World's First Vera Rubin Rack Ships, HPE Smashes Q2, Cisco: Agents Drive 9x WAN Surge

By: Network-Switch.com | June 3, 2026

Table of contents

 

Next-generation AI infrastructure and data center networking

1. Dell Delivers World's First Nvidia Vera Rubin NVL72 Rack to CoreWeave (June 1)

On June 1, 2026, Dell Technologies became the first OEM to deliver an operational Nvidia Vera Rubin NVL72 rack-scale AI supercomputer system to a commercial customer. The recipient is AI cloud provider CoreWeave. Michael Dell confirmed the shipment on X, posting photographs of the live hardware alongside system diagnostic logs showing all components passing inspection: "The world's first @nvidia Vera Rubin NVL72 server rack is here. We're thrilled to deliver the first working, liquid-cooled @Dell PowerEdge XE9812 for @CoreWeave. Built for the next era of AI infrastructure."

The system - Dell's PowerEdge XE9812 - integrates 72 Nvidia Rubin GPUs, 36 Nvidia Vera CPUs, NVLink-6 switching, Nvidia Spectrum-X Ethernet fabric, and custom direct liquid cooling in a single rack-scale unit. According to Nvidia, the Vera Rubin NVL72 delivers 10x the agent throughput at scale compared with the previous-generation Grace Blackwell NVL72, and reduces cost per token for large-scale agentic AI inferencing by up to 10x. CoreWeave's deployment includes Micron 7600 SSDs for improved energy efficiency and full system-level rack validation. CoreWeave is already one of Nvidia's closest cloud partners and holds multi-billion dollar compute supply agreements with Meta. This delivery marks the end of the Blackwell era as the frontier compute standard and the beginning of volume Vera Rubin deployments - with broader OEM availability from Dell, HPE, Lenovo, and Supermicro scheduled for H2 2026.

2. Nvidia GTC Taipei (May 31): Vera Rubin Enters Full Production, Spectrum-X CPO Switches Now Shipping

At Nvidia GTC Taipei on May 31, Nvidia announced that the Vera Rubin platform has entered full production, with Taiwan's top server manufacturers and global supply chain leaders actively manufacturing Vera Rubin-based systems at scale for AI labs, cloud providers, and hyperscalers. The announcement named Dell Technologies, HPE, Lenovo, and Supermicro as primary system builders in full-scale production - along with a broader ecosystem of 20+ ODMs and infrastructure partners including Foxconn, Quanta, Wistron, Compal, Pegatron, GIGABYTE, and Inventec.

The platform is architecturally structured as five purpose-built racks operating as a single massive AI supercomputer for agentic workloads: Vera Rubin NVL72 compute systems, the Vera CPU, the Groq 3 LPX for inference, the Vera BlueField-4 STX storage platform, and the Spectrum-6 SPX Ethernet rack. Separately, Nvidia confirmed that Spectrum-X Ethernet Photonics - the world's first co-packaged optics (CPO)-based production switches - are now in production and shipping. Built with 200 Gb/s SerDes, these switches deliver 5x better power efficiency, 5x longer AI uptime, and 1.3x faster time-to-deployment compared to networks using traditional pluggable transceivers. This marks the transition of CPO from roadmap promise to commercially available product, ahead of the broader industry ramp that market analysts projected for late 2026.

3. HPE Q2 FY2026 Results (June 1): $10.7B Revenue +40%, EPS $0.79 Doubles Guidance, Networking Routing Surges

Hewlett Packard Enterprise reported Q2 FY2026 results on June 1, 2026, posting what CEO Antonio Neri described as "record-breaking results" that exceeded every metric of the company's own guidance - and did so by a substantial margin. Revenue reached $10.7 billion, up 40% year-over-year, against the guided range of $9.6 billion to $10.0 billion. Non-GAAP EPS of $0.79 was 108% above the prior-year period and significantly above the guided $0.51-$0.55 range. GAAP gross margin hit 36.5%, up 810 basis points year-over-year. Free cash flow reached record levels for any HPE second quarter.

