HPE expands AI-native networking portfolio and outlines vision for self-driving IT operations
HPE used Discover Barcelona 2025 to lay out a bolder path for self-driving network operations across hybrid environments. The company is unifying AIOps features, hardware, and software across HPE Aruba Networking and HPE Juniper Networking, with tighter integration into GreenLake Intelligence.
This is one of the first major moves since HPE closed its Juniper Networks acquisition five months ago. The goal is straightforward: consistent operations, better visibility, and faster outcomes across compute, storage, networking, and cloud.
Why this matters for operations teams
Ops leaders are being asked to support AI-heavy workloads without adding headcount or complexity. HPE's approach leans on agentic AI to automate remediation, tighten root-cause analysis, and standardize workflows across previously separate platforms.
As Rami Rahim, executive vice president and general manager of Networking at HPE, put it: "In the era of AI, customers need networks that are purpose-built with AI and for AI to handle the rapid growth of connected devices, complex environments, and increasing security threats." The message is clear: build for autonomy, not just observability.
Unified AIOps: Aruba Central + Juniper Mist
HPE is blending capabilities between Aruba Networking Central and Juniper's Mist platform. The Mist Large Experience Model (LEM) will be available in Aruba Central, using application data from Zoom and Teams plus digital twin simulations to predict and improve user experiences.
Aruba's Agentic Mesh is being added to Mist to strengthen anomaly detection and root-cause analysis. Mist will also adopt organizational insights and NOC features from Aruba Central. For rollout flexibility, new Wi-Fi 7 access points will work across both management systems.
Aruba Central On-Premises: AIOps without the sprawl
The newest Aruba Networking Central On-Premises release introduces a redesigned interface and builds on existing AIOps capabilities. It brings intelligent remediation, actionable AI alerts, client insights, proactive monitoring, and a simpler documentation search into one environment.
For ops teams that need strict control and consistent SLOs, this helps reduce swivel-chair operations and shortens time to resolution.
New data center and edge hardware
HPE is expanding hardware for performance-heavy AI workloads, including inference at the edge. Two products to note:
- HPE Juniper Networking QFX5250 switch - Built on Broadcom Tomahawk 6 silicon, 102.4 Tbps bandwidth, liquid cooling, and AIOps-ready. Positioned as an Ultra Ethernet Transport-ready switch for GPU fabrics in AI data centers. Learn about Ultra Ethernet or see Broadcom's Tomahawk family here.
- Juniper Networking MX301 multiservice edge router - 1RU, 1.6 Tbps performance, 400G connectivity. Built for inference closer to data creation and suited for multiservice edge routing, enterprise deployments, metro networks, and mobile backhaul.
AI factory networking with NVIDIA and AMD
HPE outlined solutions with NVIDIA and AMD for high-speed, secure, low-latency connectivity between users, devices, and AI agents and for long-haul data center interconnect. The company reiterated its work with NVIDIA Spectrum-X Ethernet networking and BlueField-3 DPUs to support production AI. NVIDIA Spectrum-X
HPE also detailed collaboration with AMD on the "Helios" rack-scale architecture for trillion-parameter training, delivering 260 TB/s scale-up bandwidth and 2.9 exaflops FP4 performance. This includes a purpose-built HPE Juniper scale-up switch with Broadcom to accelerate AI training and inference using standards-based Ethernet.
Hybrid cloud AIOps: OpsRamp + GreenLake
HPE is advancing hybrid cloud operations with updates to HPE OpsRamp Software. The platform now unifies telemetry from HPE Compute Ops Management, Aruba Networking Central, and Juniper Networking Apstra for a single operational view.
New features include deeper GreenLake integration, full-stack observability via Apstra Data Center Director and Data Center Assurance, and predictive assurance to pinpoint issues faster. Operators will also see Compute Copilot, self-service root-cause analysis, OpsRamp integration for simpler troubleshooting, agentic root-causing tools, Model Context Protocol support for no-code integrations, and new AI agents in GreenLake Intelligence.
Financing options to speed adoption
HPE Financial Services introduced two programs: zero-percent financing for networking AIOps software purchased via term licenses, and the equivalent of 10% cash savings for leasing networking hardware built for AI workloads. There's an optional multi-OEM take-out service for retiring legacy gear.
Availability
- QFX5250 switch: Q1 2026
- MX301 router: December 2025
- OpsRamp integrations and related updates: rolling out from late 2025 through mid-2026
What operations leaders should do next
- Map current observability gaps across compute, storage, network, and cloud. Prioritize areas where agentic root-cause analysis could reduce MTTR.
- Decide on control-plane strategy: Aruba Central cloud vs. on-prem or Mist, given Wi-Fi 7 AP cross-compatibility.
- Plan GPU fabric readiness: evaluate Ultra Ethernet Transport, bandwidth targets (100 Tbps+), and cooling strategy for AI clusters.
- Align edge plans with inference needs: consider MX301 for metro and backhaul scenarios.
- Pilot GreenLake Intelligence agents and OpsRamp integrations to standardize runbooks and guided actions.
- Review financing options to sequence software first (AIOps) and phase hardware upgrades as workloads expand.
Skill up your team
If you're building an AIOps playbook or rolling out agent-driven operations, align team skills early. Explore role-based options here: AI courses by job.
Your membership also unlocks:
