Analyst(s): Fernando Montenegro, Tom Hollingsworth
Publication Date: June 24, 2026
A year after buying Juniper, HPE tells one AI story, with networking at the center and compute, software, security, and financing behind it. The architecture is coherent. The harder question is whether HPE can turn that breadth into momentum, and whether its security story can reach past the network.
What Is Covered in This Article:
- HPE’s networking-centered, full-stack AI architecture and the self-driving network.
- AI infrastructure and agentic operations, and the pitch to move AI from pilot to production.
- Why HPE’s network-anchored security is necessary but not sufficient.
- The virtualization reset as an ongoing opportunity, via VM Essentials and HPEFS financing.
- The two-way dependency with NVIDIA and the GSI-led services motion, with momentum as the core test.
The Event—Major Themes & Vendor Moves: HPE Discover Las Vegas 2026 (June 15-18) was the company’s first full Discover since its Juniper Networks acquisition closed in the summer of 2025, and that integration set the tone. HPE’s central message was that networking has become the control plane for enterprise AI, and that the Juniper deal is what lets it make that claim. The show drew many thousands of customers and partners to The Venetian across more than 225 sessions and hands-on labs, with over 80 customers presenting. Across keynotes from CEO Antonio Neri, networking lead Rami Rahim, and CTO Fidelma Russo, HPE laddered its announcements up to a single architecture spanning silicon, routing, AI infrastructure, operations, and governance.
AI Networking and the Self-driving Network
HPE made networking the centerpiece. It deepened Juniper integration into its AI Data Center Solutions and refreshed the switch and router lineup for AI workloads, including the liquid-cooled QFX5252 scale-up switch for AMD Helios racks, the QFX5250 scale-out switch, the QFX5140 inference switch, the MX301 inference-edge router, and the PTX12000 for data center interconnect. AI-Native Routing lets routers enforce policy on access to public AI platforms. The Self-Driving Network, framed as self-healing, self-protecting, and self-optimizing, runs through HPE Mist, with Marvis Actions now in Aruba Central and CX switches moving to Mist management, all coordinated by an agentic mesh of models and operational agents.
AI Infrastructure
On the infrastructure side, HPE refreshed Private Cloud AI and the HPE AI Factory with NVIDIA. Private Cloud AI now scales inference to 256 GPUs with NVIDIA KAI and Run:ai, shared KV cache, and an Alletra MP X10000 and Data Fabric data layer, with agentic governance built in; HPE claims it reaches value 7 to 12 months faster than do-it-yourself, with no egress or per-token fees. The AI Factory spans hundreds to tens of thousands of GPUs, makes NVIDIA Confidential Computing standard, and adds Blackwell GPUs, Spectrum-X Ethernet, and BlueField-3 networking, with the Vera Rubin NVL72 on the roadmap. HPE cited up to 17x higher throughput and 20x lower cost to first token.
Agentic IT Operations
The software story was about running the agentic enterprise. GreenLake Intelligence adds an agentic mesh that applies identity, policy, and control to any AI agent, alongside new compute, orchestration, and observability copilots. Morpheus 9 anchors a unified CloudOps family with OpsRamp and Zerto, and HPE pushed an aggressive virtualization-migration offer: a free first year of Morpheus VM Essentials and Zerto for $1. A new ServiceNow partnership ties GreenLake into service delivery.
Compute
In compute, HPE introduced the ProLiant DL394 Gen12 built on NVIDIA’s Vera CPU, the Compute Scale-up Server 3250 for high-velocity transactions and real-time analytics, and added agentic advisory intelligence to Compute Ops Management.
Security
Security was framed as built into the network rather than added on. HPE introduced SASE Orchestrator, which will unify SD-WAN and Security Service Edge in a single console and is due later in 2026, along with SASE Copilot, an AI Firewall, and AI Predictive Threat Prevention on a new quantum-safe SRX4700. It extended Universal ZTNA across policy, identity, and enforcement to non-human and agentic identities, and added a Security Director Copilot.
Financing and Lifecycle
HPE also leaned on financing as a lever. HPE Financial Services, with $13.5B in net portfolio assets and 50,000 customers, introduced a Network Migration Program offering better-than-cash terms on hardware, zero-interest networking software over three years, and revenue share on retired multi-vendor gear. The new 90/9 Advantage defers payments for 90 days, then charges 1% monthly for 9 months, aimed squarely at refresh and virtualization migrations.
