Arista Is Punished for Perfection: A Double Beat, a Raised Guide, and a 14% Drop
Key Takeaways
- Arista beat on both lines and raised the year, yet the stock fell 13.6% to $147.01. Q1 revenue of $2.709B (+35.1% YoY) topped the ~$2.61B Street number and its own $2.6B guide; non-GAAP EPS of $0.87 beat by ~7%; operating cash flow of $1.69B was the best in company history. This was a clean quarter that the market rejected on expectations, not fundamentals.
- The raise disappointed a bar the stock itself set. FY2026 revenue went to ~$11.5B (+27.7%) from +25%, and the AI target to $3.5B from $3.25B, but the Street had penciled 28–30% after a ~90% one-year rally. A "roughly in-line" Q2 guide could not clear that bar.
- Demand is no longer the question; supply is. Management called demand "the best I've ever seen in my Arista tenure" while flagging industry-wide shortages (wafers, silicon, CPUs, optics, memory), 52-week lead times, and vendor de-commits that will persist one-to-two years. The $11.5B guide reflects what Arista can ship, not what customers have ordered.
- Margins are the price of shipping. Gross margin was 62.4% (within guide, down on cloud/AI mix), and management said plainly it would "hurt our gross margins" to assure supply. Purchase commitments jumped to $8.9B and deferred revenue to $6.2B as qualification cycles stretched to 6–8 quarters, signal of backlog building faster than revenue.
- Rating: Maintaining Outperform. The Q4 initiation thesis is intact and, on the demand side, stronger. A 13.6% de-rating on a business whose only problem is that it cannot make product fast enough improves the risk/reward. We would use the pullback to build; supply-driven timing risk and the ~46% margin guide keep conviction measured rather than aggressive.
Results vs. Consensus
Arista reported fiscal Q1 2026 (quarter ended March 31, 2026) after the close on May 5. On the numbers alone, it was one of the cleaner prints in the company's recent history: a revenue beat, an EPS beat, record cash generation, and a raised full-year outlook. The stock's 13.6% drop the next session is a story about the setup, not the scorecard.
| Metric (Q1 2026) | Actual | Consensus | Beat/Miss | Magnitude |
|---|---|---|---|---|
| Revenue | $2.709B | ~$2.61B | Beat | +3.5% (above $2.6B guide) |
| Non-GAAP Gross Margin | 62.4% | ~62.5% | In line | Within 62–63% guide |
| Non-GAAP Operating Margin | 47.8% | ~47% | Beat | Above typical guide |
| Non-GAAP Operating Income | $1.29B | n/a | Beat | +35% YoY |
| Non-GAAP EPS | $0.87 | $0.81 | Beat | +7.4% |
| Non-GAAP Net Income | $1.11B | n/a | Beat | +33% YoY |
| Operating Cash Flow | $1.69B | n/a | Beat | Record |
Year-over-year comparison
| Metric | Q1 2026 | Q1 2025 | Change |
|---|---|---|---|
| Revenue | $2.709B | $2.005B | +35.1% |
| Non-GAAP EPS | $0.87 | $0.66 | +31.8% |
| GAAP EPS | $0.80 | $0.64 | +25.0% |
| Non-GAAP operating margin | 47.8% | 47.8% | Flat |
| Non-GAAP gross margin | 62.4% | ~64.1% | ~-170 bps |
| Billings growth | +54% | n/a | Up from +43% in Q4 |
Quarter-over-quarter comparison
| Metric | Q1 2026 | Q4 2025 | Change |
|---|---|---|---|
| Revenue | $2.709B | $2.488B | +8.9% |
| Non-GAAP gross margin | 62.4% | 63.4% | -100 bps |
| Non-GAAP EPS | $0.87 | $0.82 | +6.1% |
| International % of revenue | 15.5% | 21.2% | -570 bps |
| Operating cash flow | $1.69B | $1.26B | +$0.43B |
| Deferred revenue | $6.2B | $5.37B | +$0.83B |
| Purchase commitments | $8.9B | $6.8B | +$2.1B |
- Revenue: Broad-based and organic, led by AI and specialty providers, with enterprise strong in both data center and campus. Billings accelerating to +54% YoY (from +43%) is the cleanest forward-demand tell in the print. The international drop to 15.5% is a booking-geography artifact (large global customers served out of the Americas), not a demand signal.