The networking business continued its transformation. Overall networking revenue grew 148% year-over-year, while routing revenue surged from $1 million in Q2 FY2025 to $775 million in Q2 FY2026 - effectively a new line of business created by the Juniper integration. Campus and branch orders hit a record high (upper-20% normalized growth), and enterprise data center switching orders rose nearly 20%. Network operating margin came in at 21.6%. HPE raised its cumulative fiscal 2026 Networks for AI order target to at least $2 billion. The Cloud and AI segment generated $7.7 billion in revenue, up 23%, with AI systems bookings of $1.8 billion bringing cumulative bookings to $16.4 billion; the company entered Q3 with a record $5.9 billion AI systems backlog. HPE raised full-year FY2026 non-GAAP EPS guidance to $3.35-$3.45 and free cash flow guidance to at least $3.5 billion - a target Neri noted is two years ahead of the plan HPE committed to just eight months ago. Q3 FY2026 revenue guidance was set at $11.5 billion to $12.1 billion.

4. Cisco Report (May 26): Agentic AI Will Drive 9x Enterprise WAN Traffic Growth by 2035

On May 26, Cisco published AI Impact on Wide Area Networks: Cisco Report 2026 - a study combining live service-provider traffic measurements, Cisco Crosswork Assurance telemetry, and empirical agent testing to quantify how AI is already reshaping WAN infrastructure. The headline finding: without agentic AI, enterprise network traffic is projected to grow approximately 2.5x between 2026 and 2035. With agentic AI adoption, enterprise traffic could grow 9x - driven by persistent machine-to-machine communication, continuous model inference, and autonomous task execution at software speed.

Key empirical findings from the report: AI agents generate 450% more traffic per task than a human performing the same task manually. Approximately 70% of agent-generated traffic is AI inference. Median flow duration for AI inference traffic is twice as long as regular web transactions (1,292 ms vs 643 ms). Real-world service-provider data shows AI inference traffic growing 4x in just eight months. The structural implications are equally notable: today's networks were optimized for heavy downstream, light upstream - but 9% of AI inference flows already carry more upstream than downstream traffic, versus only 0.5% for typical HTTP. Cisco projects that AI inference will account for 25% of all network traffic by 2035. The report frames "the spinal cord" between AI agents and large language models as the new critical network path - requiring resilience, QoS-differentiated treatment, and security engineering equivalent to what today's applications demand of fiber backbone infrastructure. For enterprise network architects and procurement teams, this is the most authoritative quantification to date of why WAN capacity, routing, and SD-WAN upgrades cannot be deferred.

5. Cisco Revamps CCNA and CCIE Certifications for the AI Era (May 26)

Also on May 26, Cisco announced the most significant overhaul of its flagship CCNA certification in seven years - held by more than 1.8 million networking professionals worldwide. The updated CCNA adds a new AI literacy pillar, expanded hands-on lab requirements, and a security-first approach that reflects the convergence of networking, automation, and AI operations. Simultaneously, Cisco updated the expert-level CCIE practical exam to include an AI-focused module covering agentic network operations, AI-driven observability, and automated remediation workflows.

The certification changes reflect Cisco's strategic view - backed by its WAN traffic report - that network engineers must now understand AI inference infrastructure, traffic engineering for agent workloads, and AI-assisted network management as core professional competencies, not specializations. For network engineers at organizations deploying AI infrastructure, the CCNA and CCIE updates signal that the skills required to design, operate, and troubleshoot networks are shifting materially - and that the certification landscape is adapting accordingly. Cisco's VP of Learning, Howard Carter, described the change as preparing engineers "not just for networks that carry AI traffic, but for networks that are operated by AI."

6. HPE Juniper SRX 400 Series: AI-Native Security Routing Launched in Q2

HPE disclosed in its Q2 FY2026 earnings commentary that the HPE Juniper SRX 400 Series security router was launched during the quarter, contributing to the 15%-19% normalized growth in security orders. The SRX 400 is a next-generation firewall and routing platform designed for enterprise data center perimeter, campus edge, and AI factory on-ramp deployments - combining security services (IPS, advanced threat prevention, encrypted traffic analytics) with high-throughput routing in a single appliance.

The launch is part of HPE's broader post-acquisition strategy to unify Juniper's historically enterprise-focused security portfolio with HPE Aruba's campus networking platform under a single AIOps management layer. Service provider routing orders rose nearly 30% on a normalized basis in Q2, driven by cloud service provider deployments - a signal that Juniper's MX-series and PTX-series routers are gaining traction in the AI WAN on-ramp market that Cisco's report (story 4) has now quantified as a major growth vector. The HPE Juniper SRX 400 availability timing and pricing have not been publicly disclosed as of the earnings call.

