58.5% Growth, an $18.5B 2027 Guide, and a Shrinking Deferred Pool: Downgrading Arista to Hold

Results vs. Consensus

Arista reported Q3 2026 (quarter ended September 30, 2026) after the close on November 3. On the income statement the print is close to flawless: every line beat, operating margin held above the guided range while R&D stepped up by almost half, and the full-year outlook was raised again. The question this quarter is not the size of the beat. It is where the revenue came from, and the 10-Q, filed the next day, answers that question in a way the release does not.

Q3 2026 Scorecard

Versus the company's own Q3 guidance (issued August 4)

The beat versus guide ($357.6M on revenue) is the second consecutive quarter above 8%. What differs from Q2 is the operating-margin line. In Q2 the margin beat its guide by 290 bps; in Q3 it beat by 10 bps, because non-GAAP R&D rose 46.9% sequentially. The revenue upside fell almost entirely to new-product spending rather than to margin, which management signalled when it guided Q3 operating margin to 48-49% after a 49.9% Q2.

Year-over-Year Comparison

Two comparability notes. First, Q3 2025 non-GAAP results were recast for the tax-methodology change adopted in Q4 2025; the recast added $17.1M to prior-year non-GAAP net income and $0.02 to non-GAAP EPS, so the $0.77 base is $0.02 higher than first reported and the 55.8% growth rate is slightly understated against the original print. Second, GAAP net income growth of 82.9% is flattered by a $134.5M gain on strategic investments in other income, which non-GAAP excludes; that gain is why GAAP EPS ($1.22) sits above non-GAAP EPS ($1.20) this quarter.

Quarter-over-Quarter Comparison

Revenue

The year-over-year rate is the headline, but the shape of the year is the more useful number. Nine-month revenue of $9,402.3M is up 44.3%. The Q4 guide of $3.9-4.0B puts the full year at $13.30-13.40B, or 47.7-48.8% growth against FY2025's $9,005.7M, and implies second-half growth of about 59%. In August the guide implied about 43% for the second half. The acceleration is real at the revenue line. The difficulty, developed below, is that the inflow measures did not accelerate with it: revenue plus the change in deferred revenue grew 29.2% over nine months, well below the 44.3% revenue rate.

Margins

The margin story changed character this quarter. Through Q2, the operating-margin expansion came from opex growing far slower than revenue. In Q3, non-GAAP opex grew 32.7% sequentially on 20.5% revenue growth, and operating margin fell 80 bps from Q2 while still beating its guide. Gross margin did more of the work, at 64.0%, the high end of the 62-64% range. The 2027 framework (gross margin 62-64%, operating margin about 49%) makes clear that management is not planning further operating-margin expansion; it plans to spend the leverage on R&D for scale-up, liquid cooling and high-radix switching.

EPS

Non-GAAP EPS of $1.20 compares with $1.02 in Q2 and $0.77 a year ago (recast). The nine-month total is $3.09, and the Q4 guide of $1.26-1.30 takes FY2026 to $4.35-4.39, against our August estimate of about $4.10. The gap between GAAP and non-GAAP EPS is unusual this quarter: the $134.5M investment gain and $48.5M of tax adjustments removed from non-GAAP income exceeded the $153.4M added back for stock-based compensation ($138.1M, 3.8% of revenue) and intangible amortization ($15.3M). Investors should use the non-GAAP figure for run-rate purposes.

Segment & Mix Performance

Arista reports one operating segment. The disclosed mix is product versus service and revenue by region, supplemented by category targets management gives on the call (AI fabrics and campus) and, this quarter, by the deferred-revenue activity in the 10-Q, which is the most informative mix disclosure of the period.

Revenue composition

Deferred revenue activity: where the quarter's revenue came from

This is the table that changed our rating. Read across the columns, three things happened in Q3. The release from the opening balance jumped 65.5% sequentially, by $679.0M, which is more than the entire $621.9M sequential increase in revenue. New deferrals, the flow into the pool from shipments with acceptance terms and new support contracts, were flat at $1,688.6M (-0.9% from Q2, +27.5% from a year ago). And the sum of revenue and the net change in deferred revenue, a rough proxy for what Arista billed in the quarter, fell 1.9% from Q2 while revenue rose 20.5%.

The release includes ratable support revenue, so not all of it is product acceptance. Even if every dollar of the $467.6M of service revenue came from the opening balance, at least $1,247.4M of product revenue did, against at least $605.5M on the same basis in Q2 and at least $303.1M a year ago. The product acceptance cycle is now a first-order driver of reported revenue.

Assessment: None of this makes the revenue lower quality in an accounting sense; acceptance revenue is earned revenue. It does change what the growth rate means. The 58.5% year-over-year rate describes, in part, the conversion of a deferred pool that more than doubled from $3,088.8M at the end of March 2025, while the inflow proxy grew 23.8%. Unless the inflow accelerates, reported growth converges toward it. The 2027 guide of 38-40% sits above the nine-month inflow growth of 29.2%.