- Margins: Gross margin of 62.4% held inside guidance but is down ~170 bps YoY on customer mix, with memory/silicon cost a secondary drag. Operating margin of 47.8% was defended by genuine opex discipline (opex roughly flat sequentially on a +8.9% revenue quarter).
- EPS: The $0.87 is cleaner than Q4's. The tax rate normalized to 21.1% (versus Q4's one-off 18.4%), so this quarter's YoY EPS growth of +31.8% is operationally driven rather than tax-flattered. Other income contributed a favorable $110.8M.
Segment & Mix Performance
Growth was led by AI and specialty providers, but the story this quarter is that all three customer sectors grew, and enterprise, which management is careful to call "not quite AI yet," was strong in both data center and campus. Four named enterprise wins gave unusual texture to the diversification claim.
Q1 revenue composition
| Line | Q1 2026 | Detail |
|---|---|---|
| Total revenue | $2.709B | +35.1% YoY; +8.9% QoQ; above $2.6B guide |
| International | $418.9M (15.5%) | Down from 21.2%; global customers served out of Americas |
| Operating expenses | $396.8M (14.6%) | R&D $271.5M (10%), S&M $103.5M (3.8%), G&A $21.8M (0.8%) |
| Other income | +$110.8M | Tax rate 21.1%; net income $1.11B (40.9%) |
Cloud & AI titans and specialty providers
AI and specialty providers led the quarter. Management confirmed the fourth of its original four Ethernet AI training customers has now "officially moved from InfiniBand to Ethernet at production scale," a milestone flagged as pending at Q4. Neocloud and sovereign cloud were called out as strong and underappreciated, and management gave a first explicit nod to TPU-based architectures joining AMD in a diversifying accelerator mix.
"Our demand is actually the best I've ever seen in my Arista tenure. The supply, however, is a slightly different and opposite tale." — Jayshree Ullal, Chairperson & CEO
Assessment: The demand narrative strengthened materially versus Q4: a completed InfiniBand-to-Ethernet migration, TPU attach, and record billings. This is the pillar carrying the thesis, and it is on track. The constraint on converting it to revenue is now supply, not orders.
Enterprise & financials
Enterprise delivered strong results in data center and campus, and management spent unusual airtime on four wins: a neocloud that moved off a white-box architecture to Arista's 800G EtherLink for AMD accelerators; a regional fiber-to-the-home provider modernizing its routing backbone on the 7280 platform; an insurance company deploying the R3 monitoring fabric for observability; and a global manufacturer rolling Cognitive Campus across 100-plus factory sites after an extensive bake-off against two incumbents.
"We're very happy what we're seeing in enterprise, which I wouldn't say is quite AI yet. So let's cover that as a non-AI bucket... we're very confident across all 3 that we're seeing strong demand." — Chantelle Breithaupt, CFO
Assessment: This directly rebuts the Q4 bear point that the non-AI core was guided to roughly zero. Enterprise is the margin ballast (it carries higher gross margin than the titans) and the diversification hedge, and it is executing. The win detail is evidence, not just assertion.
Campus & adjacencies
Campus remains a $1.25B goal for 2026, and the VeloCloud acquisition is integrating well, adding a managed-service-provider channel motion. The manufacturing win showcased the single-EOS-binary value proposition across campus, data center, and WAN.
Assessment: Campus is on plan and increasingly a credible second growth vector, though it is dwarfed in the narrative by AI. The MSP channel is a quietly important distribution expansion for the enterprise/branch business.
