HPE Brings Nvidia Agent Toolkit to Private Cloud AI
Key insights
- HPE ProLiant Compute DL394 Gen12 with Nvidia Vera CPUs is expected in early 2027, adding a new server to the Private Cloud AI Factory lineup.
- Nvidia Confidential Computing now spans HPE's entire Private Cloud AI hardware portfolio, creating a cryptographic chain of trust for sensitive workloads.
- HPE Compute XD700 supports up to 128 Rubin GPUs per rack, targeting high-density on-premises autonomous agent deployments.
Why this matters
Regulated enterprises have held back from agentic AI deployments because hyperscaler-hosted runtimes conflict with data residency and compliance mandates, and HPE's combined hardware and software stack addresses both the compute and governance layers in a single on-premises package. Nvidia Confidential Computing as a standard feature, combined with Zerto's agent rollback capability, moves enterprise AI governance from an optional add-on to a baseline expectation for the category. For cloud hyperscalers, this positions on-premises as a structurally separate market segment for autonomous agents, one they cannot easily serve through standard public cloud offerings.
Summary
HPE at Discover in Las Vegas on June 16 expanded its Private Cloud AI Factory with the ProLiant Compute DL394 Gen12, a new server built on Nvidia Vera CPUs arriving in early 2027, and the Nvidia Agent Toolkit including Nemotron models and OpenShell runtime now rolling out across its existing Private Cloud AI servers.
Essentially: (HPE, Nvidia) are jointly targeting regulated enterprises that want autonomous agent infrastructure on-premises rather than in hyperscaler clouds.
- Nvidia Confidential Computing now covers HPE's full Private Cloud AI hardware line, adding a cryptographic chain of trust for sensitive workloads.
- Zerto Software integration enables real-time detection and rollback of agent actions.
- HPE Compute XD700 supports up to 128 Rubin GPUs per rack for dense on-premises GPU deployments.
The stack is designed to keep AI workloads safely behind corporate firewalls, a direct pitch at regulated sectors where hyperscaler-hosted agents remain a non-starter.
Potential risks and opportunities
Risks
- HPE ProLiant DL394 Gen12's early 2027 availability window gives cloud hyperscalers (AWS, Azure, Google Cloud) time to release competing on-premises or dedicated-instance agent offerings before HPE ships
- If Nvidia Vera CPU or Rubin GPU supply comes in below expectations, HPE enterprise customers already committed to Private Cloud AI roadmaps face infrastructure gaps with no near-term hardware alternative
- Enterprises deploying Zerto-based agent rollback without mature governance policies risk false confidence in oversight, potentially masking systemic agent errors before they compound across sensitive workloads
Opportunities
- Enterprise AI governance and compliance tooling vendors can position around HPE's secure local agent registration and model-vetting policy hooks as the on-premises agent governance layer matures
- Systems integrators specializing in regulated industries can build differentiated practices around the HPE Private Cloud AI Factory stack ahead of the 2027 hardware availability, locking in deployment contracts early
- Competitors offering on-premises AI infrastructure (Dell, Cisco UCS, Lenovo ThinkSystem) face pressure to match Confidential Computing and agent rollback as baseline features or risk losing regulated enterprise accounts to HPE
What we don't know yet
- Whether the Nvidia Agent Toolkit components (Nemotron models, OpenShell runtime) are available across existing servers now or tied to the early 2027 DL394 Gen12 timeline was not clarified
- Pricing or licensing terms for Nvidia Confidential Computing as a standard feature across the Private Cloud AI portfolio were not disclosed in the announcement
- No named enterprise customers or specific regulated verticals were identified as early adopters or design partners for the new agentic stack
Originally reported by siliconangle.com
Read the original article →Original headline: HPE Expands Private Cloud AI Factory With NVIDIA Vera CPUs and Full Agent Toolkit for On-Premises Agentic AI