Non-GAAP operating expense detail

AI fabrics: $3.5B target for 2026, about $7B guided for 2027

The AI fabrics target for 2026 stayed at $3.5B for the third straight update, while total 2026 revenue has risen $1.8B since May. For 2027, management guided AI fabrics to be the fastest-growing category and to double to about $7B. The deployment frame is large: more than 10 gigawatts across 10 or more accelerator types by the end of 2027, and more than 100 Etherlink customers already.

"Within this envelope, we expect our AI fabrics target to be the fastest-growing segment and to double to approximately $7 billion, while our cognitive campus represents a healthy growth of 60%, much faster than the market to approximately $2 billion in 2027." — Jayshree Ullal, Chairperson & CEO

Assessment: The doubling is measured against a target, not an actual. Management said plainly it is "staying with the target numbers because we only know actuals after we ship everything end of the year." If 2026 AI revenue lands above $3.5B, as the unallocated $1.8B of raises suggests it may, then $7B represents growth below 100%. The dollar figure is what matters for the model, and $7B is a strong number either way.

Scale-out and scale-across: the 70-30 split shifts toward scale-across

Management put the 2026 split of the AI business at roughly 70% scale-out and 30% scale-across, and said the scale-across share will probably rise in 2027. Scale-up is expected to contribute nothing material in 2027. Revenue concentration in scale-across favors the 7800 AI spine at the largest customers; location count favors the smaller 7280 platforms at colocation and smaller sites.

"The scale-up will still be in early trials. So I don't think there's a material number associated with scale-up in '27. I do think it could contribute in '28." — Jayshree Ullal, Chairperson & CEO

Assessment: This is useful clarity for the 2027 model: the $7B AI number rests on scale-out and scale-across only, both categories where Arista already ships at volume. Scale-up becomes a 2028 option rather than a 2027 dependency, which is the correct way to frame a category with no Arista revenue today.

Campus: $1.25B to about $2B

Campus remains the diversification story. The 2026 target is unchanged at $1.25B; 2027 is guided to about $2B, or 60% growth. Management attributed share gains to reliability and to the security patching burden that AI-driven vulnerability discovery is placing on enterprise networks, and noted that long campus refresh cycles make this a multi-year process.

"In a lot of ways, I think the enterprise campus market is coming to us." — Todd Nightingale, President & Chief Operating Officer

Assessment: Sixty percent growth on a $1.25B base in a market growing in single digits is a share-gain claim, and it is the most diversified line in the guide. It is also the line least exposed to the deferred-acceptance dynamic, since campus products are mature and carry shorter lead times. We treat the campus target as the highest-quality component of the 2027 number.

Core data center, front end and routing: about 11% implied for 2027

Subtracting the two named categories leaves about $8.55B of 2026 revenue in the core business (cloud and enterprise data center front end, routing, and everything else) and about $9.5B in 2027, or roughly 11% growth. Management said every category will grow double digits, and that front-end spending at the largest customers is cyclical against back-end AI spending.

"Right now, in general, we're seeing more back end and AI fabrics, and I think the front end will catch up later." — Jayshree Ullal, Chairperson & CEO

Assessment: In August management said the core front end "isn't going to be flat," and we read the unallocated $1.1B raise as weighted toward core and front end. This quarter's framing points the other way for now: back end leads, front end "will catch up later." Our August inference was too generous. Because the 2026 AI target is stale, the $8.55B core figure is itself an upper bound, so the 11% rate for 2027 is approximate in both directions.

Services

Service revenue of $467.6M grew 17.9%, the slowest line in the release, and fell to 12.8% of revenue. Support revenue follows the installed base with a lag, so the slowdown in its share is arithmetic, not a sign of attrition. Unbilled multi-year support contracts rose to $497.1M from $427.8M in June.

Key Financial & Balance-Sheet KPIs

Two forward indicators point the right way and deserve weight against the deferred decline. Evaluation inventory, which is product sitting at customer sites under trial, rose $258.3M to $874.3M and has more than doubled since December. Binding customer agreements for future product shipments rose 7.8% to $943.5M. Both are revenue candidates for 2027. The cash line points the other way: operating cash flow of $1,200.9M was 5.3% below the prior-year quarter on a 58.5% revenue increase, because the inventory build consumed $674.6M in the quarter alone. Management warned that "quarters of elevated inventory balances" will affect "the timing of cash flow from operations ahead of customer deployments."

Key Topics & Management Commentary

Overall Management Tone: The most celebratory call of our coverage, and the least patient with balance-sheet questions. Management delivered the 2027 framework as a confident, carefully prepared announcement, but treated repeated questions about deferred revenue and purchase commitments as a distraction from demand, answering with framing rather than figures. Compared with August, when operations leadership was brought on to itemize supply fixes, the tone moved from demonstrating to asserting.

