HPE is one of the vendors raking in bushels of cash from AI hardware, so there will be little surprise that its latest showcase comprises updates to its AI Datacenter infrastructure portfolio, now infused with technology from its Juniper networking acquisition.

The IT systems and services giant will use its Discover event in Las Vegas to tell everyone how important it thinks agentic AI is going to be, and how organizations will need to have the right infrastructure to be ready for AI agents – a need it intends to meet, of course.

Ahead of the event, HPE told The Reg it now has a range of AI infrastructure options, covering turnkey, sovereign and large scale deployment models, including Private Cloud AI, AI Factory at-scale and Sovereign AI Factory.

"What we're looking at within HPE is to really help our customers with their transformation to an agentic AI infrastructure, and we're happy to announce that we have made significant progress around our AI factory portfolio," said EVP and CTO Fidelma Russo.

Private Cloud AI, the turnkey platform co-engineered with Nvidia, gains the latter’s Nemotron models, NemoClaw security stack and OpenShell secure runtime to provide an agent operating environment. It also gets HPE's Alletra Storage MP X10000 device, something that the firm signposted at its earlier Discover event in December.  These features will be available from July.

The jolly green giant's AI Factory at-scale and Sovereign AI Factory products now integrate Nvidia Confidential Computing to protect user models and private data while running workloads, available from Q4 2026. On the hardware side, they are also integrated to work with Nvidia's RTX PRO 6000 Blackwell Server Edition GPUs, Spectrum-X networking, BlueField-3 DPUs and ConnectX-NICs.

Russo said that HPE’s Data Fabric Software will soon have integration points with agentic workloads, which “allows agents to see across your enterprise data, whether it's in the cloud, whether it's in third party storage, and have automated workflows to allow your agents to access that data with all of the identity access management and security criteria that you would expect.”

The Texan titan was also keen to promote the integration of its Juniper networking technology into its AI Datacenter stacks, following its assimilation of the firm less than a year ago.

 “We're bringing our networking portfolio to HPE AI factories and AI infrastructure. We're accelerating innovation by bringing the best innovations from HPE MIST into HP Aruba Central and the other way around. We're connecting AI ops across the entire portfolio to create a simpler, more unified experience with the Marvis AI engine as this common thread, throughout all the different layers,” said Rami Rahim, HPE’s EVP and general manager for Networking.

What this means is that Marvis actions are now supported in Aruba Central, while CX switches can now be managed via Mist, as HPE continues to cross-pollinate capabilities from its two AIOps platforms.

HPE also revealed new hardware including the Juniper QFX5140, a 16 Tbps switch built on Trident5 and aimed at inference clusters and edge AI, and the QFX5252 Switch tray for AMD Helios, a scale-up module for AMD’s rack-scale AI platform.

HPE also says it has integrated SD-WAN and Security Service Edge (SSE) into a unified console, Networking EdgeConnect, for simpler management and consistent policy enforcement.

This being HPE, you can’t get away without a mention of Greenlake, the firm’s cloud-like IT services platform.

“We're bringing HPE Juniper datacenter networking into the Greenlake platform, and this is a significant milestone because it creates a unified cross-domain operating experience across compute, storage, and networking,” said Rahim.

“Customers can now manage infrastructure more consistently with simplified operations and accelerate deployment of modern AI and enterprise workloads across the entire stack.”

Summing up, Rahim said HPE is “clearly moving fast and making progress” in support of the agentic infrastructure layers, as well as the intelligent hybrid operations that are needed for its agentic AI vision. ®