Nvidia Tops the $15.4B Ethernet Market, HPE Ships the First UALink Switch, Quantum-AI Computing Arrives at Dell and HPE

network-switch.com

By: Network-Switch.com | July 15, 2026

Table of contents

1. IDC Q1 2026 Ethernet Switch Tracker: Nvidia Becomes No.1 at 21.5% Share in a $15.4B Market

IDC's Quarterly Ethernet Switch Tracker for Q1 2026, published this month, delivered the most significant market share realignment in the networking industry in decades. Nvidia became the number-one data center Ethernet switch vendor by revenue, generating $2.1 billion in the quarter - up 192.7% year-over-year - and capturing 21.5% of the data center segment. This is up from less than 4% share just two years ago. The overall Ethernet switch market reached $15.4 billion in Q1 2026, up 39.8% year-over-year, with the data center segment alone hitting a record $10 billion - up 61% year-over-year and now representing roughly two-thirds of the total market.

Cisco retained the largest share of the total Ethernet switch market at 29.3%, but the majority of its revenue remains in non-data center environments. In the data center segment specifically, Arista Networks held 20.7% share with $2.2 billion in revenue (+37.3% YoY) - and 92% of Arista's switch revenue came from data centers, making it the incumbent most exposed to AI infrastructure spending. HPE (including Juniper) held 6.4% of the total market at $985 million (+15.4% YoY), with 70.5% of its Ethernet switch revenue still in campus and branch environments. Within the data center segment, 800G switches now represent 35.8% of revenue - and IDC notes that 800G adoption is tracking even faster than 400G did at the same stage. IDC senior research manager Brandon Butler noted the structural dynamic that explains Nvidia's rise: "Networking is increasingly purchased alongside GPUs, software, and cluster architecture rather than as a separate layer of infrastructure."

Among the announcements at HPE Discover 2026 last month, the one with the largest long-term strategic implications arrived with the least fanfare: the HPE Juniper Networking QFX5252 switch tray - which we covered briefly in our June 22 edition for its AMD Helios support - also ships with native support for UALink-over-Ethernet (UALoE). UALink is the open-standard GPU scale-up interconnect developed by a consortium of 85-plus companies, ratified in April 2025, as a direct alternative to Nvidia's proprietary NVLink. The consortium includes HPE, AMD, Google, Microsoft, Apple, Cisco, Intel, Meta, AWS, and Alibaba - but notably not Nvidia. The QFX5252 is the first production switch to implement UALoE at commercial scale.

The distinction is foundational. NVLink is Nvidia's closed interconnect for scaling GPUs within a pod - it ties large GPU clusters together at high bandwidth and low latency, but only within Nvidia's own hardware ecosystem. UALink enables the same scale-up GPU connectivity on an open standard, allowing AMD Instinct, Intel Gaudi, and future accelerators to achieve NVLink-comparable intra-rack scaling on commodity Ethernet infrastructure. HPE shipping a QFX switch that speaks UALoE natively is a structural bet that the AI infrastructure market will eventually standardize on open interconnect - the same way it standardized on Ethernet for data center networking a decade ago, displacing proprietary alternatives in the process. For enterprise buyers planning multi-vendor GPU environments - a growing reality as AMD and Intel chip alternatives gain traction for inference workloads - the QFX5252 is the first switch that delivers a production-grade open path to scale-up GPU connectivity not owned by Nvidia.

3. Dell and HPE Both Pursuing Quantum-Classical AI Data Centers - Equal1's RacQ Demonstrated at DTW 2026

A July 13 analysis from The Next Platform highlights that both Dell Technologies and HPE are now actively pursuing hybrid quantum-classical computing architectures for HPC and AI data centers, signaling that quantum computing is moving from laboratory demonstration toward enterprise rack-scale integration. At Dell Technologies World 2026 in May, Equal1 demonstrated its RacQ quantum computer - a silicon-spin qubit system that fits in a standard Dell 42U rack frame, runs at 0.3 Kelvin via an integrated cryocooler, and operates at approximately 1.6 kilowatts. Dell CTO John Roese stated: "I could roll that in right next to an Nvidia NVL72 rack and connect the two things together and have all the basic compute to do a CPU-GPU-QPU system."