1. The Deferred Pool Stops Growing

Total deferred revenue declined $26.4M to $6,839.5M, and product deferred revenue declined $86M. Management did not hide the decline; it reframed it. The CFO described the quarter as having "the largest churn in and out" while the balance "still remained flat," and offered a year-end comparison as the reassurance.

"The only thing I'd add, like I said before, ... was that we anticipate the end of this year from a deferred balance to be higher than we ended last year, which shows that the P&L did what the P&L needed to do without deferred." — Chantelle Breithaupt, CFO

The year-end benchmark is low. Deferred revenue ended 2025 at $5,372.4M. The commitment is therefore satisfied even if the balance falls by up to $1,467.1M in Q4. Management chose that benchmark over the obvious one, a Q4 balance above Q3's.

Assessment: This is the most important disclosure of the quarter, and it is the trigger we published in August. The deferred table above shows the quarter's sequential growth was funded by the release of earlier-quarter revenue, while the inflow into the pool was flat. The year-end framing invites a further drawdown rather than ruling one out. We take management at its word that new products will bring new acceptance deferrals; we no longer assume the pool will keep growing while revenue accelerates.

2. The Fourth Raise: $12.6B to About $13.3B

FY2026 has now moved from a $10.5B Analyst Day goal to $11.25B, $11.5B, $12.6B and now just over $13.3B at the midpoint of the Q4 guide. At $13.35B the year grows 48.3%, and the $3.95B Q4 midpoint is up 58.8% year over year.

"With this backdrop, we are now increasing our annual guidance for the fourth time in 2026 to $13.3 billion, which represents a 48% annual growth." — Jayshree Ullal, Chairperson & CEO

Assessment: The beat-and-raise pillar delivered again. The raise is smaller than August's ($0.7B against $1.1B), which is natural with one quarter left. The cumulative $1.8B of raises since May remains unallocated between AI, campus and core, which leaves the quality of the mix unverifiable for a second quarter.

3. The First Look at 2027: About $18.5B

The 2027 framework is 38-40% revenue growth to about $18.5B, gross margin 62-64%, operating margin about 49% and a tax rate near 21%. It is roughly 10-13% above the Street's pre-print 2027 revenue estimate, and management itself described it as a strong guide given unusually early. Our estimate of what it implies for non-GAAP EPS is about $5.95 (an operating margin of 49%, other income near $0.6B, a 21% tax rate and about 1.285B diluted shares).

"Our first look at 2027 guidance range puts us in the 38% to 40% range for growth annually to approximately $18.5 billion." — Jayshree Ullal, Chairperson & CEO

Assessment: The number is impressive and management's track record on its own guides is excellent. The problem is reach. Management repeated that it has two quarters of visibility, and this guide covers five. On the nine-month figures, revenue plus the change in deferred grew 29.2%. A 38-40% revenue year from that starting point requires either an inflow acceleration management has not yet shown, or a drawdown of the deferred pool, which management declined to address.

4. The Core Remainder: About 11% Growth in 2027

With AI fabrics at about $7B and campus at about $2B, the rest of the business is about $9.5B in 2027 against about $8.55B in 2026, roughly 11% growth. That is well below the company-wide rate and was the subject of direct questioning.

"So I wouldn't read any conservatism into these numbers. These are wildly higher and optimistic than we had once projected. So enjoy them for a moment, and hopefully, they'll get better." — Jayshree Ullal, Chairperson & CEO

Assessment: Management explicitly declined to call the 2027 guide conservative. For a team whose guides have been raised four times this year, that is a notable change of posture. We read it as a guide set closer to management's real expectation than the 2026 starting point was, which reduces the room for the kind of raises that drove the stock in 2026.

5. R&D Steps Up to 11.2% of Revenue

Non-GAAP R&D rose to $408.5M from $278.1M. The 10-Q attributes most of the year-over-year increase to new-product introduction costs ($137.0M of the $160.9M GAAP increase), not personnel ($18.5M). The CTO tied the spend to co-design work with hyperscalers and liquid-cooling infrastructure.

"We're building so much new stuff. The opportunity is humongous in AI, these AI accelerating systems are -- we are -- our success with the hyperscaler AI companies depends on our ability to partner." — Kenneth Duda, Founder, President & CTO

Assessment: This answers our August question about whether 9.2% R&D was a durable level: it was not, and the CFO's 9-11% framework is the right planning range. Because the increase is prototype cost rather than headcount, it can fall back in a quarter with fewer programs. The margin held at 49.1% through the spike, which is a credible demonstration of the 2027 operating-margin target.

