Enterprises are transferring agentic AI from proof of idea to manufacturing — and the subsequent era of AI factories are constructed for the period of brokers.
At HPE Uncover Las Vegas, working by means of Thursday, June 18, NVIDIA and HPE are increasing the HPE AI Factory with NVIDIA, together with NVIDIA Vera CPU and NVIDIA Agent Toolkit for HPE Non-public Cloud AI.
Plus, NVIDIA Confidential Computing extends throughout HPE AI Factory and enhanced full-stack NVIDIA integration — with NVIDIA accelerated computing, NVIDIA AI software program and NVIDIA networking — is offered all through the total portfolio.
NVIDIA Vera CPU Out there With HPE Non-public Cloud AI
The HPE ProLiant Compute DL394 Gen12 with the NVIDIA Vera CPU shall be out there in 2027 with HPE Non-public Cloud AI, a turnkey AI manufacturing facility co-engineered with NVIDIA. Vera is the first CPU constructed for brokers — designed for the software calls, orchestration and real-time knowledge processing required throughout the agent loop — bringing deterministic, low-latency efficiency into HPE Non-public Cloud AI.
The New York Inventory Change, in collaboration with Redpanda and HPE, is an early enterprise buyer exploring Vera CPU with the HPE ProLiant Compute DL394 Gen12 server.
The Vera CPU is an element of the NVIDIA Vera Rubin platform, which is ramping into full manufacturing with the NVIDIA Vera Rubin NVL72 rack-scale system out there from HPE. Vera Rubin was constructed for frontier-scale fashions bigger than 1 trillion parameters and can ship with full-stack NVIDIA Confidential Computing throughout each chip.
HPE can also be bringing the HPE Compute XD700 — constructed on NVIDIA HGX Rubin NVL8 — to the HPE AI Factory, supporting as much as 128 Rubin GPUs per rack.
NVIDIA Agent Toolkit Now Out there With HPE Non-public Cloud AI
NVIDIA Agent Toolkit — together with NVIDIA Nemotron open fashions, the NVIDIA OpenShell safe runtime and NVIDIA NemoClaw blueprints — shall be out there with HPE Non-public Cloud AI. Collectively, they provide enterprises an agentic AI working system for monitoring agent conduct, implementing governance insurance policies, and safely constructing and working autonomous, long-running multi-agent techniques.
HPE Non-public Cloud AI provides safe native agent registration, letting clients approve AI fashions, expertise and instruments towards centralized governance and safety insurance policies earlier than they run. New HPE Zerto Software program capabilities detect rogue agent actions and use steady knowledge safety to rewind to a clear state.
On the knowledge aspect, HPE Alletra Storage MP X10000 — which achieved the basis stage of NVIDIA-Licensed Storage — mechanically applies metadata and governance insurance policies to arrange unstructured knowledge for AI pipelines, enhancing token throughput.
NVIDIA Confidential Computing Throughout All HPE AI Factory Options
NVIDIA Confidential Computing is now out there throughout the HPE AI Factory by means of HPE Companies — together with HPE AI Factory at Scale, HPE Sovereign AI Factory and HPE Non-public Cloud AI.
AI functions entry and use personal and delicate knowledge that must be protected and secured. As well as, fashions skilled with proprietary knowledge or methods have to be safeguarded from exfiltration. Confidential computing is important for these fashionable AI workloads, because it protects fashions and personal knowledge throughout execution for on-premises and sovereign deployments, establishing a series of belief by means of cryptographic attestation and encryption at each stage.
As well as, HPE ProLiant Compute DL380a achieved certification as half of the NVIDIA-Licensed Programs for NVIDIA Confidential Computing program, which validates strong utility efficiency with confidential computing. These techniques present hardware-based safety for AI workloads and delicate knowledge property whereas sustaining optimum NVIDIA acceleration.
Throughout the HPE AI Factory options, NVIDIA BlueField DPUs and NVIDIA DOCA present in-silicon zero-trust coverage enforcement, runtime menace detection and community encryption — defending AI workloads, brokers and knowledge with out efficiency tradeoffs.
Enhanced Full-Stack NVIDIA Integration Throughout the Portfolio
HPE AI Factory at Scale, HPE Sovereign AI Factory and HPE Non-public Cloud AI are actually out there with NVIDIA RTX PRO 6000 Blackwell Server Version GPUs, NVIDIA Spectrum-X Ethernet networking, NVIDIA BlueField-3 DPUs and NVIDIA ConnectX-8 SuperNICs.
For next-generation AI factories, each Vera Rubin NVL72 system will ship with NVIDIA networking inbuilt — NVIDIA Vera BlueField-4 DPUs, NVIDIA ConnectX-9 SuperNICs and NVIDIA Spectrum-X Ethernet — with NVIDIA Spectrum-6 switching delivering 1.6x larger networking efficiency for AI communication versus off-the-shelf Ethernet.
Spectrum-X Ethernet networking is the normal for HPE AI Factory with NVIDIA — together with at-scale, sovereign and turnkey AI manufacturing facility options out there now. For giant-scale and sovereign workloads, HPE introduced at NVIDIA GTC in March that it’s additionally including NVIDIA InfiniBand networking choices — together with NVIDIA Quantum-X800 InfiniBand with the HPE Cray Supercomputing GX5000.
These configurations are primarily based on NVIDIA reference architectures and help use instances from AI improvement by means of production-scale deployment, with NVIDIA AI Enterprise software program and the HPE Unleash AI ecosystem.
At HPE Uncover this week, the Unleash AI accomplice program is increasing with almost a dozen new AI software program companions — together with Aizen, BridgeTEK, deepset, Deliverance, Faclon Labs, Gallop, Rocket, Supervity, Thales, Trustwise and Vortiqx.
Attendees can discover these options all week at the present and be taught extra about the HPE AI Factory with NVIDIA, half of the NVIDIA AI Computing by HPE portfolio.
Source link
#HPE #Factory #NVIDIA #Expands #Era #Agents
