Arista Networks delivered its 19th consecutive record quarter with Q3 2025 revenue of $2.31 billion, representing 27.5% year-over-year growth. Non-GAAP gross margin came in at 65.2%, with operating margins reaching 48.6% and diluted EPS increasing 25% to $0.75. Performance was driven by strong momentum in AI and Cloud Titan sectors, with software and services contributing 18.7% of revenue. The company raised its full-year 2025 revenue outlook to approximately $8.87 billion (26-27% growth) and set a 2026 target of $10.65 billion (20% growth), citing an 'undeniable and explosive AI megatrend' and a total addressable market expanding to over $100 billion. Despite supply chain lead times of 38-52 weeks, management expressed high confidence in demand, highlighting record cash flow from operations of $1.3 billion and a remaining share repurchase authorization of $1.4 billion.
| Metric | Value | Change |
|---|---|---|
| Q3 Revenue | $2.31 Billion | +27.5% YoY |
| Q3 Non-GAAP Gross Margin | 65.2% | +0.6% pp YoY |
| Q3 Operating Margin | 48.6% | N/A |
| Q3 EPS (Diluted) | $0.75 | +25% YoY |
| Cash & Investments | $10.1 Billion | N/A |
| Purchase Commitments | $7.0 Billion | +$1.3B QoQ |
| Deferred Revenue | $4.7 Billion | +$0.6B QoQ |
| Operating Cash Flow | $1.3 Billion | N/A |
Arista is aggressively positioning itself as the primary alternative to proprietary InfiniBand solutions for AI back-end networks. Management highlighted the 'sea change' in 2025 where they are increasingly invited to construct 800G and future 1.6T back-end networks, moving from 'outside looking in' to a central participant. The introduction of the Etherlink portfolio and leadership in the Ultra Ethernet Consortium (UEC) and Ethernet Scale-Up Networking (ESUN) standards indicates a strategic shift to capture the 'Scale-Up' and 'Scale-Across' AI use cases, which management believes will materialize significantly in 2026 and 2027.
The company is executing a 'Good, Better, Best' strategy to combat white-box commoditization. While acknowledging that commodity white boxes are 'good enough' for simple use cases, Arista is pushing 'Blue Boxes' (hardware with NetDI but no full EOS software license) for high-performance AI workloads where customers want operational control but need Arista's hardware reliability. This strategy aims to defend market share against merchant silicon vendors while maintaining a path to higher-margin 'Best' branded EOS sales. Management noted this Blue Box mix is factored into the 2026 gross margin guidance of 62-64%.
Significant organizational restructuring signals a dual focus on deepening AI penetration and expanding enterprise reach. The promotion of Kenneth Duda to President and CTO and the hiring of Tyson Lamoreaux (formerly AWS) to lead Cloud and AI segments demonstrate a commitment to engineering-led customer intimacy for the largest AI builders. Simultaneously, the elevation of Todd Nightingale to COO and the integration of the VeloCloud acquisition highlight a strategic priority to capture the $30 billion campus and enterprise TAM, leveraging AVA (Autonomous Virtual Assist) to differentiate through AI-driven network operations.
Arista is evolving its business model toward full-stack 'rack-level' solutions, moving beyond individual switches. CEO Jayshree Ullal described involvement in 5-7 projects integrating cabling, co-packaging, power, and cooling with various XPU cartridges. This shift implies a move up the value chain, potentially adopting a JDM (Joint Design Manufacturing) model for IP contribution rather than just OEM manufacturing. This positions Arista as a systems integrator for AI data centers, potentially locking in customers and increasing switching costs, though it may introduce margin variability depending on the mix of hardware vs. software sold.
Sequential growth is decelerating rapidly, from 10% in Q2 to approximately 5% in Q3, with Q4 guidance implying only ~1.6% growth at the midpoint. While management attributes this entirely to supply constraints ('demand is greater than our ability to ship'), the persistence of this slowdown into Q4 raises concerns about whether supply issues are being resolved or if demand digestion is occurring. The guidance for Q4 revenue of $2.3-$2.4B suggests a significant flattening of growth trajectory heading into 2026.
Gross margins are facing structural headwinds from a shift in product mix toward lower-margin AI/Cloud hardware and potential tariffs. Q4 guidance calls for GM to drop to 62-63%, and 2026 guidance is 62-64%, down from current levels. Management admitted that 'Blue Box' sales (hardware without full software) carry lower margins and that heavy mixes with Cloud/AI titans depress margins. This indicates that the massive AI revenue growth may be dilutive to overall profitability in the near term.
Supply chain lead times remain extended at 38-52 weeks, forcing the company to carry significant inventory risk. Purchase commitments have ballooned to $7 billion, up from $5.7 billion in Q2. While this indicates strong backlog, it also creates risk if customer acceptance timelines slip or if technology specifications change during long lead times. Management noted that AI deployments do not follow traditional seasonality and 'land when they land,' creating potential for quarterly volatility and earnings misses.
