HPE Declares Itself a Networking Company, Dell's AI Servers Surge 757%, Marvell Ships 102.4T Switch

By: Network-Switch.com | June 23, 2026

Table of contents
HPE Discover 2026

1. HPE Discover 2026 Closes: "I Think HPE is Very Definitely Now a Networking Company"

HPE Discover Las Vegas 2026 closed on June 18 after four days at The Venetian Convention and Expo Center, and it will be remembered as the event where HPE definitively repositioned itself around networking rather than servers or storage. CEO Antonio Neri's opening keynote on June 16, titled "Architecting AI Starts with Your Network," argued that AI factories are fundamentally about "turning electrons into tokens" - and that the network, not compute alone, is both the critical enabler and defining constraint of that process. Networking president Rami Rahim, the former Juniper CEO, followed with a cautionary parable about San Francisco's Millennium Tower - a building that slowly collapsed under its own weight due to inadequate foundations - to argue that enterprises building agentic AI without solid network architecture are setting themselves up for the same fate.

In a press Q&A after the keynote, Neri made the company's new self-image explicit: asked whether networking is now HPE's primary growth engine, he stated plainly, "in my mind, together with our competitor, Cisco, we have the most complete networking portfolio" - a remarkable statement from a company that, as recently as 2015, would not have named Cisco as a direct competitor. Analyst Steven Dickens, CEO of HyperFrame Research, echoed the shift on-site: "I think HPE is very definitely now a networking company." The numbers back the rhetoric: HPE management now expects the company to grow into an $11 billion-plus networking business, with the Juniper acquisition delivering what Neri called "the whole package, from scale up to scale out" across both data center and campus environments.

2. New HPE Juniper Switches: QFX5252 for AMD Helios, QFX5140 for Inference at the Edge

HPE unveiled two new Juniper-branded switches at Discover, both available immediately as part of the Dell-style "validated design" approach within the HPE AI Factory. The HPE Juniper Networking QFX5252 is a switch tray purpose-built for the AMD Helios AI rack-scale platform, delivering 1,024 Tbps of total switching capacity, running the open-source SONiC network operating system with AI-native operations, and 100% liquid-cooled for dense AI rack environments. HPE was among the first vendors to offer AMD Helios infrastructure, and the QFX5252 gives that platform the low-latency, high-bandwidth switching fabric needed to maximize performance at scale.

The HPE Juniper Networking QFX5140 targets a different use case entirely: distributed AI inference at the network edge. Delivering 16 Tbps of switching capacity in a single rack unit on a Broadcom Trident 5 ASIC, the QFX5140 is - in Rahim's words - "truly purpose built for the next wave of AI infrastructure with a special focus on inferencing clusters at the edge." The switch is natively integrated with HPE Marvis, the AI-driven network assurance engine, and reflects a broader industry recognition (also voiced by Cisco at Cisco Live two weeks earlier) that AI inference workloads are moving out of centralized hyperscale data centers and into distributed, edge-adjacent deployments where 16 Tbps in 1RU is a more practical fit than the 100+ Tbps switches built for centralized training clusters.

3. HPE Brings Quantum-Safe Firewalls and Confidential Computing to the AI Factory Perimeter

Two HPE Discover announcements carry outsized long-term significance for regulated-industry buyers. First, the HPE Juniper Networking SRX4700 - a 1.4 Tbps quantum-safe firewall in a single rack unit - brings post-quantum cryptography directly to the AI factory perimeter, mirroring the quantum-safe-by-default commitment Cisco made at Cisco Live earlier this month. Combined with the Juniper PTX routing family (handling 800 Gbps with ZR/ZR+ coherent optics for distributed AI site interconnect) and the new MX304 router for inference edge deployments, HPE now offers quantum-safe protection across the entire AI networking stack - from rack to WAN.

Second, and arguably the most consequential announcement of the week for enterprise AI deployment, HPE confirmed that Nvidia Confidential Computing will become standard across the entire HPE AI Factory with NVIDIA portfolio in Q4 2026 - at no additional configuration burden. Confidential computing protects data in use (not just at rest or in transit) by executing sensitive workloads inside hardware-based Trusted Execution Environments, with cryptographic attestation proving to remote systems that the enclave is running trusted, unmodified code. Until now, enabling confidential computing in production AI infrastructure required custom security architecture; HPE making it a default removes that barrier specifically for healthcare, financial services, government, and sovereign AI deployments - sectors where data-in-use protection is often a hard compliance requirement rather than a nice-to-have.