HPE's approach is broader in scope. At HPE Discover 2026, the company announced partnerships with Intel, IQM Quantum Computers, Quantinuum, QuEra, and Rigetti Computing to build an integration platform for quantum systems of multiple modalities within its Cray supercomputer family. HPE SVP Trish Damkroger described HPE's role as the integration point for CPU, GPU, and quantum hardware in a unified rack infrastructure. The combined strategy across both vendors suggests that by 2027-2028, enterprise data centers will need to plan for a three-tier compute environment - CPU, GPU, and QPU - running as a unified system. For networking teams, the implication is significant: interconnects between classical GPUs and quantum coprocessors introduce new latency and interface requirements that current InfiniBand and Ethernet fabric designs were not built to address, opening a new hardware design challenge that both HPE Juniper and Arista will need to navigate.

4. Arista Q2 2026 Results Confirmed for August 4 - Analysts Expect First $2.8B+ Quarter

On July 7, Arista Networks officially confirmed that Q2 2026 financial results will be released after U.S. market close on Tuesday, August 4, 2026, with an executive call at 1:30 PM PT. The announcement coincides with Arista's first-ever inclusion on the Fortune 500 list - a milestone the company marked this month - and comes two weeks after the stock surged 7% on July 8 following analyst upgrades from KeyBanc, Morgan Stanley, and Bank of America with price targets moving toward $200.

Wall Street consensus for Q2 2026 sits at approximately $2.8 billion in revenue, with gross margin between 62% and 63% and non-GAAP EPS of approximately $0.88. Achieving this would represent 29% year-over-year growth and a 3.3% sequential increase from Q1's record $2.709 billion. The key swing factor is supply: Arista management noted at Q1 that "supply is now the binding constraint," with $8.9 billion in purchase commitments and $6.2 billion in deferred revenue waiting to convert into shipped and recognized product. Analysts are watching for any update on component lead-time improvements - particularly in switch silicon and optics - that would signal whether Arista can accelerate shipment velocity in H2 2026. The AI networking TAM framing from the Bank of America June conference ($105 billion today, $160 billion by 2030) gives the August 4 report elevated stakes: this is the first quarterly print since that TAM expansion was articulated publicly, and Arista's 1.6T 7060XE7 Series - validated by Microsoft, Meta, and Oracle - will be the key product commentary to watch.

5. Dell PowerSwitch SN6000 Now Globally Available - Nvidia Spectrum-6, 1.6 Tbps, CPO-Ready

As announced at GTC 2026 in March and scheduled for July 2026 global availability, the Dell PowerSwitch SN6000 series is now shipping to customers worldwide. Built on Nvidia Spectrum-6 silicon, the SN6000 is Dell's flagship AI network switch, delivering 1.6 Tbps switching capacity with support for up to 512 ports of 800G connectivity. It is the first Dell switch to offer co-packaged optics (CPO) as a standard configuration option - the same technology Nvidia confirmed is now in production with Spectrum-X Ethernet Photonics (covered in our June 2 edition). Compared to switches using traditional pluggable transceivers, Nvidia's published benchmarks for the CPO variant cite 5x better power efficiency, 10x higher reliability, and 5x longer AI uptime.

The SN6000's July availability is significant for enterprise buyers who had been evaluating it since March. The switch is managed through Dell's Integrated Rack Controller and is the core networking component of Dell PowerRack - the pre-validated AI infrastructure rack system set for general availability in September 2026. Together with the PowerCool CDU C7000 liquid cooling unit (Q3 2026) and PowerRack (September), the SN6000 availability now means the complete Dell AI Factory rack-scale infrastructure stack - hardware, networking, and cooling - will be commercially complete for the first time before Q3 2026 ends. For procurement teams evaluating on-premises AI deployment timelines, this convergence marks the practical starting point for full-scale Dell AI Factory deployments outside of early-adopter configurations.

