HPE Targets 50%+ Data Center Networking Growth as AI Orders Top $3 Billion

HPE is setting aggressive growth targets for its Networking business as the company moves the former Juniper portfolio deeper into AI data centers, projecting low-to-high 50% compound annual growth for Data Center Networking from fiscal 2026 through fiscal 2029 and raising its cumulative Networks for AI order outlook to more than $3 billion. Speaking at HPE Networking Investor Day, Rami Rahim, executive vice president and president and general manager of Networking at HPE, positioned AI infrastructure as the fastest-growing part of the combined networking portfolio, spanning Scale Up accelerator connectivity, Scale Out data center fabrics, AI on-ramps and Scale Across data center interconnect.

HPE now expects overall Networking revenue to grow at a high-teens CAGR from FY26 through FY29, while Routing is projected to grow at a low-to-high 20% CAGR and Campus & Branch and Security at high-single-digit rates. For FY27 specifically, HPE raised its Networking revenue growth outlook to the high-teens to low-20% range, up from the 14%-17% framework provided with its Q3 results. Operating margin is expected to increase from the low 20% range in FY26 to the mid-to-high 20% range in FY27. HPE said Q3 networking orders were growing 3.5 times faster than revenue and that it doubled networking supply purchase commitments during the quarter to improve order conversion as it enters FY27. The Investor Day presentation places HPE’s long-term Networking revenue CAGR in the high teens with operating margins remaining in the mid-to-high 20% range from FY27 through FY29.

The central technical shift is HPE’s expansion beyond conventional Ethernet switching into rack-scale AI networking. HPE’s architecture now spans QFX switching for Scale Out, QFX-based networking for the AMD Helios Scale Up fabric, PTX routing for Scale Across/DCI and MX routing for AI on-ramp and data center edge connectivity. HPE is also carrying Juniper’s Trio and Express silicon into the AI infrastructure strategy. Rahim said the company has been early to 800GbE and 1.6TbE and is extending Ethernet into Scale Up systems that tightly interconnect accelerators within AI racks. The Investor Day architecture maps QFX and PTX into AI training and inference clusters, PTX into DCI, MX into AI on-ramp and edge roles, and Marvis across the infrastructure as the agentic operations layer.

HPE Networking Investor Day — Key Takeaways

Networks for AI — Cumulative FY25-FY26 order outlook raised to more than $3 billion

Data Center — Low-to-high 50% CAGR projected for FY26-FY29

Routing — Low-to-high 20% CAGR; AI on-ramp and DCI/Scale Across among the principal drivers

Scale Up — Purpose-built QFX/AMD Helios Ethernet fabric extends HPE networking inside the AI rack

FY27 Growth — High teens to low 20% Networking revenue growth guidance

FY27 Margin — Networking operating margin targeted in the mid-to-high 20% range

Juniper Synergies — FY28 annualized cost-synergy target increased from at least $600 million to $800 million

“You can invest billions of dollars into GPUs, but if the network can’t keep up, those resources are not fully utilized,” Rahim said.

Scale Up Becomes a New Networking Market for HPE

AMD Helios provides one of the clearest examples of HPE’s expanded networking strategy. HPE is combining Juniper-derived networking technology with HPE’s rack-scale engineering, compute and direct liquid cooling to provide an integrated Ethernet Scale Up fabric. HPE said the Helios opportunity represents more than $1 billion of networking potential over the next two years, with networking tray orders already exceeding $200 million. The company separately announced a $1.2 billion order from Vultr for AMD Helios AI Rack by HPE systems, its first commercial order for the platform. Each Helios rack incorporates six HPE Juniper Networking QFX5252 Scale Up Ethernet switch trays connecting 72 AMD Instinct MI455X GPUs and supporting UALink over Ethernet.

That order moves the HPE-AMD architecture beyond the development stage and provides HPE with a commercial entry into a networking domain traditionally located much closer to accelerators than conventional data center Ethernet. HPE’s strategy also extends beyond selling the networking as part of a complete HPE rack: Rahim said the Scale Up networking trays can be sold independently to other solution providers. Converge Digest separately covered the Vultr deployment and HPE’s earlier development of the QFX5252 for AMD Helios.

