LAS VEGAS – Hewlett Packard Enterprise (HPE) used this week’s Discover 2026 event to continue flooding the market with AI-angled updates and features, with the latest focused on Nvidia-based work around its AI Factory and Private Cloud AI platforms, moves targeted at accelerating enterprise returns on their AI-angled investments.

HPE’s AI Factory and Sovereign AI Factory platforms are based on Nvidia’s reference architectures and include Nvidia’s AI Enterprise software and HPE Unleash AI program partners. Those platforms are gaining immediate access to Nvidia’s RTX Pro Blackwell Server Edition graphic processing units (GPUs), Spectrum-X Ethernet, BlueField-3 digital processing units (DPUs), and ConnectX-8 SuperNICs (network interface cards).

The platform also gains integration of Nvidia’s Confidential Computing service that will be available through HPE Services. The Nvidia service, which was initially unveiled in mid-2023 and went public in early 2024, is designed to protect AI models and private data during execution for on-premises or sovereign deployments. This uses cryptographic attestation and encryption to establish a “train of trust” in securing hardware, software, and datasets.

HPE’s Private Cloud AI, which also runs an Nvidia architecture, gains support for secure local agent registration that allows customers to approve AI models, skills, and tools that match centralized governance and security policies, something increasingly important for sovereignty and edge-based AI use cases. It will also gain insight from HPE’s Alletra Storage MP X1000 that allows customers to automatically apply metadata and governance policies to prepare data for AI applications and boost token throughput, a notion that comes at a time of increasing token use concerns.

There is also a new unified model gateway for governed frontier model access, active workload prioritization, and multimode inferencing for up to 256 graphic GPUs. There can support the fine-tuning of pre-trained AI models, including Nvidia Nemotron open models to help control token costs.

The Private Cloud AI offering will also gain support for Nvidia’s Agent Toolkit and NemoClaw later this year, and next year be available on the vendor’s ProLiant Compute DL394 Gen12 hardware, a combination that will run Nvidia’s Vera central processing unit (CPU) as a “compute-optimized foundation” for agentic AI and data processing.

These Nvidia-focused updates come on the heels of HPE tying the chip giant’s Vera Rubin products into its rack-scale, compute, and Cray Supercomputing platforms, including that last one gaining Quantum-X800 InfiniBand support for the Cray GX5000.

HPE also recently announced it is broadening its AI factory solutions with NVIDIA Vera Rubin NVL72 by HPE rack-scale system, HPE Compute XD700 built on NVIDIA HGX Rubin NVL8, the HPE Cray Supercomputing GX240 Compute blade designed with NVIDIA Vera CPUs, and NVIDIA Quantum-X800 InfiniBand support for the HPE Cray Supercomputing GX5000.

Data fabric, Zerto gain features

Other new updates include HPE’s Data Fabric Software extending model context protocol (MCP) support to Apache Airflow and access to enterprise AI inventory that can inject metadata into distributed data sets to power agentic workflows. There is also a standalone Data Fabric appliance that runs on HPE’s ProLiant Compute servers that help simplify deployments.

HPE’s data protection Zerto Software will gain support for agent action monitoring and continuous data protection. This provides the ability for user to identify when agents go rogue and uses continuous data protection that can rewind data back when an anomaly occurs.

These updates follow a slew of new features HPE has rolled out over the past several months, with a big push coming late last year around its European-based Discover event and continued into this year with a constant cadence of new features and services.