Arista's Q3 2025: Contradictions Emerge on Growth Projections, AI Market Share, and Supply Constraints

Date of Call: September 30, 2025

Financials Results

• Revenue: $2.31B, up 27.5% YOY, above guidance of $2.25B
• EPS: $0.75 per diluted share, up 25% YOY (diluted shares 1.277B)
• Gross Margin: 65.2% (non-GAAP), above guidance of 64%, down from 65.6% sequentially, up from 64.6% YOY
• Operating Margin: 48.6% of revenue

Guidance:

• Q4 revenue expected $2.3B–$2.4B; gross margin 62%–63% (inclusive of known tariff scenarios).
• Q4 operating margin ~47%–48%, effective tax rate ~21.5%, diluted shares ~1.281B.
• FY25 revenue growth ~26%–27% (~$8.87B midpoint); campus $750M–$800M; AI center at least $1.5B.
• FY26 outlook: ~20% revenue growth to $10.65B; campus $1.25B; AI center $2.75B; gross margin ~62%–64%; operating margin ~43%–45%.

Business Commentary:

• Strong Revenue Growth and AI Focus:
• Arista Networks reported revenue of $2.31 billion for Q3 2025, with software and services contributing 18.7% of revenue.
• The growth was driven by AI-specific products and technologies, reflecting the company's strategic focus on AI networks and partnerships with major cloud titans and AI companies.

• Non-GAAP Financial Performance:
• The non-GAAP gross margin for Q3 was 65.2%, influenced by favorable mix benefits.

• This performance was supported by strong enterprise segment strength and favorable product mix impacts.

• Leadership and Organizational Changes:

• Arista announced the promotion of Ken Duda to President and Chief Technology Officer and Todd Nightingale to President and Chief Operating Officer.

• These appointments reflect the company's emphasis on growth and innovation across its technical and operational leadership.

• Supply Chain Constraints and Demand Outpaces Supply:

• Arista experienced variability in shipment levels due to long lead times for components like memory and merchant silicon, ranging from 38 to 52 weeks.
• Despite supply constraints, strong demand, particularly in AI, remains consistent, and the company is committed to achieving 20% revenue growth in 2026.
  

  Sentiment Analysis:

  Overall Tone: Positive

  • Management described a "19th consecutive record quarter," reported nearly $2.31B revenue and 65.2% non-GAAP gross margin, and reiterated confidence in AI momentum with a FY26 revenue target of $10.65B and an AI center target of $2.75B, emphasizing strong demand and continued investment.

Q&A:

• Question from Tal Liani (Bank of America): If sequential growth has decelerated this year (10% to 5% to 1.6%), what are the underlying drivers and should we be concerned about future growth?
  Response: Management: No demand concern—apparent deceleration reflects shipment constraints (supply/lead-times), not weakening demand; company remains committed to 20% growth targets.

• Question from Aaron Rakers (Wells Fargo): Can you unpack Q4 gross margin drivers — how much is tariff related versus mix, and does this change expectations going forward?
  Response: Management: Guidance reflects product-margin pressure from a heavy cloud/AI mix (product margins can be <60% when cloud/AI dominates); mix, not a structural change, drives margin variability.

• Question from Michael Ng (Goldman Sachs): How is Arista positioned for full-rack solutions and the convergence of compute and networking — partnership model vs direct offering?
  Response: Management: Engaged in multiple rack projects via partnerships, JDM/reference designs and both blue-box and Arista-branded offers; expect material emergence in 2026–2027.

• Question from Atif Malik (CITI): For LLM providers (OpenAI, Anthropic) linked to cloud titans, who drives networking hardware decisions and is Arista's share stable?
  Response: Management: Decisions are joint between LLM developers and cloud titans; Arista's share remains stable and strong versus bundlers and white-box alternatives.

• Question from Sammik Chatterjee (J.P. Morgan): Is the quarter-to-quarter variability due to changes in tier-one customers' cluster plans or supply-driven issues?
  Response: Management: Variability is largely supply-driven—component lead times (notably standard memory/merchant silicon) can be very long (38–52 weeks), so demand outstrips ability to ship at times.

• Question from Amit Daryanani (Evercore ISI): What is driving the apparent deceleration from high-20s to ~20% growth in 2026 — why not continued acceleration given purchase commitments?
  Response: Management: Rejects 'deceleration' framing—expects sustained double-digit/20%+ growth but timing is lumpy due to project acceptance and customer timing; long-term targets unchanged.

• Question from David Vogt (UBS): Outside AI and campus, how should we think about growth in the core business through 2026?
  Response: Management: AI and campus expected to grow substantially; core business may be flat or single-digit growth but remains strategically important and receives investment.

• Question from Ben Ritchie (Melius Research): Confidence in participating in private LLM and emerging hyperscaler builds—will Arista capture a lot of that growth?
  Response: Management: Strong participation today in scale-out and scale-across; scale-up participation will evolve more slowly (likely into 2027) but company is an important participant.

• Question from Tim Long (Barclays): Economics and deployment scenarios for Blue Box as it grows — margins and where it will be used?
  Response: Management: Blue Box initially carries lower margins than full EOS-branded systems but economics improve with scale; best suited for sophisticated cloud/NeoCloud and certain scale-up cases.

