Inside HPE's Agentic Enterprise Push: Siemens Energy Deal, QFX Switches, and Self-Driving Networks

For HPE, the Siemens Energy engagement is a marquee proof point that companies with strict data sovereignty and low-latency simulation requirements are choosing dedicated private clouds rather than relying solely on public hyperscalers.

HPE’s Bet on Networking: QFX Switches and an Expanded AI Portfolio

Networking dominated the Discover 2026 product barrage. HPE introduced two new HPE Juniper Networking QFX switches, integrating the Juniper stack directly into its AI Data Center Solutions to create what it calls "full-stack, AI-native infrastructure with proactive operations" .

• HPE Juniper Networking QFX5140 Switch — Purpose-built for AI inference clusters at the edge. The 1RU fixed-configuration switch delivers 16 Tbps total capacity and is designed for AI fabric, spine, leaf, and border leaf roles at the post-training stage . HPE networking president Rami Rahim called it “truly purpose built for the next wave of AI infrastructure with a special focus on inferencing clusters at the edge” .
• HPE Juniper Networking QFX5252 Switch Tray — A scale-up switch for AMD Helios supercomputing and AI training clusters. It offers 1024 Tbps total capacity, is 100% liquid-cooled, supports open standards and SONiC with AI-native operations, and runs Junos OS .

Both switches are now available as part of the HPE AI Factory, managed through the Data Center Director, signaling tighter integration between HPE’s compute and networking offerings .

Beyond the switches, HPE unveiled a broader networking arsenal:

• PTX 12,000 routing platform for high-speed data center interconnect .
• SRX 4700 quantum-safe firewall with quantum-resistant cryptography .
• MX 301 edge router to extend AI connectivity to the edge .
• Integration of Marvis Actions into HPE Aruba Networking Central and HPE Mist AIOps, using agentic reasoning to automate root-cause analysis and remediation .
• A unified AI-native SASE platform that brings together security and networking under one AI-driven fabric .

AI Factories and Private Cloud AI Get an Upgrade

HPE also updated its core AI infrastructure stack to support the shift toward agentic workloads. The HPE AI Factory portfolio is now integrated with Juniper networking and has been endorsed under the NVIDIA Cloud Partner program, positioning it for NVIDIA Cloud Provider certification to help cloud service providers streamline validation .

HPE Private Cloud AI received enhancements aimed at agentic AI, including capabilities that boost observability and efficiency for agentic deployments. The company also added the NVIDIA Agent Toolkit—including Nemotron open models, NemoClaw, and OpenShell—to the Private Cloud AI stack . HPE is now co-testing agentic AI workflows built on Accenture’s AI Refinery with NVIDIA, running on HPE Private Cloud AI, with initial use cases covering sourcing, procurement, and risk analysis .

On the hardware side, new GPU server options support NVIDIA Vera CPU and Blackwell Ultra architectures in the HPE ProLiant Compute XD685 with direct-liquid cooling .

The “Agentic Enterprise” Vision: AI Agents in Control

HPE CEO Antonio Neri used his keynote to frame all of these announcements under a single, ambitious banner: the agentic enterprise . In this model, AI agents—not just humans or traditional automation—drive operations across the enterprise. Networking becomes the foundational layer because it connects GPUs, inference clusters, edge devices, and users into a single AI-driven fabric.

The vision builds on HPE’s broader self-driving networking strategy, which extends AI-native autonomous operations from AI factories to data centers, campus, and branch locations . New AIOps features use agentic reasoning to shorten troubleshooting cycles, and self-healing automation in Aruba Central lets the network fix problems without manual intervention . In essence, HPE is betting that the enterprises that win in AI will be those whose infrastructure can manage, optimize, and remediate itself.

HPE’s Discover 2026 message was unmistakable: networking is no longer a supporting actor in AI infrastructure. It is the stage on which the agentic enterprise will be built.