6. Purchase Commitments Move Past 12 Months

Purchase commitments rose to $11.4B from $9.7B. The 10-Q split shows the entire increase sits beyond 12 months: commitments due within a year were $9.3B against $9.4B in June, and commitments due later rose to $2.1B from $0.3B. The CFO attributed the addition to a multiyear memory agreement and said commitments will rise further in Q4.

"Purchase commitments from that perspective, there are some things additionally we added and now we have memory in there, a memory agreement, which has done that. So that's a multiyear agreement that you'll see as the Q comes out." — Chantelle Breithaupt, CFO

Assessment: Securing memory for 2027 and beyond is sensible given an industry shortage management expects to last to 2028, and it supports the supply side of the 2027 guide. It also extends the forecast risk: $2.1B of commitments now reach beyond the window management can see, against $0.3B three months ago. At roughly 1.7 years of 2027 cost of revenue at the guide, total commitments are a large forward bet on the 2027 number being right.

7. Gross Margin: 64.0%, and 62-64% Again for 2027

Gross margin of 64.0% was the top of the long-standing range. Management attributed the sequential gain to end-customer mix and the year-over-year decline of 120 bps to product-cost inflation partly offset by supply-chain efficiency. Q4 is guided at 63-64% and 2027 at 62-64%.

"The overall gross margin in Q3 was 64%, down from 65.2% in the prior year, with supply chain efficiency working to partially offset product cost inflation, and up from 63.4% in the prior quarter, driven by end customer mix." — Chantelle Breithaupt, CFO

Assessment: The cost-shock narrative that opened our coverage is now fully absorbed: margin sits at the top of the range despite memory inflation. The 2027 range also settles one bull hope, that pricing power would lift margin above 64% as supply normalized. Management is not planning for that, although the 10-Q notes "targeted price increases from time to time" in response to cost inflation.

8. Scale-Up Gets a Timeline: Trials in 2027, Production in 2028

At the Open Compute Project conference in October, Arista showed rack-scale scale-up designs built on Ethernet for Scale-Up Networks (ESUN), including the Etherlink SU-144 architecture, co-designed with several accelerator partners. Management gave a specific timeline for the first time.

"We expect scale-up trials to begin in 2027, driving Ethernet to larger radix of 126 to 144 within a rack and across 1,000 across racks, with production expected for scale-up in 2028." — Jayshree Ullal, Chairperson & CEO

Assessment: The timeline is consistent with the InfiniBand-to-Ethernet analogy management has used: a multi-year migration, now with dates. It adds a 2028 growth vector after 2027's scale-out and scale-across year, and the partner list is entirely non-NVIDIA, which is where scale-up is contestable. The fulfillment model (whole rack versus switch only) is still undecided, so the economics remain unknown.

9. Ten Gigawatts and Ten Accelerators

Management framed the AI opportunity as a shift from compute-centric to network-centric design and gave a deployment number: more than 10 gigawatts across 10 or more accelerator types by the end of 2027. It clarified that the figure includes some existing gigawatts connected by scale-across, not only new builds.

"By the end of '27, we are forecasting greater than 10 gigawatts of deployment across 10 or more different types of compute accelerators." — Jayshree Ullal, Chairperson & CEO

Assessment: Accelerator diversity is the structural case for an independent network vendor: the more XPU types a customer runs, the more valuable a neutral, standards-based fabric becomes. The gigawatt figure is a forecast, not a backlog, and its mix of new and existing capacity was left loose, so we treat it as direction rather than as a revenue anchor.

10. Security Becomes a Demand Driver

The call gave unusual time to network security. The CTO described work with frontier AI model developers to add AI-driven vulnerability discovery to Arista's security pipeline, and argued that a single operating system with process isolation and hitless patching is the right defense against AI-accelerated attacks. The COO tied enterprise wins in healthcare and hospitality to the same theme.

"True network security has become a must-have, not a nice-to-have, and Arista's EOS operating system is the ideal architectural foundation based on 4 key principles." — Kenneth Duda, Founder, President & CTO

Assessment: This is the strongest version of the enterprise argument we have heard from Arista: the patch burden that AI-driven vulnerability discovery creates favors the vendor with the fewest vulnerabilities and the least disruptive upgrades. It supports the campus guide. The monetization of AVA, the agentic operations assistant, is still at alpha with no pricing model, so it adds nothing to 2027.

11. Ten-Percent Customers: "At Least 3 and Maybe 4"

For the fourth consecutive quarter, management said it expects to add one, maybe two, 10% customers to Microsoft and Meta, now phrased as "at least 3 and maybe 4" in total by year-end, contingent on shipments through December 31.

