tapebrief
ANET · Q3 2025

ANET · Q4 2025 Earnings

Neutral

Arista Networks

Reported February 12, 2026

30-second summary

SENTIMENT: Constructive Arista printed $2.49B in Q4 (+28.9% YoY, +7.8% QoQ), beating its own $2.30–2.40B guide by $88–188M, with non-GAAP gross margin of 63.4% landing above the 62–63% guide. The bigger story is forward: management formally raised FY26 revenue from $10.65B / +20% to $11.25B / +25%, raised the FY26 AI center target from $2.75B to $3.25B, and raised the FY26 non-GAAP operating margin outlook to approximately 46%. Q1 FY26 guidance of $2.6B implies roughly 30% YoY growth. The texture is supply: memory costs are running "order of magnitude exponentially higher" with shortages "expected to last multiple years," and Q1 non-GAAP operating margin is guided to 46%, down 150bps from Q4's 47.5%. The raise was delivered despite — not in ignorance of — that headwind.

Headline numbers

EPS

Q4 FY2025

$0.82

Revenue

Q4 FY2025

$2.49B

+28.9% YoY

Gross margin

Q4 FY2025

62.9%

Operating margin

Q4 FY2025

41.5%

Key financials

Q4 FY2025
MetricQ4 FY2025YoYQ3 FY2025QoQ
Revenue$2.49B+28.9%$2.31B+7.8%
EPS$0.82—$0.75+9.3%
Gross margin62.9%—64.6%-170bps
Operating margin41.5%—42.4%-90bps

Guidance

Arista beat Q4 FY2025 guidance across revenue and margins, raising full-year FY2025 gross margin but significantly lowering reported operating margin; Q1 FY2026 revenue raised to $2.6B with stable margin guidance.

Guidance is issued for both next quarter and the full year. Both may appear below.

Actuals vs prior guidance

MetricPeriodPrior guideActualΔResult
RevenueQ4 FY2025$2.3-$2.4 billion$2.488 billion+$0.088-0.188 billion above guideBeat
Non-GAAP Gross MarginQ4 FY202562-63%63.4%+0.4-1.4 percentage points above guideBeat
Non-GAAP Operating MarginQ4 FY202547-48%47.5%-0.5 percentage points below guide high end; within rangeBeat

New guidance

MetricPeriodGuideYoY
RevenueQ1 FY2026$2.6 billion—
Non-GAAP Gross MarginQ1 FY202662-63%—
Non-GAAP Operating MarginQ1 FY202646%—

Changes to prior guidance

MetricPeriodPrior guideNew guideΔResult
Gross Margin
FY2025
approximately 64%64.1%+0.1 percentage pointsRaised
Operating Margin
FY2025
approximately 48%42.8%-5.2 percentage pointsLowered

Product revenue

Q4 FY2025
SegmentQ4 FY2025YoY
Product Revenue$2.096B+30.3%
Service Revenue$0.392B+21.6%

Management tone

Q2 anchor: "Once in a lifetime opportunity" → Q3 anchor: FY26 disaggregation and AI organizational restructuring → Q4 anchor: FY26 raise paired with multi-year supply discipline.

From "demand exceeds shipments" as a one-line caveat to memory as the binding multi-year constraint. At Q3, supply chain was framed as a manageable variability driver. This quarter it became the central margin narrative. From the UBS Q&A: memory prices are "horrendous" and "order of magnitude exponentially higher," with the shortage "expected to last multiple years." Management absorbed the cost in 2025 but is signaling 2026 will be worse — which is precisely why Q1 non-GAAP operating margin is guided 150bps below Q4 actual despite revenue accelerating, and why the FY26 operating margin guide settled at ~46% rather than holding nearer the 48% FY25 print.

From AI revenue inevitability to AI deployment timing realism. Q3's tone implied AI center revenue would track linearly. This quarter, in response to JP Morgan probing why FY26 growth at 25% on $9B appears more measured than Q1's ~30%, Jayshree was explicit that "network deployment lags infrastructure build-out (power, GPUs)" and that demand may shift into 2027 from 2026. The Wells Fargo exchange on scale-up vs scale-out reinforced this — the eSUN spec needed for at-scale scale-up production isn't expected until Q4 2026 at the earliest, with majority scale-up revenue now a 2027 event. The FY26 framework was raised; the calendarization of scale-up specifically is later than Q3 suggested.