4. Dell Q1 FY2027: AI Server Revenue Up 757% to $16.1B, Stock Jumps 32.76% in a Single Day

On May 28, Dell Technologies reported first-quarter fiscal 2027 results that triggered the company's best single trading day since returning to public markets in 2018, with the stock surging 32.76%. Total revenue hit a record $43.8 billion, up 88% year-over-year - beating Wall Street consensus by more than $8 billion. The headline figure: AI-Optimized Server revenue reached $16.1 billion, up 757% year-over-year, with $24.4 billion in new AI orders booked in the quarter alone, pushing the AI backlog to $51.3 billion.

The Infrastructure Solutions Group (servers, storage, networking) posted record revenue of $29.0 billion, up 181% year-over-year, while Traditional Servers and Networking revenue reached $8.5 billion, up 92%. Storage revenue hit a record $4.3 billion, up 8%. Gross margin compressed to 18.1% due to AI product mix, but operating income more than doubled on disciplined expense management. COO Jeff Clarke was unambiguous about the demand environment: "The demand is not a demand issue for us. We are supply-constrained in the second half," citing persistent shortages in DRAM, NAND, microprocessors, and hard drives as the binding constraint on near-term growth - not customer orders. Dell raised full fiscal 2027 revenue guidance by $27 billion to a midpoint of $167 billion (up ~47% year-over-year) and lifted its full-year AI server revenue forecast to $60 billion, up 144% from FY2026.

5. Marvell Ships Industry's First 102.4 Tbps Switch Purpose-Built for AI and Cloud Data Centers

On June 1, Marvell Technology announced general availability of what it describes as the industry's first 102.4 Tbps switch silicon purpose-built for AI and cloud data center infrastructure. The chip extends Marvell's position as a core supplier of networking and custom-compute silicon for hyperscale and AI infrastructure providers, competing directly in the same high-radix switching tier where Broadcom's Tomahawk 6 (used in HPE's QFX5250) and Nvidia's Spectrum-X currently lead. Marvell's switch targets the same fundamental problem driving every other announcement in this edition: AI training and inference clusters require ever-higher per-chip bandwidth density to keep GPU utilization high and avoid network-induced bottlenecks in distributed training jobs.

The announcement reinforces a broader 2026 trend that Dell'Oro Group and other analysts have tracked throughout the year: the AI switching silicon market is no longer a two-horse race between Broadcom and Nvidia. Marvell's deepening relationships with Nvidia (a $2 billion investment earlier in 2026) and its existing RAN silicon business with Samsung and Nokia give it a credible multi-front position spanning data center switching, optical interconnect, and telecom radio silicon - three of the fastest-growing segments in the broader AI infrastructure buildout. For enterprise buyers and OEM partners, the emergence of a credible third silicon supplier at the 100+ Tbps tier should, over time, support more competitive pricing and supply diversification - a meaningful consideration given the supply constraints Dell and other OEMs have flagged for the second half of 2026.

6. HPE Aruba and Juniper Cross-Pollinate: Marvis Comes to Central, CX Switches Come to Mist

Beyond new hardware, HPE Discover delivered the clearest evidence yet that the Aruba-Juniper integration - nearly a year after the $14 billion acquisition closed - is moving from roadmap promise to shipped reality. HPE announced that HPE Networking CX switches (the Aruba-branded switching line) are now manageable through HPE Mist (Juniper's AI-driven cloud management platform), and conversely that Marvis actions - Juniper's AI-powered automated remediation engine - are coming to Aruba Central. HPE calls this "cross-pollination," and it directly addresses the question every enterprise customer has been asking since the acquisition closed: will Aruba and Juniper remain separate product lines indefinitely, or will the platforms genuinely converge?

The agentic AI angle was layered throughout: HPE announced secure agentic operations for the AI Factory, meaning autonomous agents now run within bounded, secure environments with built-in guardrails, alongside support for Nvidia NemoClaw (an open agent framework) and Zerto-based data recovery if an agentic workflow goes wrong. HPE Mist also gained proactive maintenance with predictive analytics and an advanced reasoning agent for high-confidence automated remediation - continuously analyzing telemetry across power, temperature, optics, and system health to flag potential failures before they cause outages. Rahim summarized the philosophy directly: "This leverages more than a decade's worth of AI expertise that we've built, and it helps customers improve uptime, reduce risk, and operate just far more efficiently."