Key Financial & Balance-Sheet KPIs
| KPI | Q1 2026 | Q4 2025 | Read |
|---|---|---|---|
| Cash & marketable securities | $12.35B | $10.74B | Net cash, no debt |
| Operating cash flow | $1.69B | $1.26B | Record; deferred build helps |
| Deferred revenue (total) | $6.2B | $5.37B | Product deferred +$643M QoQ |
| Purchase commitments | $8.9B | $6.8B | Multiyear supply lock-in |
| Inventory | $2.38B | $2.25B | Turns 1.7x (from 1.5x) |
| DSO | 64 days | 70 days | Shipment linearity |
| Buyback | $0 | $620.1M | $817.9M remains; paused this quarter |
Two lines tell the supply story. Purchase commitments leapt $2.1B to $8.9B as Arista locked multiyear supply at elevated cost, and deferred revenue rose to $6.2B as customer qualification cycles stretched to 6–8 quarters. The record $1.69B of operating cash flow was helped by that deferred build. The absence of any buyback, with $817.9M still authorized, is notable given the stock traded near its highs entering the print; management chose to conserve cash against the working-capital demands of the supply squeeze.
Key Topics & Management Commentary
Overall Management Tone: Confident on demand and candid on supply, with a clear tonal shift from Q4's framing of memory as a cost problem to Q1's framing of a broad, multiyear supply ceiling. Management leaned into the demand strength (the "best I've ever seen" line was unprompted and emphatic) while repeatedly conditioning the guide on shippability, not order book. It was assured about the business and disciplined about not over-promising the shape of the year.
1. The Beat-and-Raise the Market Punished
Arista beat, raised the full year to 27.7% growth, lifted the AI target, and guided Q2 above consensus, then watched the stock fall 13.6%. The gap between a good raise and a great expectation is the entire event. The Street had modeled 28–30% growth after a ~90% one-year run; 27.7% was not enough.
"It gives us confidence to increase our forecasted growth slightly to 27.7%, aiming now for $11.5 billion for 2026. We also increased our AI target now to $3.5 billion this year, thereby more than doubling our AI sales annually." — Jayshree Ullal, Chairperson & CEO
Assessment: The word "slightly" did real damage. This is an expectations reset, not a fundamental deterioration, and the historical Arista pattern is that the February/May full-year number is a floor that rises through the year. We read the reaction as opportunity rather than warning.
2. Demand Greater Than Supply: The Central Story
The defining message of the call was inverted from Q4. Then, the worry was the cost of memory; now, it is the availability of everything. Management described industry-wide shortages across wafers, silicon, CPUs, optics, and memory, with 52-week lead times and vendor de-commits, and framed it as a one-to-two-year condition.
"We are experiencing 52-week lead times pretty reliably with reservation needs beyond that, and our customers certainly do not want to wait that long." — Kenneth Duda, Co-President & CTO
Assessment: Supply-capped growth is a high-class problem, but it is still a governor. The $11.5B guide is a shippability number, so upside to it depends on the supply chain easing rather than demand appearing. The $8.9B of multiyear purchase commitments is management buying down that risk, and it is the right move even at the cost of margin.
3. Gross Margin: Willing to "Hurt" It to Ship
Gross margin was 62.4%, inside guidance, down on mix. The striking disclosure was management's explicit willingness to sacrifice margin to secure supply and keep customers' GPUs from sitting idle. Unlike peers, Arista has taken only modest price increases so far and continues to absorb much of the cost.
"We are going to do everything, including hurt our gross margins to supply to that demand this year and next year because we... certainly don't want to keep GPUs idle and AI infrastructures underutilized because Arista didn't supply the network." — Jayshree Ullal, Chairperson & CEO
Assessment: This is a share-and-relationship strategy over a near-term margin strategy, and for a category leader building a decade-long installed base, it is the correct trade. The offset is that pricing power is being deliberately withheld ("price increases really come into play once our backlog starts to reduce"), so margin relief is a later-cycle event, not a 2026 one.
4. AI Target to $3.5B; Scale-Across Emerges as One-Third
The AI fabric target rose to $3.5B, more than doubling annually. The architectural news is that scale-across (connecting distributed data centers via the 7800R3/R4 and coherent optics) is emerging faster than expected and will contribute at least one-third of the AI number, while scale-up remains a 2027–28 story gated by the ESUN spec and 1.6T silicon.