From firm "$10.65B / +20%" framework to formally raised "$11.25B / +25%" framework with AI center stepped up to $3.25B. Q3 introduced an explicit FY26 dollar target ($10.65B, ~+20%) with disaggregated AI center ($2.75B) and campus ($1.25B) components. This quarter, in CFO prepared remarks, management raised the full-year framework to $11.25B / +25%, raised the AI center target to $3.25B, held the campus target at $1.25B, and lifted operating margin guidance to ~46%. This is a formal three-metric FY26 raise, not Analyst Day foreshadowing.

From "rising tide" competitive framing to multi-model-builder partnership narrative. Q3 framed the AI competitive set as a market-share-stable rising tide. This quarter, Evercore drew out a more textured customer architecture: Arista now works with Gemini, XAI, Anthropic, and OpenAI, with model builders moving from single-data-center deployments to multi-colo partnerships with cloud titans. This is the first quarter the named model-builder roster has been articulated this explicitly — a signal that the customer count beyond the original tier-one cloud titans is broadening into a definable set of accounts.

Q&A highlights

Amita Marshall · Morgan Stanley

Inquiry into the drivers for achieving one or two additional 10% customers in 2026, specifically asking about bottlenecks in customer building and acceptance criteria.

Jayshree explained that confidence in the forecast is based on multiple variables including deferred revenue acceptance criteria timing, ongoing demand, and supply chain/shipping constraints. Actual customer concentration may land in high single digits to low double digits depending on these factors.

One or two potential new 10% customers expected in 2026Variables include deferred revenue acceptance criteria and timingSupply chain allocation and inflation constraints affect shipping abilityFinal concentration may be high single digits to low double digits

Samit Chatterjee · JP Morgan

Questions about the apparent conservatism in 25% guidance despite 30% Q1 guidance, asking whether cautious outlook is driven by new customer ramps or supply chain constraints.

Jayshree clarified that he is providing realistic visibility on what can be shipped versus what is demanded. Network deployment lags infrastructure build-out (power, GPUs); many new use cases are still in early innings and demand may ship over multiple years. Growth is still strong given the large base ($9B revenue).

25% growth guidance on $9B base is strong given large scaleNetwork deployment timing lags infrastructure buildout by monthsMany AI use cases still in first innings of developmentQ1 guidance at ~30% reflects early-year strength

David Vogt · UBS

Request to quantify revenue and gross margin impacts from memory constraints and supply chain pressures embedded in guidance, and potential upside if supply resolves.

Jayshree explained that memory prices are 'horrendous' and 'order of magnitude exponentially higher' due to worsening shortages. Arista absorbed costs in 2025 but faces worse conditions in 2026 expected to last multiple years. Chantel confirmed 62-64% gross margin guidance holds despite pressures, with path to guidance intact through mitigation efforts.

Memory prices increased order of magnitude exponentially in 2026Costs absorbed in 2025; situation worsened significantly in 2026Supply shortage expected to last multiple yearsGross margin guidance of 62-64% maintained despite pressure

Aaron Rakers · Wells Fargo

Questions on AI revenue decomposition between scale-up and scale-out, timing of revenue recognition, and why non-AI/non-campus segments are guiding to low single-digit growth.

Jayshree clarified that current configurations are mostly scale-out/scale-up combinations using 800G and smaller radix. Scale-up production at scale will be 2027, tied to eSUN specification completion (expected Q4 2026 at earliest), primarily using 1.6T speeds. New products and specs are required before major scale-up ramp.

Current AI deployments mostly scale-out and scale-up combinationsScale-up production scale expected 2027, not 2026eSUN specification completion expected Q4 2026 at earliest1.6T switching emerging in 2026 but majority scale-up revenue in 2027

Amit Daryanani · Evercore ISI

Discussion of multi-billion dollar AI clusters being built by companies like Anthropic independently, and Arista's participation in these buildouts plus opportunities with cloud titan partnerships for these model builders.

Jayshree explained that AI model builders are moving beyond single data center silos to multi-location, multi-colo partnerships with cloud titans. Arista works with multiple model builders (Gemini, XAI, Anthropic, OpenAI) and expects co-pilot versions with cloud titans. Network infrastructure critical for multi-protocol algorithms and minimizing collective degradation.