Quantum
Finally, HPE expanded its quantum partner ecosystem across multiple hardware and software vendors and reiterated post-quantum cryptography readiness work targeting 2031.
HPE Discover 2026: A Coherent AI Story That Now Has to Convert
Analyst Take—A Coherent Story that now has to Convert: A year after the Juniper acquisition closed, HPE has done the hard part. It arrived at Discover with one coherent full-stack AI story, with networking as the organizing principle tying compute, storage, software, and operations together. The strategic logic holds: as enterprises add AI agents to their existing user-driven workloads, the network becomes where performance, security, and governance are determined.
The harder question is conversion. HPE’s enterprise footprint is broad and durable; turning that presence into spending acceleration is the work ahead.
Networking as the AI Control Plane
The networking story is the strongest, most defensible part of HPE’s hand. Few rivals can present campus, data center, routing, and SASE as a single portfolio, and fewer still pair it with compute, storage, and operations underneath. Making the network the place where AI traffic is routed, secured, and observed is a credible way to stay strategically relevant in deployments that might otherwise be defined by the chip vendor.
The honest tension is integration. Two heritage management planes, HPE Mist and Aruba Central, still coexist, and bridging features like Marvis Actions do not settle which console owns what. The promised simplification has to show up as fewer decisions for operators, not just a unified diagram.
Breadth Pays Off in Operations if It Holds Up
Operations is where HPE’s size actually helps. Running network, compute, and storage from one operating model and applying identity and policy to AI agents is something that narrower vendors struggle to copy. That is the job of GreenLake Intelligence, with Morpheus and OpsRamp handling management and observability.
It also targets the right problem. Most enterprises can stand up an AI pilot; far fewer can run AI in production. HPE’s argument is that the operating model is what gets them across that line. The company leaned hard on cost, too, talking up tokens as the unit of spend and a much lower cost to first token.
The demand is there, but HPE holds only a slice of it. In ETR’s AI Product Series, about 21% of organizations building AI applications use HPE for hosting and deployment, and plan to stay, with another 14% considering it. Most of the field is still open.
The catch is execution. This has to work in the mixed, multi-vendor, half-migrated environments most companies actually run, not just in a clean reference design.
Security: Necessary but Not Sufficient
HPE is right to build security into the network. The pieces are real: Universal ZTNA stretched to cover non-human and agentic identity, an AI firewall, predictive threat prevention, and quantum-safe routing. Using the network to see, detect, and enforce plays to what HPE already does well.
The point is less a criticism than a question of scope. To win the larger, more strategic AI deployments HPE is after, it will share the table with vendors that already sell explicit security-for-AI lines. Cisco tells a similar network-as-security story and has added AI Defense. NVIDIA is a subtler case: its reference designs increasingly carry security, partly its own, through Confidential Computing on its GPUs, NeMo Guardrails for models, and the Morpheus and BlueField stack, and partly through security partners that build on those primitives. In that company, owning the network layer is a strong foundation, though probably not the whole story that customers will ask for.
There is a related positioning question. HPE has been clear that it does not aim to go head-to-head with the key security vendors, which is a sensible focus. But those vendors now sell purpose-built security-for-AI offerings: Palo Alto Networks with Prisma AIRS across the agentic lifecycle, CrowdStrike with AI agent discovery and runtime protection, and Microsoft with AI security woven into Defender and Entra, among others. A vendor positioning itself as a strategic AI provider tends to win more easily when it can speak fluently about AI security across the stack, even when customers ultimately buy their controls from one of those specialists. The goal here is a credible point of view, not necessarily a competing product line.
That breadth need not come through acquisition. Strong partnerships, of the kind HPE already has with several security vendors, could cover much of the ground, provided the integrations run deep, and the joint story is clear. NVIDIA’s own approach is instructive: much of the security in its AI designs comes from partners building on its platform rather than from NVIDIA alone.
It is also worth noting a structural limit to any network-anchored approach. Not all agentic traffic behaves like a user on the corporate network. A good deal of it will be application-to-application, often between external SaaS providers, and may never cross an HPE switch or router. Security that lives in the network sees less of those flows. What HPE has built is a strong start; rounding it out, organically or with partners, would extend the story to the data, the models, the agents, and the SaaS-to-SaaS paths between them, and bring security into every AI-infrastructure conversation.