"Scale across will definitely contribute at least 1/3 of our AI number." — Jayshree Ullal, Chairperson & CEO
Assessment: Scale-across plays directly to Arista's routing and DCI heritage and is higher-value, routing-heavy hardware. That it is materializing in 2026 (ahead of the prior expectation) is an underappreciated positive, and scale-up remains free optionality on top of the guide.
5. The Named 10% Customers: Microsoft and Meta
For the first time in our coverage, management named its two greater-than-10% customers: Microsoft and Meta, both 10%-plus for over a decade. On the promised new 10% customers, the message was unchanged from Q4: one, maybe two are there in demand, but crossing the threshold depends on shipments.
"In terms of the new entrants, we still expect at least one, maybe two... certainly, in demand, we see 1 or 2. We shall see... how we do on shipments to see if we can achieve the greater than 10%." — Jayshree Ullal, Chairperson & CEO
Assessment: Naming Microsoft and Meta removes some of the opacity that fed the concentration critique, and both are among the best-capitalized buyers on earth. The new-10%-customer question is now explicitly supply-gated, consistent with the quarter's theme; it remains a pending, not a broken, item.
6. Deferred Revenue and the 6–8 Quarter Qualification Cycle
Deferred revenue rose to $6.2B, with product deferred up $643M. Management explained the build as a function of both customer readiness (space, power, GPUs, and literally months of manual cabling) and new-product qualification, with acceptance cycles that have lengthened from 2–4 quarters historically to 6–8 quarters now.
"The length of time to qualify this, which used to be 2 to 4 quarters has extended more like 6 to even 8 quarters. So it's gotten much longer." — Jayshree Ullal, Chairperson & CEO
Assessment: A rising deferred balance with lengthening acceptance is backlog building faster than reported revenue, a demand positive, but it also means revenue timing is increasingly disconnected from bookings. The CFO's reminder that deferred is recognized in and out every quarter (not simply piling up) is the right nuance; the balance is a leading indicator, not a warning.
7. XPO: A New Optics Innovation Narrative
Arista unveiled its extended pluggable optics (XPO) form factor at the Optical Fiber Conference: 12.8 Tbps per module, 204.8 Tbps per OCP rack unit, 400W cooling, and now endorsed by more than 100 vendors. Management positioned XPO as the successor to OSFP for the next decade and a partner (not replacement) at 400G/800G, becoming the connector of choice at 1.6T/3.2T.
"Just as the last decade was greatly influenced by OSFP. The next decade will be greatly influenced by XPO. And remember, 99% of the optical market today that we connect to is all pluggable optics." — Jayshree Ullal, Chairperson & CEO
Assessment: XPO is an ecosystem-enabling standard that Arista profits from as it grows, and it undercuts the co-packaged-optics bear narrative by offering a standard, interoperable path to liquid-cooled density. It is optionality on content-per-rack, not a 2026 revenue driver.
8. Diverse Accelerators: AMD, TPU, and Blue Box
The accelerator mix keeps broadening. After highlighting AMD at 20–25% of deployments last quarter, management added a first explicit nod to TPU attach in scale-across, multi-tenant use cases. The neocloud win detailed on the call moved from white box to Arista's blue box specifically to connect AMD MI-series XPUs.
"The diversity of accelerators is creating tremendous multi-accelerator opportunity and multi-protocol features that we can provide for them in our network." — Jayshree Ullal, Chairperson & CEO
Assessment: Every incremental accelerator (AMD, TPU, custom XPU) increases the value of a vendor-neutral Ethernet fabric and blunts the "GPU vendor bundles the network" bear case. Blue box adoption in neocloud is also evidence Arista can defend against white box with EOS diagnostics and SDK depth.
9. Front-End to Back-End Ratio Shifts to ~1:1
Management updated its framing of front-end versus back-end networking intensity. Previously described as a 2:1 pressure (back-end driving front-end), it is now closer to 1:1 as customers pour resources into the back end first, with front-end refreshes following. The single-EOS-binary advantage across both was reiterated as a differentiator only Arista offers.
