Arista Q2 2026 Revenue Tops $3 Billion, but AI Mix Is Undisclosed | Magica
Arista’s First $3 Billion Quarter Shows Growth, Not Yet an AI-Revenue Mix
Editorial Team
••📖5 min read
Arista Networks reported $3.036 billion in Q2 2026 revenue and forecast about $3.3 billion for Q3. The filing confirms rapid product-led growth and strong operating margins, but it does not break out sales from AI fabrics, customers or other end markets—leaving the breadth and durability of the AI thesis unproven.
Arista reported its first quarter above $3 billion in revenue, with Q2 sales up 37.7% from a year earlier.
Product sales accounted for about 86% of Q2 revenue and roughly 88% of its year-over-year revenue increase; the filing does not identify how much came from AI fabrics.
The $3.3 billion Q3 target is a company forecast. Its credibility will depend on results, mix and the concentration and supply risks Arista flags.
Arista Networks, a supplier of client-to-cloud networking for large AI, data-center, campus and routing environments, reported $3.036 billion in second-quarter 2026 revenue, its first $3 billion quarter. The company forecast approximately $3.3 billion in third-quarter revenue, a target that would be about 8.7% higher than the just-reported quarter if achieved.
The result establishes a larger sales base. It does not establish that AI-fabric products are the driver, or that demand is broadly distributed. Arista reports product and service revenue, but not revenue by AI, data-center, campus or routing use case, customer, or new platform.
The financial result is clear; the end-market mix is not
Revenue rose $830.9 million from the year-earlier quarter, based on the company’s unaudited statements. Product revenue increased $728.2 million to $2.6052 billion, while service revenue increased $102.7 million to $430.5 million. That makes products 85.8% of Q2 revenue and 87.6% of the year-over-year increase; those percentages are calculations from the reported product, service and total-revenue lines.
Samsung has shown an HBM5 mock-up built around a 2nm base die and a new heat-management design, but it has not provided a production timetable or comparable performance data. Its nearer challenge is turning HBM4E samples into qualified, high-volume supply as rivals ship competing HBM products and expand capacity.
Editorial Team
Product revenue
$2.6052 billion
$1.8770 billion
$728.2 million
Service revenue
$430.5 million
$327.8 million
$102.7 million
GAAP operating margin
45.4%
44.7%
0.7 percentage points
GAAP gross margin
62.9%
65.2%
-2.3 percentage points
The financial release also reported $0.95 in GAAP diluted earnings per share and $1.02 on its non-GAAP measure. The reported revenue exceeded the $2.82 billion LSEG consensus cited in market coverage by about $216 million, or 7.7% of that estimate.
That is a useful measure of the quarter’s surprise, not a measure of AI-specific demand. Chantelle Breithaupt, Arista’s chief financial officer, called the growth “broad-based” in the company’s release, but the financial tables provide no segment or customer data with which to test that characterization.
New AI-fabric hardware is a product claim, not a sales breakdown
Jayshree Ullal, Arista’s chairperson and chief executive, has led the company since her 2008 appointment. Before joining Arista, she was a Cisco senior vice president responsible for its $10 billion data-center, switching and services group, according to Arista’s appointment announcement. Her role and that background explain why the company’s current argument centers on networking infrastructure for large computing installations.
Alongside the results, Arista said it had introduced its 7060XE7 Series, a 1.6-terabit-per-second Etherlink portfolio for AI fabrics. The company says the platforms can deliver up to 100 Tbps of system bandwidth and that support for Linear Pluggable Optics can reduce interconnect power use by approximately 60% versus traditional pluggable optics. Those are Arista product specifications and a company comparison, not independently reported deployment results or a price comparison.
The release also cites techniques for scale-up, scale-out and scale-across fabrics, including multi-planar leaf-spine designs, Multipath Reliable Connection and segment routing over IPv6. It does not identify customer deployments, contract values, sales of the new series, or comparable offerings from named competitors. The strongest supported conclusion is therefore narrower than an AI-demand victory: Arista has placed new hardware and networking approaches next to a quarter of strong overall sales.
Margin strength sits beside real operating constraints
Arista’s GAAP operating margin improved despite the 2.3-percentage-point decline in GAAP gross margin. Its 49.9% non-GAAP operating margin should not be treated as interchangeable with the GAAP figure: the company excludes stock-based compensation, intangible-asset amortization, strategic-investment gains or losses and related tax effects, and it recast prior non-GAAP periods after changing its tax methodology in late 2025.
For Q3, Arista forecast a 48% to 49% non-GAAP operating margin and non-GAAP diluted earnings per share of $1.06 to $1.08. A market report said those forecasts compared with a $2.94 billion revenue estimate and 91-cent earnings-per-share estimate, respectively, while attributing those consensus figures to Benzinga Pro.
The company’s own forward-looking-statement disclosure supplies the counterweight to the forecast: it cites large purchases from a limited number of customers, reliance on a predominant merchant-silicon vendor, sole- or limited-source components, third-party manufacturers, tariffs and export controls, and potentially reduced infrastructure spending. The disclosure establishes exposure to those risks, not that any one has occurred or will change Q3 results.
What the next report needs to answer
The next decision point is the Q3 report. First, readers can test whether actual revenue reaches the company’s approximately $3.3 billion target and whether the forecast non-GAAP margin and earnings range hold.