Editor's Summary

This was one of the densest weeks of the year for enterprise networking news, anchored by HPE Discover's unambiguous declaration that HPE now sees itself - and wants customers to see it - as a networking company on par with Cisco, not merely a server-and-storage vendor with a networking division. The new QFX5252 and QFX5140 switches, the quantum-safe SRX4700 firewall, and Q4 2026 confidential computing defaults collectively give HPE one of the most complete AI networking stacks in the industry, spanning from inside the GPU rack to the WAN edge. Dell's Q1 FY2027 results - AI server revenue up 757% in a single quarter - confirm that the demand side of this story remains supply-constrained rather than demand-constrained, a dynamic Marvell's new 102.4 Tbps switch is directly aimed at addressing by widening the silicon supplier base. For enterprise IT buyers, the message from this week is consistent across every vendor: network infrastructure decisions made in the second half of 2026 will be constrained primarily by component supply, not by a shortage of capable products to choose from.

Network-Switch.com is a global professional distributor of networking equipment from Cisco, Huawei, Ruijie, H3C, and our own NS brand - including switches, routers, firewalls, wireless APs, optical modules, and fiber patch cables. Our CCIE, HCIE, H3CIE, and RCNP certified engineers deliver end-to-end technical support and complete network solutions for enterprise, campus, and large-scale infrastructure deployments worldwide.

Visit us: https://network-switch.com

Frequently asked questions (FAQs)

What new networking products did HPE announce at Discover 2026?

At HPE Discover Las Vegas 2026 (June 15-18), HPE announced the HPE Juniper Networking QFX5252 switch tray for the AMD Helios AI rack platform (1,024 Tbps capacity, SONiC, liquid-cooled), the QFX5140 inference switch (16 Tbps in 1 rack unit, Marvis-integrated), and the SRX4700 - a 1.4 Tbps quantum-safe firewall for the AI factory perimeter. HPE also confirmed Nvidia Confidential Computing will become standard across its entire AI Factory with NVIDIA portfolio starting Q4 2026.

What were Dell's Q1 fiscal 2027 financial results?

Dell Technologies reported Q1 FY2027 results on May 28, 2026: record revenue of $43.8 billion (up 88% year-over-year), beating Wall Street estimates by more than $8 billion. AI-Optimized Server revenue reached $16.1 billion, up 757% year-over-year, with $24.4 billion in new AI orders booked, pushing the AI backlog to $51.3 billion. Dell stock surged 32.76% in a single day - its best trading day since returning to public markets in 2018. The company raised full-year FY2027 revenue guidance to a $167 billion midpoint and lifted its AI server revenue forecast to $60 billion.

What is the Marvell 102.4 Tbps switch and why does it matter?

On June 1, 2026, Marvell Technology announced general availability of the industry's first 102.4 Tbps switch silicon purpose-built for AI and cloud data centers. It positions Marvell as a credible third major supplier in high-radix AI switching silicon, alongside Broadcom (Tomahawk 6) and Nvidia (Spectrum-X). The launch is significant for enterprise buyers because it widens the silicon supplier base at a time when AI networking components face supply constraints, potentially supporting more competitive pricing and availability.

How are HPE Aruba Networking and Juniper Networking being integrated after the acquisition?

At HPE Discover 2026, HPE announced concrete "cross-pollination" between its Aruba and Juniper platforms, nearly a year after completing the $14 billion Juniper acquisition. HPE Networking CX switches (Aruba-branded) are now manageable through HPE Mist (Juniper's AI-driven cloud platform), and Marvis actions (Juniper's automated remediation AI) are coming to Aruba Central. This represents the first major evidence that the two previously separate networking product lines are converging into a unified management platform rather than operating independently.

Is HPE now considered a networking company rather than primarily a server and storage vendor?

At HPE Discover 2026, CEO Antonio Neri explicitly repositioned the company around networking, stating that HPE - together with Cisco - now offers "the most complete networking portfolio" in the industry. Analyst Steven Dickens of HyperFrame Research stated on-site that HPE is "very definitely now a networking company." This marks a significant shift from HPE's historical identity as primarily an enterprise server and storage vendor, driven by the Juniper Networks acquisition and HPE's argument that the network - not compute alone - is the defining constraint on enterprise AI deployment.

Sources