The core non-AI business appears to be stagnating or growing slowly, acting as a drag on the overall growth rate. When discussing the 2026 outlook, management implied that the 'core business' (outside of AI and Campus) would be 'flattish' or 'single digit' growth. This suggests that Arista's legacy high-margin cloud networking business is maturing or saturating, placing immense pressure on the AI and Campus segments to execute perfectly to maintain the company's 20% growth targets.
Overall: Management exhibited a highly confident and assertive demeanor throughout the call, particularly CEO Jayshree Ullal, who actively pushed back against analyst narratives regarding growth deceleration. The tone shifted from celebratory regarding product milestones to defensive when questioned about sequential growth rates and margin guidance, yet remained resolute in the long-term vision. There was a clear emphasis on 'pragmatic' optimism, balancing excitement over AI opportunities with operational discipline.
Confidence: HIGH - Management consistently reiterated that demand exceeds supply, committed to aggressive 2026 growth targets (20% CAGR) despite a larger revenue base, and made specific references to 'record' backlogs and customer engagement levels. The defensiveness regarding 'deceleration' suggests they view the current dip as a temporary supply/timing issue rather than a demand problem.
$2.30 billion - $2.40 billion
62% - 63%
~$8.87 billion (26-27% growth)
At least $1.5 billion
$10.65 billion (20% growth)
62% - 64%
$2.75 billion
Hedging & Uncertainty: Management utilized specific temporal and probability hedges to manage expectations regarding the timing of AI revenue. Phrases like 'I don't know whether it will land in '26 or '27' and 'it lands when it lands' were used to explain the lack of traditional seasonality and the unpredictability of large AI deal closures. They also used qualifiers like 'variability' and 'lumpiness' to describe quarterly results, effectively hedging against future revenue misses by blaming supply chain logistics rather than demand weakness. However, they were less hedged on long-term targets, firmly stating 'we are committed to' 20% growth.
We are experiencing a golden era networking with an increasing TAM now of over $100 billion in forthcoming years. - Jayshree Ullal, CEO
I don't like the word deceleration... call it, variability across quarters, and demand is great. - Jayshree Ullal, CEO
The demand and scale of AI build-outs is clearly unprecedented. - Jayshree Ullal, CEO
We're talking about big, big numbers here, guys. - Jayshree Ullal, CEO
It's all about good, better and best, where in some simple use cases, a commodity white box is good enough. - Jayshree Ullal, CEO
We're feeling pretty good on our large GPU deployments... At the same time, the variability I was stating is demand is greater than our ability to ship. - Jayshree Ullal, CEO
We're putting some goals for ourselves with the AI... It doesn't mean we're not focused on the rest. But I don't think it's the right approach to assume everything is going to be 100%. - Chantelle Breithaupt, CFO
Analyst Sentiment: Analysts were skeptical regarding the sharp deceleration in sequential growth and the sustainability of margins amidst a product mix shift. Questions focused heavily on distinguishing between supply constraints and true demand softness, as well as the financial impact of the new 'Blue Box' strategy.
Management Responses: Management was combative but informative, repeatedly correcting analysts' use of the word 'deceleration' and insisting that demand remains robust. They provided detailed technical explanations to justify margin guidance and emphasized that supply chain lead times (38-52 weeks) are the primary bottleneck to revenue recognition.
Analysts pressed heavily on the sequential growth slowdown (Q2 10% -> Q3 5% -> Q4 ~1.6%), questioning if this signaled a broader market pause. Management attributed this entirely to an inability to ship due to supply constraints, citing $7B in purchase commitments as proof of demand.
There was significant focus on 'Blue Box' economics and how the shift to selling hardware without full software (EOS) would impact 2026 margins. Management clarified this is a strategic necessity for scale-up AI workloads and is already factored into the 62-64% GM guide.
Questions arose regarding the convergence of front-end and back-end networks. Management explained that while back-end AI networks are the primary driver, the sheer scale is forcing upgrades to front-end cloud infrastructure, creating a 'halo effect' for Arista's broader portfolio.
Arista Networks remains a premier play on the AI infrastructure build-out, successfully leveraging its relationships with hyperscalers to capture share in both back-end AI and front-end cloud networks. While near-term margins face pressure from mix shifts and tariffs, and supply chain constraints are causing lumpy quarterly results, the fundamental demand drivers are intact. The company's strategic pivot to 'Scale-Up' networking via the Etherlink portfolio and its leadership in open standards (UEC) position it well for the next phase of AI growth. The 2026 target of $10.65B implies a 20% CAGR, which, given the $100B+ TAM opportunity and management's confident execution history, supports a positive long-term view despite current valuation concerns.
Management described AI build-outs as 'unprecedented' and 'explosive,' with customers planning facilities in terms of 'gigawatts' rather than megawatts. This indicates a sustained, multi-year capex cycle driven by LLM training and inference.
Lead times for key components remain elevated at 38-52 weeks. This is acting as a bottleneck on revenue recognition, forcing Arista to hold $7B in purchase commitments to secure supply, which ties up capital but signals strong future revenue.
Management cited 'known tariff scenarios' as a specific factor impacting Q4 and FY25 gross margin guidance (62-63%), indicating that trade policy is a tangible headwind to profitability.