"At the moment, I would say it's more one-to-one. And at the moment, I'd also say the scale across in the back end has become a bigger use case than we imagined this time last year." — Jayshree Ullal, Chairperson & CEO
Assessment: The near-term back-end weighting is where the dollars are, and the front-end refresh is deferred demand that supports multi-year durability. The unified operating model across front and back end is a genuine, and possibly unique, competitive moat.
10. Agentic AI and Inference: The "Calm Before the Storm"
On the longer arc, management characterized training as still the dominant workload, with distributed inference and agentic AI in early trials, deploying in clusters of a few thousand accelerators rather than the 100,000-GPU titan scale, often HPC-like and CPU-inclusive.
"This is the calm before the storm, if you will. And as the AI gets more distributed, I think it doesn't need GPUs alone, it's going to need more high-performance compute... I think it's going to take a couple of years to fully happen." — Jayshree Ullal, Chairperson & CEO
Assessment: This frames a second, later demand wave (enterprise inference/agentic) behind the current training buildout. It is not in numbers yet, but it extends the runway and supports the argument that the AI networking cycle is multi-year rather than a single capex pulse.
11. Record Cash Flow and a Paused Buyback
Operating cash flow of $1.69B was the strongest in Arista's history, and cash climbed to $12.35B. Yet the company repurchased no stock in the quarter despite $817.9M of remaining authorization.
"We generated approximately $1.69 billion of cash from operations in the period, the strongest in the history of Arista." — Chantelle Breithaupt, CFO
Assessment: Pausing buybacks with the stock near highs and cash needs rising for supply commitments is defensible capital discipline. With the stock now 13.6% lower, we would watch for a resumption; a buyback into this pullback would be a constructive signal of management's own valuation view.
Guidance & Outlook
| Metric | Prior (Q4 print) | New (Q1 print) | Change |
|---|---|---|---|
| FY2026 revenue | ~$11.25B (+25%) | ~$11.5B (+27.7%) | Raised (below Street's 28–30%) |
| FY2026 AI fabric | $3.25B | $3.5B | Raised |
| FY2026 campus | $1.25B | $1.25B | Maintained |
| FY2026 gross margin | 62–64% | 62–64% | Reiterated |
| FY2026 operating margin | ~46% | ~46% | Maintained |
| Q2 2026 revenue | n/a | ~$2.8B | Above ~$2.78B Street |
| Q2 2026 gross margin | n/a | 62–63% | New |
| Q2 2026 operating margin | n/a | 46–47% | New (below Q1's 47.8%) |
| Q2 2026 EPS | n/a | ~$0.88 | New metric; above ~$0.86 Street |
The guide is a shippability forecast, which is the single most important interpretive point. Management repeatedly said the $11.5B reflects what it can supply, not what customers have ordered, so upside now hinges on the supply chain rather than demand. The new addition of quarterly EPS guidance is a modest transparency improvement.
Implied ramp: With H1 tracking near $5.5B ($2.709B plus the ~$2.8B Q2 guide), the ~$11.5B full year implies roughly $6.0B across H2, i.e., continued sequential growth. Street at: the Q2 guide edged consensus on both revenue and EPS, but the full-year growth rate of 27.7% landed below the Street's 28–30% zone, which is what drove the reaction. Guidance style: conservative by Arista's own pattern, now doubly so because it is capped by supply; we expect the number to rise through the year if lead times improve at all.
Analyst Q&A Highlights
Is supply the ceiling on this year's and next year's growth?
The most important exchange of the call went straight at whether the 27.7% guide is a demand number or a supply number. Management was unusually direct that it is the latter, and that the constraint is multi-year.
Q: "With the supply outstripping demand, I'm wondering how much does your current supply allow you to grow this year and next? Is the updated top line growth guide of 28% growth a good reflection of how much supply you've secured for this year?"
— Antoine Chkaiban, New Street Research
A: "This is a reflection of not just demand, but how much we can ship this year... We started out at 20%, were at 25%, now we are at 27.7%. Could we improve to the tail end of the year? We'll see. But the amount of de-commits we're seeing doesn't feel good. So we think a lot of this will continue into next year and keep us constrained for the next couple of years." — Jayshree Ullal, Chairperson & CEO
Assessment: This is the quarter in one answer. Demand is not in doubt; the question is throughput. The successive raises (20% to 25% to 27.7%) support the "floor that rises" read, but the "de-commits don't feel good" candor is why the market got nervous about the pace of conversion.
