Hewlett Packard Enterprise Co. is using its Discover Barcelona 2025 event today to announce a broad expansion of its networking, hybrid cloud and data infrastructure portfolio, highlighting early integration milestones following its recent acquisition of Juniper Networks Inc.

The company detailed new switches and routers for artificial intelligence workloads, expanded AI operations capabilities and updates across its GreenLake and Morpheus platforms. Executives said the announcements are intended to align compute, storage and networking under a consistent architecture for enterprise and cloud environments.

HPE said the networking updates reflect progress made only five months after closing the Juniper acquisition. The company introduced shared AIOps features for HPE Aruba Networking Central and HPE Juniper Networking Mist and described plans to unify user experiences across both platforms. The company said the goal is a consistent operational model across cloud and on-premises deployments.

Rami Rahim, executive vice president and general manager of HPE Networking and the former chief executive officer of Juniper Networks, said the combined portfolio reflects changes in how AI clusters are built and connected. “In the front end, this connects clusters to users, to applications and microservices,” he said. “In the back end, we’re connecting [graphic processing units] across racks.”

Rahim said InfiniBand, the high-performance networking standard that uses a switched fabric to provide high throughput and low latency, “is gradually migrating to Ethernet,” and that this shift has created new opportunities for Ethernet-based products.

“InfiniBand will be around for a while,” he said. “That does not change the fact that Ethernet is a huge and growing opportunity.”

Rahim said the distinction between the Aruba Central and Juniper Mist platforms will diminish over time. “It’s going to disappear because the experience to the end customer is going to be identical irrespective of which platform you’re on,” he said.

Networking hardware

HPE introduced multiple hardware platforms aimed at scale-out and edge routing for AI clusters. The HPE Juniper Networking QFX5250 switch uses Broadcom Inc.’s Tomahawk 6 silicon and provides 102.4 terabits per second of bandwidth. The company said the switch is liquid-cooled and ready to support Ultra Ethernet Transport, a new protocol designed for high-performance, scalable data center networks. Rahim called it “the world’s highest-performance, 100% liquid-cooled, Ultra Ethernet transport-ready switch built on the Broadcom Tomahawk 6 processor.”

HPE also announced the MX301 multiservice edge router, positioned as a compact on-ramp for distributed inference clusters. The single rack unit router “packs all of the performance, 1.6 terabits per second, and all of the flexibility that our customers have come to love about DMX,” Rahim said.

The company said it is also developing an Ethernet-based scale-up switch for Advanced Micro Devices Inc.’s Helios rack-scale AI reference design. Rahim said the switch is “an industry-first scale-up solution using standard Ethernet” and is intended to support open standards.

Expanded partnerships

HPE said it has extended partnerships with Nvidia Corp. and AMD to support AI factories, long-haul data center interconnects and rack-scale systems. The company said the joint work with Nvidia now includes routing layers using the MX and PTX high-speed routing platforms. “These joint solutions will give our customers the assurance they need to deploy our routing technology alongside Nvidia’s cutting-edge products with full confidence,” Rahim said.

He said Helios is “the first rack-scale AI reference design from AMD, built fully on the open rack-wide standard,” an open-source data center rack standard specifically managed by the Open Compute Project and meant to meet the extreme power, cooling and serviceability demands of AI systems.

New AIOps features are designed to unify management across compute, storage, networking and cloud environments. The company said HPE OpsRamp integrations now bring together telemetry from its Compute Ops Management cloud-based server management console, Aruba Networking Central and Juniper Apstra networking management platform into a single view.

HPE also added Model Context Protocol support to GreenLake and OpsRamp to allow third-party AI agents to interpret context drawn from OpsRamp’s multi-vendor integrations. Additional updates include a new Compute Copilot interface for ProLiant systems. “You now have a contextual text-based interface that allows you to ask questions and get recommendations,” said Fidelma Russo, HPE’s chief technology officer.

AI-ready storage

The new HPE Alletra Storage MP X10000 Data Intelligence Node, built on an Nvidia AI Data Platform reference design, enriches and structures data as it enters storage. “It does metadata tagging, embedded vector generation, and vector-ready formatting, and all of these happen automatically,” Russo said. The company said the system aims to accelerate data preparation for GPU workloads.

HPE also added two new models to its StoreOnce backup and data protection systems. The all-flash StoreOnce 7700 delivers up to 300 terabytes per hour for data ingest. “We’ve redesigned the entire I/O path to optimize for flash,” Russo said, noting that modern recovery workloads increasingly involve random access patterns. The hybrid StoreOnce 5720 targets broader enterprise workloads.

Expanded virtualization capabilities are supported by additions to the Morpheus hybrid cloud management platform. Morpheus VM Essentials and Morpheus Enterprise Software. Updates include software-defined networking, micro-segmentation, stretched clusters, continuous data protection with HPE Zerto and image-based backup using software from Veeam Software Corp. The company said these features are intended to provide a full alternative for customers evaluating virtualization alternatives.

Morpheus now also integrates with Juniper Apstra to automate switch configurations and maintain consistent VLAN and security policies. The company said the update is intended to eliminate configuration errors when virtual machines move across environments.

Performance and security enhancements to HPE’s Private Cloud AI platform include Nvidia RTX Pro 6000 GPUs across configurations, hardened Nvidia Inference Microservices, air-gapped deployment options and GPU fractionalization using Nvidia Multi-Instance GPU. Russo said the configuration is “ideal for mixed HPC and AI workloads” and supports systems based on the ARM reduced instruction set computing architecture.

Pricing and availability

HPE Financial Services will provide zero-percent financing for networking AIOps software and additional programs that provide the equivalent of 10% savings for customers leasing networking systems built for AI workloads. Availability timelines are early 2026 for full OpsRamp MCP support, January 2026 for the X10000 Data Intelligence Node and spring 2026 for StoreOnce 5720 and 7700.

Rahim said HPE will measure the success of the Juniper acquisition over the next year by evaluating integration progress and market growth. “Success means that we bring the two businesses together without in any way disrupting our customers,” he said. He said the combined companies aim to grow faster than competitors across routing, campus and branch networking, data center solutions and security.

Photo: HPE