasons related to compliance and concerns around data leaks. Risk-averse companies can’t afford to take the risk to store sensitive data and run mission-critical workloads in the cloud. With HPE Private Cloud AI, enterprises will be able to spin up the high-performance turnkey environments AI agents require and ensure their proprietary data and workloads runs safely behind their corporate firewalls, while benefiting from low-latency processing.

Image: Nvidia