Capitalizing on the Virtualization Reset
None of this is new. Broadcom’s rework of VMware pricing has been pushing customers to look elsewhere for a couple of years now, and plenty of vendors, HPE among them, have been mining that discontent ever since. What Discover added was more incentive, not a new strategy.
The product piece is Morpheus VM Essentials, which HPE pushed hard with a free first year and near-free migration tooling. The financing piece is HPEFS, with programs like the 90/9 Advantage and the Network Migration Program that let customers spread or defer costs rather than pay up front.
It helps to be realistic about how this plays out. Few enterprises rip out VMware wholesale. The usual path is coexistence: keep VMware for existing workloads, put new ones on HPE’s hypervisor, and migrate in waves. That is where the real pitch sits, in a single console that manages both VMware and the HPE hypervisor through the transition, rather than a clean swap.
Neither piece is groundbreaking on its own, and the field of VMware replacements is crowded. Bundled, the offer lowers the cost and friction of starting that migration, which is the point. The harder part is execution, because winning here is a sales-motion problem more than a technology one.
A Two-Way Dependency With NVIDIA
HPE’s AI infrastructure leans heavily on NVIDIA, from Spectrum-X networking and BlueField to Blackwell and the coming Vera Rubin systems. That is worth watching, because NVIDIA’s reference designs keep reaching further up the stack, into networking and security, the same ground HPE wants to own.
But this runs both ways. NVIDIA needs partners like HPE to get its technology into regulated, on-premises, and traditional enterprises, and that is exactly where HPE is strong: systems integration, lifecycle services, financing, sovereignty, and a large installed base. HPE is not a passenger here.
It also helps to remember that most enterprise workloads are still not AI. HPE’s bread and butter, compute, storage, networking, and operations for everything that is not a GPU cluster, is real revenue and real leverage that the AI story tends to overshadow.
That value-add increasingly lives in the services layer. The largest AI projects are shaped by global system integrators and consultancies, not bought off a price list, and HPE is better positioned here than the keynote conveyed, with its own professional-services arm and co-sell ties to Deloitte, Accenture, HCL, and TCS, among others. The open question is depth: whether that motion scales past a handful of flagship offers.
The job is to keep adding value on top of NVIDIA’s silicon, so that integration, operations, and trust stay HPE’s own rather than getting absorbed into the reference design.
What to Watch:
- Does the self-driving network deliver in brownfield environments, and do Mist and Aruba Central merge? HPE’s operations pitch rests on agents cutting real work. Watch whether that holds in mixed, multi-vendor environments, and whether the two management planes resolve into a single experience.
- Can HPE extend its security story above the network? The network is a strong base, but data, models, agents, and SaaS-to-SaaS traffic sit beyond it. Watch whether HPE builds or partners its way to a stronger AI security position.
- How fast does SASE Orchestrator evolve, and does it truly unify SD-WAN and SSE? It is slated for later in 2026. The test is one real console and policy model across both, not two products under one name.
- Can HPE keep its value-add as NVIDIA climbs the stack? As reference designs absorb more networking and security, HPE’s edge shifts to integration, operations, and the GSI co-sell motion. Watch whether that differentiation deepens or gets commoditized.
For more information, read the full announcement from HPE.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Other Insights From Futurum:
Can HPE’s Unified Private Cloud and Data Platforms Really Deliver AI Simplicity?
HPE Closes Juniper Acquisition, Combining AI-Native Networking Portfolios
HPE CEO Antonio Neri on Innovation in AI, Cloud, and Networking – Six Five On The Road
Author Information
Fernando Montenegro serves as the Vice President & Practice Lead for Cybersecurity & Resilience at The Futurum Group. In this role, he leads the development and execution of the Cybersecurity research agenda, working closely with the team to drive the practice's growth. His research focuses on addressing critical topics in modern cybersecurity. These include the multifaceted role of AI in cybersecurity, strategies for managing an ever-expanding attack surface, and the evolution of cybersecurity architectures toward more platform-oriented solutions.
Before joining The Futurum Group, Fernando held senior industry analyst roles at Omdia, S&P Global, and 451 Research. His career also includes diverse roles in customer support, security, IT operations, professional services, and sales engineering. He has worked with pioneering Internet Service Providers, established security vendors, and startups across North and South America.
Fernando holds a Bachelor’s degree in Computer Science from Universidade Federal do Rio Grande do Sul in Brazil and various industry certifications. Although he is originally from Brazil, he has been based in Toronto, Canada, for many years.
