HPE has expanded its HPE AI Factory with NVIDIA portfolio with new tools so enterprises can deploy AI agents securely and under governance, including MCP APIs and servers in HPE Private Cloud AI, a rack with 256 NVIDIA Vera CPUs and support for the NVIDIA OpenShell architecture. The company announced it on October 21, 2026, at GTC Berlin, together with an agreement with Schneider Electric to bring industrial AI agents to the edge.
HPE and NVIDIA’s agentic AI push in 20 seconds
- HPE Private Cloud AI adds governed APIs and MCP servers for agents.
- It already supports mixing H200 and RTX PRO 6000 Blackwell GPUs in a single cluster.
- A new rack packs 256 Vera CPUs for agentic workloads.
- Schneider Electric will bring industrial AI agents to the edge with HPE.
HPE’s message is that enterprises want to move AI agents into production without losing control. Cheri Williams, senior vice president and general manager of private cloud and Flex solutions at HPE, said agents are becoming an essential part of how enterprises operate and that customers need control over security, sovereignty, and data access to move from experimentation to production.
Private Cloud AI Opens Up to Agents and Developers
The biggest software news lands in HPE AI Essentials, within Private Cloud AI. HPE is adding production-grade APIs and MCP servers so that applications, developers, software vendors, and agents can access inference services, models, and knowledge under centralized governance. The MCP servers are versioned and controlled by role, tenant, and project, with auditing. Interfaces are defined with OpenAPI, which generates documentation, developer kits, and code samples, and access control is based on roles, machine identities, and audit logs for both people and autonomous agents. Everything also works in isolated, offline deployments designed for regulated and sovereign environments.
These capabilities will arrive by the end of 2026, and a self-service experience for developers to register applications and agents, discover APIs and MCP services, and manage credentials is planned for 2027. Already available is the ability to mix NVIDIA H200 and RTX PRO 6000 Blackwell Server Edition GPUs in the same cluster, along with NVIDIA’s open-source KAI Scheduler, which allocates GPUs by quota and priority across projects.
A Rack With Vera CPUs, OpenShell, and DSX POD
On the infrastructure side, HPE is introducing the NVIDIA Vera CPU Rack by HPE, a liquid-cooled system with 256 NVIDIA Vera CPUs built for agentic and reinforcement-learning workloads that depend on the CPU. According to the announcement, it supports 22,500 sandboxes, the isolated environments where agents run code, and delivers 80% more performance in CPU-bound agentic workloads, without specifying which system it’s compared against. HPE already brought the NVIDIA Vera CPU to ProLiant back in June.
HPE will also support the NVIDIA OpenShell reference architecture for secure agents, which combines the OpenShell execution environment on Vera CPUs, NVIDIA Sentry with BlueField-4 to enforce policies in hardware, and Vera Rubin NVL72 for inference. It will also deliver AI factories based on NVIDIA DSX POD, NVIDIA’s design for modular, containerized data centers, through its AI Mod Pod services. And it is extending NVIDIA’s confidential computing to its ProLiant Compute DL380a Gen12 servers with RTX PRO 6000 Blackwell GPUs, to protect models and data while they are being processed on sovereign infrastructure.
Schneider Electric Brings Agents to the Factory Floor
Schneider Electric is the only partner company named in the announcement. The French company will combine HPE Private Cloud AI with its industrial and operational technology expertise to develop industrial AI agents at the edge, so decisions get made closer to the operation and sensitive data stays under control. The announcement gives no dates or details on the first deployments of that collaboration, nor on the availability of the Vera rack.
Frequently Asked Questions
What did HPE announce with NVIDIA at GTC Berlin?
APIs and MCP servers for agents in HPE Private Cloud AI, a rack with 256 NVIDIA Vera CPUs, support for OpenShell and DSX POD, and expanded confidential computing.
When are the new HPE Private Cloud AI features coming?
The APIs and MCP servers will arrive by the end of 2026, and the self-service experience for developers in 2027. Mixed GPU support is already available.
What is the NVIDIA Vera CPU Rack by HPE?
A liquid-cooled rack with 256 Vera CPUs for agentic and reinforcement-learning workloads, with capacity for 22,500 sandboxes.
What will Schneider Electric do with HPE?
It will combine HPE Private Cloud AI with its industrial expertise to develop AI agents at the edge.
Sources:
