Volume 343, Issue 3 · 7 articles from 2 issues
At OCP Global Summit, HPE makes the case for open, standards-based, heterogeneous AI infrastructure.
Writing for the Open Compute Project Global Summit, HPE outlines its strategy for future AI across compute, networking and large-scale infrastructure, expecting AI to run as a service across edge, cloud and sovereign environments as workloads shift from training to inference and from generative to agentic AI.
Different workloads will need a diverse mix of accelerators and processors, and the ecosystem of neoclouds, colocation providers and sovereign programs must interoperate.
HPE argues an open compute, standards-based approach is essential.
HPE argues AI resilience must go beyond availability to cover models, agents, data pipelines and infrastructure together.
HPE says resilience has traditionally been measured by availability and recovery time, but AI raises the bar, since a model can stay online and appear healthy even after the context it depends on is disrupted.
Because models, agents, data pipelines, vector databases and inference services all affect outcomes, a disruption may not look like a traditional outage.
The post outlines what customers should demand from providers beyond uptime.
HPE expands HPE AI Factory with NVIDIA with agentic AI capabilities, NVIDIA Vera CPU racks and secure agent architecture support.
At NVIDIA GTC Berlin 2026, HPE announced innovations across the HPE AI Factory with NVIDIA portfolio to help enterprises securely deploy and operate autonomous AI across hybrid and sovereign environments.
HPE Private Cloud AI gains expanded agentic AI capabilities, governed agent integration and more GPU flexibility.
New infrastructure includes the NVIDIA Vera CPU Rack by HPE and modular designs, and HPE will support the NVIDIA OpenShell Reference Architecture for Safe Agents with NVIDIA Vera Rubin, Vera CPUs and BlueField-4 for full-stack governance and isolation.
Tech Field Day podcast: HPE Networking on alert fatigue, microsegmentation and AI-driven automation that keeps engineers in control.
In this Tech Field Day Podcast episode from Networking Field Day 41, Tom Hollingsworth and Michael Balasko talk with Yash Nagaraju, director of technical marketing engineering at HPE Networking, about why securing enterprise networks has become too much for human operators alone.
They cover alert fatigue, sprawling microsegmentation demands and agentic AI threats, and why high-fidelity telemetry is critical.
Nagaraju explains how HPE Mist and Central use large experience models to enable self-driving actions without taking control away from network engineers.
The episode is presented by HPE Networking.
HPE Spaceborne Computer-4 will run the first HPC benchmarks and AI workloads on the lunar surface aboard Astrolab's FLIP rover.
HPE announced that HPE Spaceborne Computer-4, a compute module built for space, plans to power AI and the first high performance computing benchmarks on the surface of the Moon.
Working with NASA and Deloitte, the system will test LLM retrieval-augmented generation, vision inference and LiDAR data processing.
It will ride on Astrolab's FLIP rover, delivered by Voyager Technologies' Griffin-1 lander near the lunar South Pole as part of NASA's Commercial Lunar Payload Services initiative, with launch targeted for late 2026.
HPE outlines how enterprises can respond as AI shrinks the window between vulnerability discovery and exploitation.
HPE says AI is collapsing the gap between finding a software flaw and exploiting it, pointing to Anthropic's Claude Mythos finding a decades-old OpenBSD bug and thousands of other unknown vulnerabilities.
It argues mature organizations respond by building security into every stage of development through threat analysis, code reviews, static analysis and penetration testing.
HPE also calls for continuous vulnerability management and urges action on regulation, confidential computing and quantum risk so enterprise security can operate at machine speed.
HPE argues private AI security depends on managing the platform lifecycle, not just secure deployment.
HPE says most organizations evaluate AI platform security like buying a new car, checking certifications, benchmarks and references, then assuming the job is done.
Attackers, however, operate in the months and years after deployment, exploiting slipped patches, drifting dependencies and configurations that become exposures.
HPE argues the gap is widest in AI, where agents increasingly take consequential actions and attackers use the same technology to move faster.
It says private AI gives enterprises more control but requires them to manage the platform lifecycle, and urges buyers to evaluate how reliably platforms maintain security over time.