"We fully expect our top prior 10%-plus customers to continue to be with us in '26, and we hope beyond. And we expect to add at least 1, maybe 2." — Jayshree Ullal, Chairperson & CEO

Assessment: Unchanged and unconfirmed. The 10-K in February will settle it. A third or fourth 10% customer would read as diversification at the top of the customer list, but it would also confirm that the 2026 growth is concentrated in a handful of very large AI builds, which is exactly where the acceptance-clause dynamic lives.

12. Capital Allocation: A Third Quarter With No Buyback

Arista repurchased no shares in the first nine months of 2026, leaving $817.9M of the May 2025 authorization unused, while cash and marketable securities rose to $14.36B. Arista joined the S&P 100 in September.

"Of the $1.5 billion stock repurchase program approved in May of 2025, $817.9 million remain available for repurchase in future quarters." — Chantelle Breithaupt, CFO

Assessment: With $11.4B of purchase commitments and an inventory build to fund, holding cash is defensible. But a management team guiding 38-40% growth for 2027 with $14.36B of cash and no debt has left the authorization untouched for nine months. The stock then fell 24.3% from its October 9 record close to the November 2 close of $164.14; whether Arista bought any stock in that window will only be visible in the Q4 report, and nothing on the call suggested it did.

13. Optics: Pluggables Stay the Default Through 2027

On co-packaged and near-packaged optics, management reiterated a "copper when you can, optics if you must" stance, backed pluggable optics as the majority technology, and placed open co-packaged designs with socketed optical engines in the 2028-2030 timeframe.

"We think they'll come on to the market in the '28 to '30 time frame. It's not imminent, but meanwhile, long live pluggable optics." — Jayshree Ullal, Chairperson & CEO

Assessment: A reasonable position that keeps Arista out of a proprietary integration race it does not need to run in 2027. The risk is competitive timing: if a vertically integrated supplier makes co-packaged optics standard sooner at the largest customers, Arista's open approach could arrive late. Nothing in the 2027 guide depends on it either way.

Guidance & Outlook

The strategic framing is that Arista's addressable market now exceeds $200B in 2030 and that 2026 and 2027 are "the exponential ramp." The financial framing is narrower and more useful: management guided the rest of 2026 and all of 2027 at a confidence level it said required unusual internal work, and it tied the confidence to supplier agreements and two years of technical engagement with customers rather than to orders in hand.

Implied ramp: The Q4 guide is +6.6% to +9.4% sequentially and +56.8% to +60.8% year over year. For 2027, $18.5B against about $13.35B in 2026 implies an average quarter of about $4.6B, against $3.95B guided for Q4. The sequential path is not demanding if the inflow resumes; the issue is that the inflow proxy was $3,631.2M in Q3, already below both Q3 revenue and the Q4 guide.

Street at: Q4 revenue consensus sat near $3.57-3.60B and EPS near $1.06-1.15 across sources, so the guide midpoint is about 10% above on revenue and 11-21% above on EPS. 2027 revenue consensus was about $16.4-16.75B, against the $18.5B guide.

Guidance style: Historically conservative, and management explicitly asked investors not to read conservatism into 2027. We take that at face value: the 2027 guide is closer to a central estimate than the 2026 guide was at the same stage.

Analyst Q&A Highlights

Whether deferred revenue has stopped predicting revenue

The sharpest question of the call asked whether revenue growth was coming from new orders or from recognizing older ones, and whether the link between deferred revenue and revenue had broken. Management rejected the framing and declined to decompose demand, while confirming large flows in both directions.

Q: "What are the drivers for the growth? Is it orders? Or is there some part there of recognition of historical orders? And then is deferred revenue actually losing some of its predicting ability just because I see some decoupling between the deferred revenue trends and revenue growth trends?"
— Tal Liani, Bank of America

A: "Deferred is a strong function of a new use case or a new product that eventually is not a new use case or a new product. So we had millions of dollars come out of deferred because the use case is satisfied. We had millions of dollars go in as deferred."
— Jayshree Ullal, Chairperson & CEO

Assessment: The answer is accurate and the 10-Q quantifies it: $1,715.0M out, $1,688.6M in. It does not address the question, which was about direction. The decoupling the question described is real in the filing, and management's response was to tell investors not to look at it.

Whether the 2027 guide needs the deferred balance drawn down

A related question asked directly whether the 2027 outlook requires an aggressive drawdown of deferred revenue, and whether it depends on new power and data-center capacity being financed. The CFO declined to guide deferred revenue.

Q: "And as you address that, can you speak to whether you require an aggressive drawdown of your deferred revenue balance to achieve your outlook? I think I know the answer, but I would love your view."
— Karl Ackerman, BNP Paribas

A: "So we don't guide deferred, but I would say that we see a very strong demand signal and demand environment back to the momentum theme."
— Chantelle Breithaupt, CFO

Assessment: A non-answer to the single most relevant question about the 2027 number. The CEO added that 2027 will be "filled with lots of new products," which implies new acceptance deferrals; that is plausible, but it means the pool would grow only if new deferrals outrun releases, which did not happen in Q3.