• Question from Meta Marshall (Morgan Stanley): How are front-end (inference) network upgrades progressing relative to expectations from a year ago?
  Response: Management: Back-end is moving to 800G→1.6T, which pressures front-end upgrades; Arista uniquely addresses both and benefits from the convergence of front- and back-end networks.

• Question from Carl Ackerman (BNP Paribas): How should we think about disaggregated scheduled fabrics (DSF) versus non-scheduled fabrics and customer adoption?
  Response: Management: Both architectures have massive production deployments; Arista supports both, offering customers choice and consistent software across topologies.

• Question from Simon Leopold (Raymond James): What Blue Box assumptions are baked into 2026 margin guidance and how big will Blue Box adoption be in 2026?
  Response: Management: Blue Box is expected to remain limited to a small set of sophisticated customers (single-digit to low-double-digit customers), is factored into 2026 margin guidance, and will not be mainstream volume next year.

• Question from Sebastian Najarian (William Blair): Where are you investing sales & marketing for enterprise growth—geography, channel, cross-sell?
  Response: Management: Investing in campus (VeloCloud), geographic expansion (notably Asia), channel and new-logo acquisition to attack large white-space opportunities.

• Question from Ben Boland (Cleveland Research): How have customer engineering/delivery lead times evolved and how confident are you in meeting needs over 12–18 months?
  Response: Management: Rely on early customer forecasts and larger purchase commitments to plan; confident for large/long-lead AI projects but continuing to improve campus lead times and monitor component constraints.

Contradiction Point 1

Growth and Demand Variability

It highlights discrepancies in the explanation of growth and demand variability, which are crucial for understanding Arista's financial performance and strategic positioning.

Why is growth decelerating compared to last year's consistent growth? Should we be concerned about future growth? - Tale Liani(Bank of America)

2025Q3: No concern on demand, but variations in shipments due to supply constraints. Demand is strong, and we're able to ship more when supply aligns. - Jayshree Ullal(CEO)

What factors enabled Arista to raise its growth guidance to 25% from 17%? - Amit Daryanani(Evercore ISI)

2025Q2: Another incredibly strong quarter as our enterprise campus continues to record demand and AI investments from large titans and enterprises contribute significantly. - Jayshree V. Ullal(CEO)

Contradiction Point 2

AI and Cloud Market Share

It addresses potential inconsistencies in Arista's market share and penetration in the AI and cloud sectors, which are key growth drivers for the company.

How confident are you in participating in large AI builds, and what factors are affecting your confidence? - Ben Ritchie(Melius Research)

2025Q3: We believe we will exceed $10 billion in 2026. - Jayshree Ullal(CEO)

Is cloud growth driven by front-end or back-end demand? - Meta A. Marshall(Morgan Stanley)

2025Q2: We continue to execute on our strategy in the cloud, and we believe we have increased our market share in the cloud this quarter and will continue to do so. - Jayshree V. Ullal(CEO)

Contradiction Point 3

Demand and Supply Constraints

It directly impacts expectations regarding the company's ability to meet demand and the potential impact on revenue and investor expectations.

Why is growth slower compared to last year's consistent rate and current sequential rate? Should we worry about future growth? - Tale Liani (Bank of America)

2025Q3: No concern on demand, but variations in shipments due to supply constraints. Demand is strong, and we're able to ship more when supply aligns. Continued commitment to 20% growth reflects confidence in demand. - Jayshree Ullal(CEO)

Could you clarify how tariffs impact your revenue outlook? - David Vogt (UBS)

2025Q1: When we began dealing with tariffs, we had initially planned to move our production out of Mexico due to anticipated tariffs. However, the situation has changed, and we are grateful for the pause in tariffs until July 9. - Jayshree Ullal(CEO)

Contradiction Point 4

AI back-end Deployments

It involves the progress and expectations for AI back-end deployments, which are critical for the company's growth strategy.

How confident are you in participating in large AI builds, and what factors are influencing your confidence? - Ben Ritchie (Melius Research)

2025Q3: Arista's participation is strong in scale-out and scale-across. Scale-up will take longer, mainly proprietary technologies. Confident about exceeding $10 billion in 2026, AI a significant part. - Jayshree Ullal(CEO)

Can you update us on the four Tier 1 customers for AI back-end deployments? - Amit Daryanani (Evercore)

2025Q1: All four customers are progressing well with AI deployments. Two are expected to reach 50,000 GPU deployments by year-end, with some potentially reaching 100,000. The fourth is well underway. The front-end to back-end ratio remains generally 1:1. - Jayshree Ullal(CEO)

Contradiction Point 5

AI Market TAM and Arista's Role

It involves the company's perception and forecasts for the AI market, which is a critical growth area for Arista, affecting investor expectations and strategic planning.

What is causing the slowing growth, given that product commitments indicate growth? - Amit Daryanani (Evercore ISI)

2025Q3: AI market is roughly 1/3 of the $70 billion TAM. The 750 million in AI back-end is driven by customer deployments, with confidence in achieving this in FY 2025. - Jayshree Ullal(CEO)

What is the upside potential for the $750 million AI back-end sales goal with the stalled fifth customer? How does AI align with the $70 billion TAM by 2028? - Atif Malik (Citi)

2024Q4: AI market is roughly 1/3 of the $70 billion TAM. The 750 million in AI back-end is driven by customer deployments, with confidence in achieving this in FY 2025. - Jayshree Ullal(CEO)