More importantly, evidence for a durable AI-networking expansion would require information the Q2 release does not provide: sales or deployment evidence for the new AI-fabric portfolio; an explanation of demand across end markets and customers; and enough detail to assess whether customer concentration, component availability and infrastructure budgets are constraining the higher run rate. Until then, the first $3 billion quarter is a verified financial milestone, not a disclosed measure of AI-fabric adoption.
Anthropic is reported to have agreed to spend $10 billion over six years on Volta computing capacity. The disclosed Norwegian lease linked to Volta clarifies the opportunity—and the construction, financing and delivery conditions that still stand between a headline commitment and usable AI compute.
SoftBank has drawn $20 billion of a $40 billion bridge facility for its OpenAI investment and expects another $10 billion draw in October. Its Aug. 6 earnings release should show whether the group can turn a stated menu of asset-backed financing, bonds and asset sales into longer-term funding.
A newly published review of Phoenix Ikner’s yearlong ChatGPT record describes crisis-support replies alongside detailed answers in the hours before the 2025 Florida State University shooting. The record raises hard questions about OpenAI’s safeguards and account-level risk detection, but it does not establish causation or reveal what the company’s internal systems knew.
Patrick Steven Yaroch, a former FBI supervisory special agent who worked counterintelligence, is accused of using bureau-held wallet information to transfer cryptocurrency into accounts he controlled. The charging records make ChatGPT conversations part of the government’s evidence, but not the alleged means of moving the money.
Caterpillar raised its 2026 sales outlook after quarterly revenue first topped $20 billion. A surge in power-generation demand points to the data-center buildout, but stronger construction sales, pricing and a growing dealer inventory make the company’s record backlog an incomplete measure of that boom.
Cloudflare is taking reservations for cloudflare.pay handles and has outlined stablecoin wallets that would let account owners constrain AI-agent spending. The wallet is not yet available, and Mastercard’s broader machine-payments program shows that identity, limits and settlement are already a competitive infrastructure race.
Apple is seeking expedited discovery before a ruling on its preliminary-injunction request against OpenAI and former Apple hardware employees. The filing adds allegations involving other former employees, while OpenAI calls the case factually wrong; the court must decide whether Apple’s proposed evidence-gathering is justified and proportionate.
HappyRobot's $150 million Series C values the enterprise AI-agent company at $1.2 billion and funds expansion beyond logistics. The company says its deployment model can automate complex operational work, but the disclosures leave its reliability, pricing and cross-sector repeatability unmeasured.
Bending Spoons has agreed to acquire Airtable in an all-cash transaction with a $1.285 billion enterprise value. Against Airtable’s company-reported $480 million in annual recurring revenue, that is about 2.7 times ARR—a more useful, though still limited, view of the transaction than a simple comparison with the company’s 2021 funding valuation.
Wisconsin Assembly member and gubernatorial candidate Francesca Hong proposes a statewide pause on new AI data-center construction while lawmakers write rules on utilities, environmental protections, subsidies, labor and local approval. The proposal puts a clear choice before the state: set those terms before more projects advance, or keep local dealmaking open while regulations are debated.
Microsoft paid more than $20 million to 562 security researchers in the year ended June 30, a company record. But the total also reflects a broader eligibility policy, a live hacking event and a higher volume of AI-assisted research, leaving the program's operational performance unmeasured.
AMD reported $11.536 billion in second-quarter revenue as data-center sales more than doubled, but investors’ focus has shifted to whether its Helios rack-scale systems can turn that demand into shipments and cash generation on the timetable AMD projects.
The Justice Department says OpenAI and its subsidiary Statsig will pay $3.2 million to resolve allegations that they disadvantaged U.S. workers during recruitment for fewer than 10 visa-linked jobs. The agreement combines penalties, a back-pay fund and required changes to job postings and applications.
A 64-year-old Pauillac photographer has been placed under formal investigation and remanded in custody after a partial device review found several thousand suspected child sexual abuse files. The central unanswered question is which children, if any, can be matched to the alleged AI-altered material.
A reported White House framework would offer U.S. officials a 30-day pre-release review of some closed frontier AI models while exempting open models. The government has not published the rules, leaving its thresholds, participants and consequences unclear.
The Ninth Circuit lifted a preliminary injunction that had barred Perplexity’s Comet assistant from using Amazon.com, holding that Amazon was unlikely to show Perplexity itself accessed its computers. The decision is limited to the current record and leaves Amazon’s lawsuit, its terms of service and wider questions about AI-agent liability unresolved.
An Aug. 1 Arena snapshot put Alibaba's Qwen3.8 Max fourth in front-end web development at 1,668 points. The preliminary, task-specific result is a useful signal, but it cannot by itself verify Alibaba's broader claim for Qwen3.8—especially because the source record separately names a July Preview release and a Qwen3.8 Max leaderboard entry.
Goodfire’s Silico packages mechanistic-interpretability experiments into software for organizations that can access a model’s internal parameters. The launch broadens a specialist workflow, but its price, operating cost and ability to deliver reliable control outside demonstrations remain unproven.
SpaceX’s first public quarterly report showed 92% revenue growth and a smaller-than-expected loss, but six-month cash flows show that its expanding launch, connectivity and AI plans still depend far more on financing than operations.
