HPE and NVIDIA announced at HPE Discover Las Vegas on June 16 a significant expansion of their joint AI infrastructure product, called HPE AI Factory with NVIDIA. The new features are aimed squarely at companies that want to move from running AI experiments to running AI agents at scale, inside their own facilities, without depending on public cloud services.
The centerpiece of the announcement is NVIDIA's new Vera processor. This is a chip built specifically to run AI agents: programs that operate on their own, taking actions across company systems, calling external tools, and processing data in real time. The Vera processor is designed to handle the kind of fast, sequential decision-making that agents require, which is different from the work involved in generating text or images. A new HPE server built around the Vera chip will be available in early 2027. The New York Stock Exchange is already testing it, citing its ability to handle more than 1.1 trillion messages per day while optimizing response times.
The software side matters more to most operators than the hardware. HPE's Private Cloud AI, which is a pre-built, on-premises AI system that companies can deploy without assembling parts themselves, will now ship with tools to monitor agent behavior, enforce rules about what agents are allowed to do, and detect when an agent goes off-script. A new capability built on HPE's Zerto software can detect a rogue agent and rewind the system to a clean state before that agent caused harm. That feature lands at exactly the right moment: a 2026 survey found that 35% of companies admit they could not immediately shut down a rogue AI agent if they needed to.
The governance gap is the real story here. Most enterprises are deploying agents far faster than they are building controls around them. A December 2025 survey of 750 IT executives found that 88% had already experienced a suspected security or privacy incident tied to AI agents. More than half of all deployed agents run without consistent security oversight or logging. Only 14% of agents go live with full IT and security approval. These are not small companies being careless; these are large organizations moving faster than their own processes allow.
For companies in regulated sectors such as banking, insurance, healthcare, or critical infrastructure, the "Confidential Computing" feature in this announcement is worth paying attention to. It means the data an AI agent processes and the AI model itself are encrypted even while they are actively running, which is a protection that previously did not exist. This matters for any company that handles patient data, trade data, or proprietary records and cannot afford to have that information exposed during processing.
New software features for HPE Private Cloud AI will arrive in July 2026. For companies in procurement, operations, or finance that are being pushed to adopt AI agents, the practical question is not whether to use agents but whether the controls around them are in place first. The EU AI Act's high-risk obligations take effect in August 2026, and regulators in multiple jurisdictions are increasingly looking at what oversight companies have over autonomous systems. The HPE and NVIDIA announcement is essentially a commercial answer to that pressure: a pre-packaged system with the governance layer already built in.