Scale Out remains another major growth vector. HPE says its first-to-market 1.6T Scale Out solutions are contributing to AI networking wins, while the recent Oracle agreement provides a large reference deployment. Oracle is deploying HPE Juniper Networking switching and routing across a multi-year, multi-gigawatt AI infrastructure buildout, with QFX platforms serving AI data center switching requirements and PTX routing addressing connectivity across the infrastructure. HPE describes DCI and AI on-ramp traffic as major drivers behind its projected low-to-high 20% routing CAGR.

Juniper Integration Moves to the Channel and Operating Systems

Rahim also provided a detailed update on the integration of Juniper Networks, which HPE completed in July 2025. HPE has already combined the networking sales organizations under one catalog, compensation plan and set of account rules. The next major milestone comes November 1, when the former Aruba and Juniper partner programs converge under HPE Partner Ready Vantage. Rahim said only about 10% of the two partner populations overlapped before the acquisition, creating an opportunity to bring the combined portfolio to a much broader channel. HPE’s presentation confirms that the unified partner program and the first phase of its combined lead-to-cash operating model begin in November.

HPE also increased its FY28 annualized cost-synergy target from at least $600 million to $800 million, citing progress in SG&A, go-to-market harmonization and elimination of duplicated R&D costs. The broader integration strategy now has four primary growth components: expanding Campus & Branch through HPE’s go-to-market reach, increasing services attachment across the larger installed base, cross-selling data center networking into HPE enterprise accounts, and combining rack-scale systems with networking to pursue AI infrastructure opportunities.

ServiceNow Targets a Fully Autonomous Network

The Investor Day also provided a customer view of HPE’s Self-Driving Network strategy. Sajeev Nair, senior director of Digital Core Services at ServiceNow, described the company’s progression from an initially fragmented network environment to an architecture spanning wireless, wired and WAN infrastructure managed with Juniper Mist and Marvis capabilities. According to Nair, ServiceNow has reduced network incident volume by approximately 90%, while network patching work that previously consumed roughly 2,000 to 3,000 staff hours annually has fallen below 60 hours.

Nair said ServiceNow is pursuing a goal of reaching a 100% autonomous network by 2028. He emphasized a model of “graduated autonomy,” progressively allowing AI and automation to assume more operational responsibility rather than immediately handing full control to an autonomous system. HPE is meanwhile extending Marvis across HPE Mist and HPE Aruba Central, positioning agentic AI, conversational interaction and trusted self-driving actions as an operational layer across the combined networking portfolio. The Investor Day architecture explicitly extends that model from campus and branch networks into enterprise data centers and AI infrastructure.

The most consequential aspect of HPE’s Investor Day is the expansion of its networking addressable market toward both ends of the AI infrastructure topology. At one end, HPE is moving inward from conventional data center switching into Scale Up connectivity between accelerators inside rack-scale systems. At the other, PTX and MX extend the portfolio outward through AI on-ramps, backbone routing and Scale Across/DCI. Between those layers sit the QFX-based Scale Out fabrics used to connect increasingly large accelerator clusters.

This gives HPE a relatively straightforward architectural story for Networks for AI: Scale Up with QFX/Helios, Scale Out with QFX and PTX, Scale Across with PTX, and AI on-ramp and edge routing with MX. The portfolio is underpinned by a combination of merchant switching silicon and Juniper’s internally developed Trio and Express routing silicon. HPE’s Investor Day architecture also places SRX security and Marvis operations across several of these domains rather than treating security and AIOps as independent products.

The financial targets give that architecture additional significance. HPE estimates the broader Data Center Networking market will grow at a 44% CAGR from CY26 through CY29, while projecting its own Data Center Networking revenue to grow in the low-to-high 50% range from FY26 through FY29. Routing is projected to grow in the low-to-high 20% range against the 20% market growth estimate used by HPE. These are company forecasts rather than realized results, but they demonstrate how heavily HPE’s post-Juniper growth framework now depends on AI networking.

The Juniper acquisition is therefore moving beyond the original combination of Aruba enterprise networking and Juniper routing, switching and Mist AIOps. HPE is using those assets as part of a broader systems strategy that combines networking with rack-scale compute, liquid cooling and AI infrastructure engineering. AMD Helios and Oracle represent two different manifestations of that strategy: Helios takes HPE networking deeper into the rack through Scale Up Ethernet, while Oracle extends QFX and PTX across large Scale Out and Scale Across AI infrastructure. HPE’s increased Networks for AI order outlook—now exceeding $3 billion cumulatively across FY25 and FY26—provides a measurable indicator of how quickly that strategy is developing.