Editor's Summary

The week of July 9-14, 2026 is a consolidation week between two major product cycles, but the data and strategic signals it delivers are high value. IDC's Q1 2026 Ethernet switch figures confirm Nvidia's rise to the top of data center switching - driven by its end-to-end AI factory positioning - while also showing that Arista's 92% data center revenue concentration makes it the most operationally leveraged incumbent to the same AI infrastructure spending wave. The HPE QFX5252 UALink story is the most underreported development of the past month: HPE shipping the first production switch supporting UALoE means the open alternative to Nvidia NVLink has moved from specification to shipped hardware - a structural inflection that will matter most when AMD and Intel GPU deployments reach a scale where proprietary interconnect lock-in becomes a procurement constraint. The quantum-AI story is longer-horizon but directionally clear: both Dell and HPE are publicly committing to CPU-GPU-QPU rack architectures. And Dell PowerSwitch SN6000 shipping this month, with Arista's Q2 earnings on August 4, sets up the next three weeks as an important checkpoint for how the AI networking cycle is actually tracking against 2026 projections.

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)

Who is the number-one data center Ethernet switch vendor by revenue in Q1 2026?

According to IDC's Quarterly Ethernet Switch Tracker for Q1 2026, Nvidia became the number-one data center Ethernet switch vendor by revenue for the first time, generating $2.1 billion in quarterly switch revenue - up 192.7% year-over-year - and capturing 21.5% of the data center Ethernet switch segment. This is up from less than 4% share in Q1 2024. The data center Ethernet switch market reached $10 billion in Q1 2026, up 61% year-over-year. Arista Networks held second place in the DC segment at 20.7% share.

What is UALink-over-Ethernet (UALoE) and why is the HPE Juniper QFX5252 important?

UALink (Ultra Accelerator Link) is an open-standard GPU scale-up interconnect developed by a consortium of 85-plus companies - including HPE, AMD, Google, Microsoft, Apple, Cisco, Intel, Meta, and AWS, but notably not Nvidia - as a direct alternative to Nvidia's proprietary NVLink. UALink-over-Ethernet (UALoE) implements this open standard over Ethernet fabric. The HPE Juniper Networking QFX5252 is the first production switch to ship with native UALoE support, making it the first commercial hardware to provide an open, vendor-neutral path for high-bandwidth GPU-to-GPU scale-up connectivity. For enterprises running multi-vendor GPU environments with AMD Instinct or other non-Nvidia accelerators, this means production-grade GPU scaling without proprietary Nvidia hardware lock-in.

What is hybrid quantum-classical computing and how are Dell and HPE pursuing it?

Hybrid quantum-classical computing integrates traditional CPU and GPU compute with quantum processing units (QPUs) in a unified rack system, enabling workloads that benefit from quantum acceleration - such as optimization, simulation, cryptography, and drug discovery. Dell Technologies demonstrated Equal1's RacQ silicon-spin qubit computer in a standard 42U Dell server rack at Dell Technologies World 2026, with Dell CTO John Roese describing it as compatible with side-by-side NVL72 GPU rack deployment. HPE announced partnerships with IQM, Quantinuum, QuEra, Rigetti, and Intel at HPE Discover 2026 to build quantum integration platforms for its Cray supercomputer family. Both vendors are working toward rack-scale CPU-GPU-QPU systems for enterprise data centers by 2027-2028.

When will Arista Networks report Q2 2026 earnings and what is expected?

Arista Networks will report Q2 2026 financial results after U.S. market close on Tuesday, August 4, 2026, with an executive conference call at 1:30 PM PT. Wall Street consensus is approximately $2.8 billion in revenue, with gross margin of 62-63% and non-GAAP EPS of approximately $0.88 - representing about 29% year-over-year growth. Supply constraints remain the primary variable: Arista has $8.9 billion in purchase commitments and $6.2 billion in deferred revenue waiting to convert into shipped product, making component availability commentary the key swing factor for the print.

What is the Dell PowerSwitch SN6000 and is it now available?

The Dell PowerSwitch SN6000 is Dell's flagship AI network switch, built on Nvidia Spectrum-6 silicon, now globally available as of July 2026. It delivers 1.6 Tbps switching capacity with up to 512 ports of 800G connectivity and supports co-packaged optics (CPO) for 5x better power efficiency, 10x higher reliability, and 5x longer AI uptime versus traditional pluggable transceiver switches. It is managed via Dell's Integrated Rack Controller and serves as the networking layer of Dell PowerRack - Dell's turnkey AI infrastructure rack available from September 2026. Together with the PowerCool CDU C7000 (Q3 2026), the SN6000's availability completes the full Dell AI Factory rack-scale hardware stack.

Sources