Why should gross margin recover toward 63%, and how big is the product constraint?
Questioning pressed on the margin path and whether supply cost Arista revenue in the quarter. Management downplayed a near-term revenue hole and pinned any margin recovery on mix, while clarifying its deliberate pricing restraint.
Q: "Are you able to say what the [product constraint] number was in the quarter and what it's taking away in terms of the $2.8 billion guide?... And why the gross margin should go back up to 63%. What is it that you guys are doing that gives us confidence that it can actually expand a tad from here?"
— Benjamin Reitzes, Melius Research
A: "The only chance for mix expansion or margin expansion would be due to mix... while we would consider and have raised prices a little bit, unlike our competitors, we haven't done 2 pricing increases... the price increases really come into play once our backlog starts to reduce... So our gross margins are a strong factor of cost going up and are still eating a lot of the costs and giving our customers the benefit." — Jayshree Ullal, Chairperson & CEO
Assessment: Margin recovery is a mix story, not a pricing story, in 2026. Management is intentionally sitting on its pricing power to protect customer relationships while backlog is elevated. That caps near-term margin upside but strengthens the multi-year share position.
Where do the new 10% customers stand, and who are the current two?
A recurring line of questioning revisited the Q4 promise of one-to-two new 10% customers. Management named the incumbents and reiterated the new entrants are demand-present but shipment-gated.
Q: "Last quarter, you had alluded to engagements with other hyperscale cloud titan customers... having 1 or 2 new 10% customers this year. I'm curious where we stand today?"
— Aaron Rakers, Wells Fargo
A: "2 big ones... Microsoft and Meta, they're all-time favorites. They've been 10% and greater customers for over a decade... In terms of the new entrants, we still expect at least one, maybe two... certainly, in demand, we see 1 or 2. We shall see... how we do on shipments to see if we can achieve the greater than 10%." — Jayshree Ullal, Chairperson & CEO
Assessment: Naming Microsoft and Meta is a small but real reduction in the opacity that fed the concentration critique. The new-customer answer is unchanged and now explicitly supply-gated, which is thematically consistent but leaves the diversification catalyst unrealized for now.
What has to happen for deferred revenue to convert?
With deferred nearly doubling year over year, an analyst asked what conditions trigger recognition. Management walked through the customer-readiness and product-qualification gates, and quantified the lengthening cycle.
Q: "Deferred revenues doubled in the last year... What are the conditions to recognize deferred revenues? Meaning what needs to happen for deferred revenues to be recognized over the next few quarters?"
— Tal Liani, Bank of America
A: "There's a customer aspect and a product aspect... they need to have the space, the facilities, the GPUs... in many cases, they need to manually install the cables, and that takes several months... the length of time to qualify this, which used to be 2 to 4 quarters, has extended more like 6 to even 8 quarters." — Jayshree Ullal, Chairperson & CEO
Assessment: The lengthening acceptance cycle explains both the deferred build and why revenue increasingly lags bookings. It is a demand-positive backlog signal, but it makes any single quarter harder to model, and it is one reason the market discounts the "raise" until it converts.
How material is scale-across, and how is share shaping up early?
Multiple analysts probed the emerging scale-across opportunity, its contribution to the $3.5B AI target, and Arista's early competitive position as customers design distributed fabrics.
Q: "How much revenue, if any, did [scale across] contribute last year? And how material is that to the $3.5 billion forecast you're giving this year?"
— Simon Leopold, Raymond James
A: "Last year, on scale across, we were just beginning... scale up is virtually 0 and nonexistent because it really only comes to play after the ESUN spec. So consider that more a '27, '28 kind of number... scale across will definitely contribute at least 1/3 of our AI number." — Jayshree Ullal, Chairperson & CEO
Assessment: Scale-across going from "just beginning" to one-third of a $3.5B target in a year is a fast-emerging, high-value opportunity that suits Arista's routing strength. It also implies the AI mix is shifting toward premium platforms (the 7800), a modest margin positive within the AI bucket.
