HPE AI Factory With NVIDIA Expands for the Era of Agents
18:30 · June 16, 2026 · NVIDIA
Enterprises are moving agentic AI from proof of concept to production — and the next generation of AI factories are built for the era of agents. At HPE Discover Las Vegas, running through Thursday, June 18, NVIDIA and HPE are expanding the HPE AI Factory with NVIDIA, including NVIDIA Vera CPU and NVIDIA Agent Toolkit […]
Summary
NVIDIA and HPE are extending their joint AI Factory offering to address the shift of agentic AI workloads from proof-of-concept experiments to production deployments. The updated portfolio centers on HPE Private Cloud AI and introduces the NVIDIA Vera CPU, the NVIDIA Agent Toolkit, and broader deployment of NVIDIA Confidential Computing across HPE’s AI systems. These components target the low-latency, deterministic execution and governance requirements of autonomous, multi-agent systems that interact with tools and maintain state over extended periods.
The Vera CPU, scheduled for availability in 2027 inside the HPE ProLiant Compute DL394 Gen12 server, forms part of the Vera Rubin platform. It is designed specifically for the agent loop, handling tool invocation, orchestration, and real-time data movement while delivering predictable performance inside HPE Private Cloud AI. The same platform includes the HPE Compute XD700, based on NVIDIA HGX Rubin NVL8, supporting up to 128 Rubin GPUs per rack, and the rack-scale Vera Rubin NVL72 system. Early evaluation of the Vera CPU configuration is under way at the New York Stock Exchange in collaboration with Redpanda.
Software support arrives through the NVIDIA Agent Toolkit, which bundles Nemotron open models, the OpenShell secure runtime, and NemoClaw blueprints. When integrated with HPE Private Cloud AI, the toolkit supplies monitoring, policy enforcement, and secure execution for long-running multi-agent applications. HPE adds local agent registration and Zerto-based continuous data protection to detect and recover from unauthorized agent actions. On the storage side, the HPE Alletra Storage MP X10000 applies metadata and governance policies to unstructured data, raising token throughput for AI pipelines.
Confidential Computing now spans the full HPE AI Factory portfolio, including at-scale, sovereign, and private-cloud deployments. Hardware-based isolation protects both proprietary models and sensitive data during execution, supported by cryptographic attestation and encryption at every stage. Certified systems such as the HPE ProLiant Compute DL380a maintain NVIDIA acceleration while enforcing these protections. Complementary network-level controls are provided by BlueField DPUs and DOCA software, delivering in-silicon zero-trust enforcement and runtime threat detection without throughput penalties. Spectrum-X Ethernet remains the standard interconnect, with InfiniBand options added for large-scale and sovereign configurations.
Why it matters
This article highlights key advancements in enterprise AI infrastructure, specifically focusing on secure, agentic AI and confidential computing. While highly technical, the emphasis on data security and governance aligns well with European and Dutch priorities for safe, compliant AI deployment.