How purchase commitments and deferred revenue fit the 2027 guide

Questioning asked the CFO to reconcile a flat deferred balance and a $1.7B rise in purchase commitments with a 38-40% growth guide. The answer produced the memory-agreement disclosure and a signal that commitments will rise again.

Q: "But just help us understand how to think about that metric in the context of a 38% to 40% guide, particularly also with purchase commitments going up a couple of billion dollars sequentially."
— David Vogt, UBS

A: "And I think on the purchase commitment side, you can probably expect that to go up a little bit more as we exit Q4, as we get ready for this 2027 guide we have."
— Chantelle Breithaupt, CFO

Assessment: Useful on commitments, which are being sized to the 2027 guide. On deferred revenue the CFO called the flat balance "a really healthy sign of the demand," which is a defensible reading of gross inflows but not of the net. Commitments rising toward the guide while the inflow proxy is flat is the combination our demand-exposure bear point described in August.

Whether the R&D step-up is the new run rate

Questioning asked whether the sequential jump in R&D included one-time items. The CTO attributed it to new-product work and the CFO kept the 9-11% framework.

Q: "I was just curious about the big sequential jump in R&D. It was really, really significant. I'm just wondering if that included some onetime items. Also, is this the new run rate in R&D going forward?"
— George Notter, Wolfe Research

A: "So some quarters like we've talked about, we'll hit this range as we have specific investments. And so I would still keep it in that 9% to 11%. More to come as we guide next year in more detail."
— Chantelle Breithaupt, CFO

Assessment: Clear and consistent with the 10-Q, which puts the increase in prototype and third-party engineering cost. Q3 sits just above the top of the range, so the operating-margin guide of about 49% for 2027 has room to absorb this level of spend.

Whether new 10% customers are landing

Questioning returned to the long-promised additional 10% customers. The answer repeated the prior framing with a new total.

Q: "So I just want -- you've previously referred to potentially 2 10%-plus customers into the fourth quarter. Do you mind talking about whether they're landing in that range and what the use cases are for and if you're excited about those customers still and into '27?"
— Ben Reitzes, Melius Research

A: "And we expect to add at least 1, maybe 2. And a lot of that depends on whether Todd and Eugenia can ship enough product. So we'll let them decide, and we'll give them all the way into December 31."
— Jayshree Ullal, Chairperson & CEO

Assessment: Fourth consecutive quarter of the same expectation. The tie to shipments through December 31 suggests the threshold is close for at least one customer, but the use-case half of the question went unanswered.

Why the core business slows in 2027

Questioning pointed out that the guide implies the non-AI, non-campus business slows sharply in 2027 and asked whether that reflected tough comparisons or supply conservatism.

Q: "I was just wondering if you can kind of touch on the underlying assumptions or factors that's maybe driving a little bit more moderation here into 2027 and whether that's maybe just tough compares or conservatism in terms of supply, et cetera."
— Joseph Cardoso, JPMorgan

A: "We're very comfortable that everything is going to grow double digits, some more, some less. ... So I wouldn't read any conservatism into these numbers."
— Jayshree Ullal, Chairperson & CEO

Assessment: The answer accepts the moderation and explains it as the back-end/front-end cycle. "Double digits" for the core is consistent with the roughly 11% the guide implies, so the guide's composition is internally coherent. The explicit denial of conservatism is the more important part of the exchange.

How a two-quarter-visibility company guides five quarters out

Questioning asked how a team that cites one to two quarters of visibility could give a full-year 2027 number this early. The answer distinguished visibility from engagement.

Q: "But look, I appreciate the early outlook, but you guys always talk about 1 to 2 quarters of visibility. You tend to be a very prudent team. And so maybe ... the confidence in providing the durability at this scale this early and why we're going to see this kind of strength persist, why Arista for scale-up in 2028?"
— James Fish, Piper Sandler

A: "And so we're very confident not only in our '27 guide, but a promising '28 ahead as well. And the reason we are is because even though we have 2 quarters of visibility, we have 2 years of work behind it."
— Jayshree Ullal, Chairperson & CEO

Assessment: "Two years of work" describes design engagement, not orders. That is a legitimate basis for confidence in share, but not in timing, and timing is what a calendar-year revenue guide measures. The answer confirms that the second half of 2027 rests on relationships rather than on purchase orders.

Whether supply for 2027 is fully secured

The last question asked whether all silicon and memory for the $18.5B is already planned with suppliers, given views that component tightness worsens in 2027.