Editor's Summary

The week of May 26-June 2 delivered some of the most operationally concrete AI infrastructure news of the year. Dell's first Vera Rubin NVL72 delivery to CoreWeave marks the close of Blackwell as the frontier standard - the next generation is no longer a roadmap item; it is a shipped product. Nvidia's GTC Taipei announcement that Spectrum-X CPO switches are in production means co-packaged optics has crossed from pilot to commercial deployment, reshaping the economics of every 800G and 1.6T switch purchase. HPE's Q2 results - revenue 7% above the top of its own guidance, EPS double the midpoint - confirm that the Juniper integration is executing ahead of every timeline the company set. And Cisco's WAN report is the clearest quantification to date of what agentic AI actually means for network capacity: 9x enterprise traffic growth, 450% more traffic per AI task, and a structural shift in upstream/downstream asymmetry that will force router, WAN, and SD-WAN upgrades at every tier of the network. For enterprise IT procurement teams, these five data points together define the network infrastructure investment case for the next three years.

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 is the Dell PowerEdge XE9812 and who received the first Vera Rubin NVL72 system?

The Dell PowerEdge XE9812 is the world's first commercially delivered Nvidia Vera Rubin NVL72 rack-scale AI supercomputer system, shipped by Dell Technologies on June 1, 2026. It was delivered to CoreWeave, an AI cloud provider. The system integrates 72 Nvidia Rubin GPUs, 36 Nvidia Vera CPUs, NVLink-6 switching, Nvidia Spectrum-X Ethernet fabric, and custom liquid cooling in a single rack. It delivers up to 10x the agent throughput and 10x lower cost-per-token compared to the previous-generation Grace Blackwell NVL72.

What is Nvidia Spectrum-X Ethernet Photonics and is it available now?

Nvidia Spectrum-X Ethernet Photonics is the world's first co-packaged optics (CPO)-based production Ethernet switch, announced at Nvidia GTC Taipei on May 31, 2026. It combines CPO technology with Spectrum-X switching, using 200 Gb/s SerDes. Compared to switches using traditional pluggable transceivers, it delivers 5x better power efficiency, 5x longer AI uptime, and 1.3x faster time-to-deployment. As of June 2026, it is in production and shipping - marking the first time CPO technology has transitioned from roadmap to commercially available product in the networking industry.

What were HPE's Q2 FY2026 earnings results and networking highlights?

HPE reported Q2 FY2026 results on June 1, 2026: revenue of $10.7 billion (+40% YoY), non-GAAP EPS of $0.79 (108% above the prior year), and GAAP gross margin of 36.5% (+810 basis points YoY). Networking revenue grew 148% year-over-year, while routing revenue surged from $1 million to $775 million in a single year following the Juniper Networks integration. HPE entered Q3 with a record AI systems backlog of $5.9 billion and raised its full-year FY2026 free cash flow guidance to at least $3.5 billion - two years ahead of the plan it committed to eight months prior.

What does Cisco's 2026 WAN report say about the impact of agentic AI on network traffic?

Cisco's "AI Impact on Wide Area Networks 2026" report, published May 26, 2026, found that agentic AI could drive enterprise network traffic growth of 9x by 2035, compared to 2.5x without AI adoption. AI agents generate 450% more traffic per task than humans doing the same task. Approximately 70% of agent-generated traffic is AI inference. AI inference flows last twice as long as regular web transactions. Real-world service-provider data shows AI inference traffic growing 4x in just eight months. Cisco projects AI inference will account for 25% of all network traffic by 2035.

What changes did Cisco make to CCNA and CCIE certifications in May 2026?

On May 26, 2026, Cisco announced the most significant overhaul of its CCNA certification in seven years - held by more than 1.8 million professionals globally. The updated CCNA adds a new AI literacy pillar, expanded hands-on lab requirements, and a security-first approach. The expert-level CCIE practical exam was also updated with an AI-focused module covering agentic network operations, AI-driven observability, and automated remediation workflows. The changes reflect the industry's recognition that network professionals must now understand AI inference infrastructure and AI-assisted operations as core competencies.

 

Sources