Model builders moving from single data center to multi-location deploymentsWorking with multiple model builders beyond initial 1-2 (Gemini, XAI, Anthropic, OpenAI)Expect partnerships between model builders and cloud titansNetwork handles multiple accelerators and multi-protocol algorithms

Answers to last quarter's watch list

Whether Q4 non-GAAP gross margin lands at the high end of the 62–63% guide or below it — Q4 non-GAAP gross margin printed at 63.4%, above the guide high end by 40bps. The cloud/AI mix headwind flagged at Q3 materialized in the magnitude expected, but Arista executed at the favorable end.
Resolved positively
Whether FY26 revenue guidance of ~$10.65B holds or is raised — Raised. Management formally lifted FY26 revenue to $11.25B / +25%, AI center to $3.25B, and operating margin to ~46% in CFO prepared remarks.
Resolved positively
Concrete dollar disclosure on FY25 AI center revenue achievement vs. the "at least $1.5B" floor — Jayshree stated the $1.5B AI center networking goal was exceeded, but no specific FY25 dollar achievement was quantified. Status: Partially resolved.
Purchase commitments + inventory direction from Q3 — Disclosed. Purchase commitments rose to $6.8B at end of Q4 from $4.8B at end of Q3 (+$2.0B QoQ); inventory $2.25B. The build reflects forward memory and silicon procurement against multi-year shortage. Status: Resolved.
Whether the fourth tier-one cloud customer crosses 100K GPUs and any disclosure on a fifth named AI customer — Q&A referenced "one or two potential new 10% customers in 2026" rather than confirming an existing-customer GPU milestone or a named fifth account. The diversification narrative was instead carried by the model-builder roster (Gemini, XAI, Anthropic, OpenAI) disclosed in the Evercore exchange.
Not resolved
Core (non-AI, non-campus) business trajectory disclosure — Wells Fargo asked specifically why non-AI/non-campus segments are guiding to low single-digit growth; management acknowledged the framing but did not quantify the core business as a discrete line item. The black box inside the FY26 framework remains opaque.
Continue monitoring

What to watch into next quarter

Whether Q1 non-GAAP operating margin lands at the 46% guide or compresses further — a print below 45% would suggest the memory cost absorption is running ahead of management's modeled mitigation and would force a re-rate of the 62–64% FY gross margin band.

Whether the newly raised $11.25B / +25% FY26 framework holds against worsening memory and silicon costs — the raise was delivered with explicit acknowledgement that 2026 supply conditions are worse than 2025; any softening within the year would test how much margin management is willing to spend to hold the top-line lift.

Whether the $3.25B AI center target tracks ratably or back-half-loads — given scale-up production is gated to 2027, the $3.25B is predominantly scale-out; an early-quarter shortfall would suggest deployment timing slippage rather than demand erosion.

Concrete FY25 AI center revenue dollar disclosure — Jayshree confirmed the $1.5B floor was exceeded but did not quantify the overshoot, leaving the baseline for the $3.25B FY26 trajectory imprecise.

Named fifth AI customer or a fourth tier-one cloud customer GPU milestone — would convert the "one or two new 10% customers in 2026" hedge into a falsifiable commitment.

Purchase commitments trajectory from the $6.8B Q4 print — direction and magnitude speak directly to whether the multi-year memory shortage narrative is materially inflating Arista's working-capital footprint.

Whether management quantifies the non-AI, non-campus core business growth rate — currently the largest modeling gap; Wells Fargo's pointed question on this remains unanswered after two consecutive quarters.

Sources

  1. Arista Networks Q4 2025 Earnings Release, filed with SEC: https://www.sec.gov/Archives/edgar/data/1596532/000159653226000010/ex991q425-earningsrelease.htm
  2. Arista Networks Q4 2025 earnings call (CFO prepared remarks and management Q&A).
  3. Arista Networks Q3 2025 brief (Tapebrief), for cross-quarter comparison.
  4. Arista Networks Q2 2025 brief (Tapebrief), for cross-quarter comparison.

Get the next brief, free.

We publish analyst-grade earnings briefs the same day or morning after every call — headline numbers, segment KPIs, Q&A highlights, and tone analysis. Free during beta.

This is not investment advice.