Q: "Should we -- since you are providing an initial assessment on 2027 revenue target, should we assume that you have already planned with your suppliers and silicon is secured for both compute and memory?"
— Mehdi Hosseini, Susquehanna

A: "So we're incredibly -- we have very high confidence in the ability to deliver supply to the forecast. We -- of course, we would not want to place everything to 5 quarters lead time for all of next year. So not every order is placed."
— Todd Nightingale, President & Chief Operating Officer

Assessment: Candid and sensible. Supply is no longer the binding risk on the 2027 number, which confirms our August decision to treat the supply bear point as contained. The binding risk is now on the other side of the ledger.

What They're NOT Saying

Each item below was checked against the release, the transcript and the 10-Q filed November 4.

1. Whether deferred revenue rises in Q4: The only commitment is that year-end deferred will exceed the 2025 year-end balance of $5,372.4M, which allows a Q4 decline of up to $1,467.1M. Management did not say whether it expects the balance to rise from Q3.
2. How much of 2027 depends on the deferred pool: Asked directly, the CFO said "we don't guide deferred." The 10-Q quantifies the 2026 releases but gives no view on 2027.
3. The 2026 AI and campus actuals: Both targets are unchanged ($3.5B and $1.25B) across $1.8B of raises since May, and management said it will only know actuals at year-end. The 2027 AI "doubling" is therefore measured against a target, not a result.
4. The memory agreement's terms: The CFO said the multiyear memory agreement would be visible when the 10-Q came out. The 10-Q names no counterparty, size or term and files no material-contract exhibit; the only trace is the rise in commitments due beyond 12 months, from $0.3B to $2.1B.
5. The product share of the $1,715.0M release: The 10-Q does not split revenue recognized from the opening balance between product acceptance and ratable support. Our lower bound for product is $1,247.4M.
6. Which customers approach 10%: No names or use cases for a fourth quarter. The 10-Q gives only the 2025 annual shares of the two existing 10% customers (26% and 16%).
7. Why the buyback stayed dormant: No repurchase in the first nine months, no rationale offered on the call, and no comment on whether the October drawdown changed the approach.
8. 1.6T timing: In August management placed 1.6T trials in the second half of 2026 and production in 2027. It was not mentioned on this call, in the release or in the 10-Q, so its place in the 2027 guide is unknown.
9. Scale-up economics: The gross margin of rack-scale systems was answered only with the corporate 62-64% range, and the fulfillment model is still "different strokes for different folks."

Market Reaction

• Pre-print setup: ANET closed at $169.91 on November 3, up 29.7% year to date (S&P 500 +10.6%), up 7.8% over the trailing twelve months and down 18.1% over the trailing thirty days. The stock set a record close of $216.74 on October 9, then fell 18.3% in five sessions to $176.98 on October 16, and reached a low close of $164.14 on November 2 after further one-day declines of 7.2% on October 22 and 8.3% on October 30. It rose 3.5% in the print session itself.
• After-hours move: Shares rose as much as roughly 13% after the release and stood near $185 (+8.9%) late in the after-hours session.
• Reaction session (November 4): The stock opened at $179.00 (+5.3%), reached $188.53 (+11.0%) by 9:50 AM ET, gave back about half of the move by about 11:00 AM, and traded between about $177 and $181 for the rest of the day to close at $178.48, up 5.0%. Volume of 18.5M shares was 3.4x the 30-day average.
• Peers and tape: The S&P 500 rose 1.6%, but AI hardware sold off. The semiconductor index fell 3.1%, NVIDIA 3.2%, Broadcom 2.0%, Astera Labs 4.4%, Celestica 9.1%, Ciena 9.3% and Credo 9.5%. Cisco rose 0.6%.
• Since our last note: The $178.48 close is 5.4% below the $188.67 reference close in our August recap and 17.7% below the October 9 record.

The reaction has two readings, and both are partly right. Against its own tape, ANET did very well: a 5.0% gain on a day when AI-hardware peers fell between 2% and 10% is a relative move of 7 to 15 points, and it reflects how far above the Street the 2027 number landed. The October drawdown also matters. The stock entered the print 21.6% below its record after a month in which the market questioned how AI infrastructure is being financed, and an Alphabet print on October 29 that reopened the hyperscaler cash-burn debate preceded ANET's 8.3% decline on October 30. A good print against a de-rated stock should rally, and it did.

The second reading is the intraday shape. The stock reached +11.0% in the first twenty minutes and gave back about half of that move within ninety minutes, then held the remainder all day. The broad AI-hardware selloff explains part of the fade, and the deferred-revenue debate was part of post-print coverage. We do not think the market has yet priced the deferred table in the 10-Q, which was filed on the reaction day, and the sell-side reaction we saw largely treated the product deferred decline as timing noise.

For a fundamental owner the setup is better than in August on valuation and worse on visibility. At $178.48 the stock trades at about 41x our FY2026 estimate, against about 43x for the August 7 price on the same estimate, and about 30x 2027 at the guide. That is a reasonable multiple for a business guiding 38-40% growth, if the guide is achieved without drawing down the deferred pool. It is not a margin of safety if it is not.

