HPE today introduced new AI products and features with NVIDIA designed to help organizations put agentic AI into production, adding security, governance and control as companies move beyond pilots to autonomous systems at scale. The updates, announced on June 17, 2026, centre on HPE Private Cloud AI and the HPE AI Factory with NVIDIA, targeting the operational demands of AI agents that act with a degree of independence.
Agentic AI refers to systems that can make decisions and take actions rather than simply responding to prompts, a trend that is driving new approaches to AI Agents & Automation. Customers are seeking ways to deploy these systems with stronger oversight, particularly when they handle sensitive data and support business processes at scale. The new package includes updates to computing, storage, data management and security software, along with fresh NVIDIA hardware and software integrations.
Agentic AI with built-in controls
At the centre of the announcement is HPE Private Cloud AI, a turnkey AI factory product developed with NVIDIA. It now supports NVIDIA Agent Toolkit software, including NVIDIA Nemotron open models, NVIDIA NemoClaw and the NVIDIA OpenShell secure runtime. These tools help customers monitor agent behaviour, apply policies and limit deployment risks. The product is also adding support for HPE ProLiant Compute DL394 Gen12 with NVIDIA Vera CPU, providing a compute base for agentic AI workloads and data processing with built-in security and management features.
HPE Zerto Software gains new functions to detect rogue agent actions and use continuous data protection to restore systems to an earlier clean state. Local registration for AI agents is also being added, giving customers the option to approve models, tools and skills under central governance and security policies. This approach addresses concerns about how AI agents are introduced into production and how their actions are supervised once deployed.
"As AI becomes more autonomous, organisations need a new architecture to run it securely, govern it responsibly, and scale it economically," said Antonio Neri, President and Chief Executive Officer, HPE. "Across networking, servers, storage and software, HPE is delivering full-stack AI solutions with NVIDIA that build the foundation for agentic enterprises, helping customers move from experimentation to production with control and confidence."
"Every layer of the computing stack is being reinvented for the age of AI agents," said Jensen Huang, Founder and Chief Executive Officer, NVIDIA. "Together with HPE, we are building AI factories for this new era of computing - powered by NVIDIA Vera CPUs, accelerated infrastructure, and secure AI software - to help enterprises transform their data into intelligent action."
Data pipelines and inference efficiency
HPE also highlighted data preparation and the cost of inference. HPE Private Cloud AI can turn unstructured data into pipelines for AI use and improve efficiency in how prompts and tokens are processed. Built-in intelligence in HPE Alletra Storage MP X10000 can automatically apply metadata and governance policies to prepare data for AI applications, cutting token response times by up to 20 times and increasing token throughput by up to 20%.
HPE Data Fabric Software is being expanded to support agentic workflows. Support for model context protocol is being extended to Apache Airflow, and an enterprise AI inventory is being added to enrich distributed data with metadata. A standalone HPE Data Fabric appliance on HPE ProLiant Compute servers is intended to simplify deployment. The updates also include a unified model gateway for governed access to frontier models, workload prioritisation and multi-node inferencing for systems of up to 256 GPUs. Customers will also be able to fine-tune pre-trained models, including NVIDIA Nemotron open models, through NVIDIA NeMo with secure access to enterprise data.
Sovereign AI and confidential computing
Beyond private cloud deployments, HPE is extending the HPE AI Factory at-scale and HPE Sovereign AI Factory with added security functions. A key part of that is NVIDIA Confidential Computing, which HPE plans to integrate through HPE Services for on-premises and sovereign deployments. The confidential computing approach protects models and private data during execution through cryptographic attestation and encryption across hardware, software and datasets, supporting compliance with regional and industry standards.
Zero-trust controls are also part of the package. HPE said NVIDIA BlueField and NVIDIA DOCA will be used across the AI Factory to provide policy enforcement, runtime threat detection and encryption for networking tied to AI workloads, agents and data. The AI Factory offerings will be available with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, NVIDIA Spectrum-X Ethernet, NVIDIA BlueField-3 DPUs and NVIDIA ConnectX-8 SuperNICs, based on NVIDIA reference architectures.
Why this matters for product development
For product development teams, the new capabilities mean tighter integration of governance and security into the AI lifecycle, a shift explored in resources on AI for Product Development. The HPE updates address three practical concerns: governing agent behaviour, securing data pipelines, and managing inference costs. The local registration and policy enforcement tools let teams approve which models and actions an agent can take, reducing the risk of unexpected outcomes. The data fabric and storage improvements cut token response times and speed up data preparation, which directly affects the responsiveness of AI-powered features in end products.
"The shift to agentic AI is becoming a priority for organisations across Australia and New Zealand as they look to unlock greater value from their data and automate more complex processes," said Chris Weber, Vice President and Managing Director, South Pacific, HPE. "HPE's AI factory approach enables customers to operationalise agentic AI in a way that is scalable, efficient and ready for enterprise use."
For teams building AI-driven products, the ability to fine-tune models securely and manage multi-node inference up to 256 GPUs opens the door to larger-scale deployments without losing control over data or costs. The confidential computing and zero-trust features also provide a path to meet compliance requirements when handling customer data in sovereign regions.
Your membership also unlocks:
