Hewlett Packard Enterprise (HPE) has taken advantage of its HPE Discover Barcelona 2025 event to send a clear message to the market: the networks of the future will be AI-native, autonomous, and designed to move massive volumes of data between users, data centers, and “AI factories.” Just five months after completing the acquisition of Juniper Networks, the company is already presenting a joint portfolio under the brands HPE Aruba Networking and HPE Juniper Networking, with a highly aggressive roadmap in networking, AIOps, and end-to-end observability.
The declared goal is ambitious: transform the network into the critical backbone of AI infrastructure and streamline operations in hybrid and multi-cloud environments, leveraging agent-based artificial intelligence and advanced automation.
Unified AIOps: Aruba Central and Mist begin speaking the same language
One of the key announcements is the accelerated integration between the two major network operations platforms of the group:
- HPE Aruba Networking Central
- HPE Juniper Networking Mist
HPE has created a common framework of microservices and agentic AI so that both platforms share advanced automation and diagnostics capabilities, preserving customer investments while offering a more consistent experience. Key highlights include:
- Mist Large Experience Model (LEM) arrives at Aruba Central:
This experience model, trained with billions of application data points such as Zoom or Microsoft Teams and enriched with synthetic data from “digital twins,” allows real-time detection, prediction, and remediation of video issues. It will now also be available to customers managing their networks with Aruba Central. - Agentic Mesh by Aruba debuts in Mist:
HPE Aruba’s agentic AI technology, focused on anomaly detection, root cause analysis, and autonomous or assisted actions, will be integrated into Mist to further enhance diagnostic and autonomous response capabilities. - More cohesive management experience:
Mist will adopt Aruba Central’s organizational view and global NOC-like dashboards, aiming to provide a unified control panel regardless of whether the customer operates HPE Aruba, HPE Juniper, or a combination of both. - New common Wi-Fi 7 access points:
HPE also announced new Wi-Fi 7 APs compatible with both Aruba Central and Mist, ensuring investment protection for customers wishing to upgrade their wireless network without committing to a single platform.
Additionally, HPE Aruba Networking Central On-Premises 3.0 incorporates advanced generative and traditional AIOps capabilities, smart alerts, proactive remediation, customer insights, and a redesigned user interface, while maintaining data and operation control within the customer’s own data center—crucial for environments with strong sovereignty or regulatory compliance requirements.
Tomahawk 6 Switch: networks for AI at 102.4 Tbps
The other major pillar of the announcement is the reinforcement of the AI networking portfolio. The company emphasizes that AI inference computing is shifting toward the edge due to latency, privacy, and cost considerations, requiring much more powerful and efficient switches and routers.
HPE introduces two strategic components:
- HPE Juniper Networking QFX5250
- First OEM switch leveraging Broadcom Tomahawk 6 silicon, with 102.4 Tbps bandwidth.
- Designed to connect GPUs within data centers and ready for Ultra Ethernet Transport.
- Combines Junos OS, HPE’s liquid cooling expertise, and AIOps capabilities to simplify the management of large AI clusters.
- HPE Juniper Networking MX301
- Compact 1U router with 1.6 Tbps capacity and 400G links, intended to bring inference closer to the network edge.
- Targeted at multiservice scenarios, urban networks, mobile backhaul, and high-performance enterprise routing.
Both devices align with HPE’s vision of “AI-native” networks: infrastructure prepared to carry AI traffic and to be managed internally using AI—from access to data center core.
Partnerships with NVIDIA and AMD for “AI factories”
HPE has also strengthened alliances with two key players in the AI ecosystem: NVIDIA and AMD.
On one side, it expands its AI factory networking solutions with:
- Extension of HPE Juniper Networking to the edge (edge on-ramp) and long-distance data center interconnection (DCI).
- Utilization of routing platforms MX and PTX to create high-scale, low-latency, secure links between users, devices, AI agents, and training/inference clusters, both on-premise and across multiple clouds.
- Integration with NVIDIA’s Ethernet platform Spectrum-X and DPUs NVIDIA BlueField-3 to enhance performance and workload experience in AI production.
On the other hand, the announcement highlights AMD’s “Helios” AI rack-scale architecture, which HPE positions as a turn-key rack capable of:
- Reaching 2.9 exaflops FP4 for training models with up to one trillion parameters.
- Providing 260 TB/s of scale-up bandwidth via Ethernet.
- Including a specialized HPE Juniper scale-up switch, developed with Broadcom, leveraging off-the-shelf Ethernet to maximize training and inference performance without proprietary technologies.
Toward a hybrid “command center” with agentic AIOps
Apart from hardware, HPE is using AI to unify observability and IT operations in hybrid environments. The company announced advancements in HPE OpsRamp Software and its integration with HPE GreenLake, creating a shared resource model spanning hardware to public clouds.
New capabilities include:
- Integration of HPE Juniper Networking Apstra Data Center Director and assurance tools with OpsRamp, delivering full-stack observability, predictive analytics, and proactive issue resolution across compute, storage, network, and cloud.
- Enhancements in Compute Ops Management, including integration with OpsRamp, Compute Copilot, and root cause analysis in self-service mode.
- Support for Model Context Protocol (MCP) and Agentic Root Causing to connect third-party AI agents without coding, enrich them with full-stack telemetry, and eliminate “blind spots” in highly dynamic environments.
- New AI agents in GreenLake Intelligence targeting sustainability, infrastructure health (Wellness Dashboard), and AIOps—designed to break silos and enable agentic analytics across the entire infrastructure.
Practically, HPE aims for IT operations to resemble a “hybrid control center,” where AI analyzes, correlates, and proposes (or executes) actions automatically, regardless of the underlying vendor.
0% financing to promote adoption
To ease the transition to AI-native networks, HPE Financial Services has launched two specific offers:
- 0% financing for networking software with AIOps, including licenses for HPE Juniper Networking Mist.
- An leasing program with a 10% cash equivalent savings for AI workload-focused network infrastructure (data centers and enterprise routing), with optional old equipment removal service and resale revenue sharing.
Availability
According to the company’s schedule:
- HPE Juniper Networking QFX5250: available in the first quarter of 2026.
- HPE Juniper Networking MX301: available in December 2025.
- Integrations of HPE OpsRamp and GreenLake:
- MCP: currently available to select customers; general release expected in early 2026.
- Compute Ops Management: available in December 2025.
- Storage Manager: February 2026.
- Apstra Data Center Director: Q2 2026.
This suite of announcements clearly shows that HPE’s strategy following the Juniper acquisition goes well beyond catalog expansion: it aims to redefine networking as a truly AI-native platform—autonomous, end-to-end observable, and tightly integrated with the broader hybrid infrastructure. A move designed to position itself as a leader in the AI network era.