Street Perspective

Debate: Is the deferred decline timing noise or the cushion being spent?

Bull view: The prevailing sell-side view treats the product deferred decline as lumpy acceptance timing, given supply availability and new scale-up certifications. The 10-Q gives that reading some support: evaluation inventory rose $258.3M in the quarter to $874.3M, and binding customer agreements for future product shipments rose 7.8% to $943.5M. Both are candidates for the next wave of deferrals.

Bear view: The skeptics argue that revenue is running ahead of new bookings, that the pool which cushioned the guide for a year stopped growing just as management extended the guide to five quarters, and that purchase commitments and inventory are rising for demand not yet booked.

Our take: The bears have the data for this quarter and the bulls have a plausible story for the next. One flat quarter does not prove the inflow has peaked, and evaluation inventory is a genuine leading indicator. But the bull case now requires the inflow to accelerate, and that is a forecast, while the release-driven sequential growth is a fact in the 10-Q.

Debate: Is the $18.5B 2027 guide another floor?

Bull view: The bull camp notes that every guide in 2026 was raised, that management has rarely given a full-year number this early, and that a guide 10-13% above consensus from a historically cautious team should be read as a floor, with 1.6T, scale-up trials and campus share gains as upside.

Bear view: The bear camp notes that management itself said not to read conservatism into the number, that the core business is guided to slow to about 11%, and that the guide sits above the inflow growth rate the company actually reported.

Our take: We side with management's own description: this is a central estimate, not a floor. That still makes it a very good number. It means the raise-every-quarter pattern that powered the stock in 2026 should not be assumed for 2027, and the stock has to earn its return from delivery rather than from upward revisions after the first one.

Debate: Is the October drawdown an entry point?

Bull view: Some desks argue that a stock down 18% from its record, trading near 30x 2027 earnings at the guide with no debt and $14.36B of cash, is an entry point, and that post-print target increases across the Street confirm it.

Bear view: The bear camp contends that AI-hardware multiples are compressing as a group on financing concerns, that ANET's beta to hyperscaler capital budgets is high, and that a second quarter of deferred decline would hit both the estimate and the multiple at once.

Our take: The valuation case is better than it was in August, and that is why we stop at Hold rather than going further. We would rather wait one quarter for evidence that the inflow resumed than pay 30x for a guide whose funding source is unclear.

Model Update Needed

Valuation impact: FY2026 non-GAAP EPS moves to about $4.37 from about $4.10, up 6.6%. We introduce a 2027 estimate of about $5.95 that takes the guide at face value; we have not haircut it, because management's delivery record does not justify one. At the November 4 close of $178.48 the stock trades at 40.8x FY2026E and 30.0x FY2027E. On the same FY2026 estimate, the August 7 price of $188.67 was 43.2x, so the de-rating since our last note is about 2.4 turns, not the larger drop that comparing against August's 46x on the old estimate would suggest. We still publish no formal price target. The rating change reflects the funding question on the 2027 guide, not a view that the stock is expensive.

Thesis Scorecard Post-Earnings

Grading the standing thesis carried from the Q2 2026 update.

Grading last quarter's commitments

The triggers we published in August

Overall: Thesis mixed. The business pillars strengthened: AI leadership, the guidance record and operating leverage all had a strong quarter, supply is no longer a constraint, and the valuation de-rated. Against that, the demand-exposure risk we added in August escalated, and the 10-Q revealed a new one. Revenue growth is now being supplied in part by the conversion of a deferred pool that stopped growing, so the reported growth rate overstates the growth of the inflow that will drive 2027.

Why we are honoring the trigger. In August we wrote that a sequential decline in deferred revenue would move us to Hold. The decline is small in net terms, and a reader could fairly call $26.4M noise. We did not stop at the net figure. The gross activity in the 10-Q shows the decline is the visible edge of a larger shift, in which releases rose $679.0M while new deferrals did not grow, and that shift is the condition the trigger was written to catch. We are not re-cutting the trigger after it fired, and we are not arguing around it.

Action: Downgrade to Hold from Outperform; conviction 6. This is not a bearish call: the business is executing at its best level, the valuation improved, and the 2027 guide may well be met. What has changed is that the guide's funding is uncertain, and the next two prints will resolve it. The new triggers are: (1) back to Outperform if total deferred revenue rises sequentially at Q4 (above $6,839.5M) with Q4 revenue inside or above the $3.9-4.0B guide, or if revenue plus the change in deferred revenue returns above Q2's $3,702.9M; our base case is a Q4 decline, so both sit outside it; (2) to Underperform if Q4 revenue falls below the $3.9B low end or the February update cuts 2027 below 38% growth, both outside our base case. If a Q4 deferred increase and a 2027 cut arrive together, they offset and we stay at Hold.