A look at HPE's strategy with Juniper, networking for AI

HPE's recent networking advancements and the completion of its Juniper Networks acquisition aim to reshape networking strategies for AI.

The past few weeks have been big for networking at HPE.

HPE made some major networking announcements at its Discover 2025 conference in Las Vegas and days after it wrapped up, the US Department of Justice dropped its opposition to HPE's acquisition of Juniper Networks based on an agreed-upon settlement. HPE completed the acquisition a few days later. Combined, these developments have big implications for enterprises of all sizes that are worth understanding.

Networking updates from HPE Discover 2025

While the overall theme was focused on AI, HPE ratcheted up its emphasis on networking products and technology at the conference. In fact, networking was named as one of three essential elements for HPE's strategy, alongside AI software and hybrid cloud infrastructure.  

HPE CEO Antonio Neri said during his keynote that "the future of AI success depends on networking."  Recent research from Enterprise Strategy Group, now part of Omdia, has drawn similar conclusions, finding that 90% of organizations agree that networking has become more critical with the arrival of AI.

While we have long understood that networking would have to evolve to support GPU-heavy learning clusters in the data center, the implications are far broader. They touch campus, branch, and Wi-Fi as distributed inferencing and agentic AI move toward mainstream adoption, demanding both higher bandwidth and lower latencies.

HPE's technology updates reflected its moves to support networking for AI, including the CX 10040, a second generation smart switch with AMD Pensando DPUs onboard, two new CX 8325 configurations for data center and campus core, and four new models in the CX 6300 series of layer three stackable switches, all expanding throughput and performance to improve AI deployment success. There were also solid steps forward on the Wi-Fi 7 AP front as well as a new integrated platform for private 5G deployments.

Perhaps the most consequential announcements were AI for networking, and particularly the new Agentic AI Mesh architecture for HPE Aruba Networking.  The first phase of the Agentic AI Mesh will include 15 purpose-built agents as well as a super-agent for orchestration and is primarily focused on faster and more efficient operations as part of HPE Aruba Central network management.   

This agentic mesh will also be used to tie the network into HPE's new GreenLake Intelligence offering, which is taking on the perennially thorny challenge of effective cross-domain infrastructure management spanning compute, storage and networking. The addition of OpsRamp integration into the HPE management stack is key for consistent multi-vendor operational monitoring.

HPE completes Juniper acquisition

Hot on the heels of HPE Discover, the long-anticipated acquisition of Juniper Networks suddenly and dramatically closed. The acquisition plan was first announced in January 2024, but the US Department of Justice opposed it on the grounds that it would limit competition in Wi-Fi networking. That all changed quickly when the US DoJ dropped its opposition to the deal, having reached a negotiated settlement with HPE.

When the deal was first announced, an in-depth analysis of the implications of this combination was released, which is still largely valid. Essentially, the two together will have a richer set of offerings for data center, routing, branch, campus, and Wi-Fi networking, along with one of the most mature AI for networking software, Marvis AI.

But with 18 months between announcement and closure, there are a couple of important changes in the market demand and supply landscape that affect the relevance of the acquisition now:

  • The focus of networking for AI planning and strategy has moved beyond the data center and is turning toward campus, branch, and edge, to prepare for remote inferencing and distributed agentic AI applications; 
  • The MIST Marvis AI "moat" that Juniper established is no longer as wide as it once was. Just 18 months ago, Juniper had a significant advantage versus all its competitors, HPE included, on the AI for networking front, but the arrival of agentic AI has radically changed that picture. Juniper is now actually behind other networking vendors, including HPE, Cisco, and Extreme, in building agentic AI into the platform. Combining Juniper Marvis AI with the Agentic AI Mesh of HPE brings the combined organization right back to the leading edge.
  • HPE Aruba was the first major networking vendor to embed DPUs in switches, and at this year's Discover, launched the second generation of products. With the release of Cisco's Smart Switches and Routers earlier this year, HPE is no longer unique in this offering. This will need to be a center point of strategy, as smart switching holds important promises for both AI deployments and the networking and security convergence that is fast becoming a hard requirement for enterprises.

The road ahead for HPE networking

Post-acquisition, the new organization will be called HPE Networking and will be comprised of two teams:  HPE Aruba Networking and HPE Juniper Networking. The hard work now begins in merging products, technologies, roadmaps, and teams. The scope of HPE's networking products is greatly expanded, but product lines and architectures will have to be rationalized, particularly for campus and branch. HPE's challenge is to execute without disrupting customer and partner relationships to the point that eyes start to wander toward alternatives.

There is a tremendous amount of detail that still needs to be filled in, but the promised endpoint will be a new HPE Networking that can offer more complete technologies and will be more competitive on a global scale. We'll be keeping close tabs on progress.

Jim Frey covers networking as principal analyst at Enterprise Strategy Group, now part of Omdia.

Enterprise Strategy Group analysts have business relationships with technology providers.

Dig Deeper on Network infrastructure