Is the gross-margin dip mix or memory?
A question sought to separate the customer-mix effect from the memory/silicon cost effect on gross margin. The CFO was clear that mix dominates.
Q: "[Gross margins] dipped about 170 bps year-on-year... whether it was primarily mix related or maybe if you can quantify how significant the memory and cost-related impacts were?"
— Matt Niknam, Truist
A: "The majority of the difference is mix of the customers... our larger customers have a lower gross margin accretion... And then the secondary, although not as significant, would be things depending on the quarter... the memory cost or the silicon cost. So secondary driver, but the primary driver is mix of the customer segments." — Chantelle Breithaupt, CFO
Assessment: Framing margin pressure as primarily mix, not cost, is important: mix is a function of winning the largest AI deals (a good problem), and it is structurally different from an uncontrolled cost spiral. It supports the view that margin holds inside the 62–64% guide.
Where are the holes in inventory and purchase commitments?
An analyst asked, with ~$2.4B of inventory and nearly two years of COGS in purchase commitments, where the supply gaps actually are. Management located the problem squarely in lead times and the newest silicon.
Q: "With $2.4 billion almost of inventory and almost 2 years in COGS of purchase commitments, how should we think about the supply constraints and where that inventory and purchase commitments are not satisfactory to meet demand? Where are the holes?"
— Andrew (for David Vogt), UBS
A: "The real hole is lead times. We are experiencing such significant wafer fab shortages that we're not getting the chips in time... our purchase commitments are multiyears because we're having to deal with forecasts that are out multiple years so that we get them in time." — Jayshree Ullal, Chairperson & CEO
Assessment: The constraint is timing, not access; Arista's spending power and reliability make it a priority customer, but 52-week lead times mean 2026 revenue is set by chips ordered well in advance. This is the clearest articulation of why the guide is supply-shaped.
What They're NOT Saying
- How much revenue supply is costing: Management would not quantify the revenue that constraints are deferring out of 2026, or how much higher the guide would be in an unconstrained world, beyond acknowledging demand exceeds supply.
- A firm gross-margin exit rate: The path back toward 63% was left entirely to "mix," with no commitment on timing or magnitude, and pricing power was explicitly deferred until backlog normalizes.
- Which new customers are near 10%: The one-to-two new 10% customers remain unnamed and are now shipment-gated, so the diversification catalyst has no confirmed timeline.
- 2027 growth framing: Despite multiyear purchase commitments and repeated references to a "one-to-two-year" supply condition, management declined to frame 2027 growth, saying only that constraints will persist.
- Why the buyback was paused: No explicit rationale was given for repurchasing zero stock with $817.9M authorized and the shares near highs; the working-capital-conservation read is ours, not stated.
Market Reaction
- Pre-print setup: ANET closed at $170.22 entering the print, up 29.9% YTD, up 88.3% over the trailing 12 months, and up 34.4% in just the prior 30 days, near the top of its 52-week closing range ($86.25–$177.73) at roughly 48x forward earnings. The bar was extreme.
- Reaction (T+1, May 6): The stock gapped down 10.1% to open at $153.00, traded as low as $140.58 (-17.4%), and closed down 13.6% at $147.01 on volume of 35.6M shares, about 5.1x the 30-day average. The S&P 500 rose 1.5% on the session, so the decline was entirely company-specific.
This is a textbook "punished for perfection." A double beat and a raised guide could not clear a bar the stock's own 90% run had set, and the reaction concentrated on three worries: a full-year growth raise (27.7%) below the Street's 28–30% zone, a multi-year supply ceiling that caps how fast demand converts, and gross-margin pressure that management is choosing to absorb. A director's small pre-planned 10b5-1 sale on May 1 added marginal noise. The important point for a fundamental owner: none of the three worries is a demand problem, and the de-rating from ~48x to ~40x forward improves the entry into a business whose order book is accelerating (billings +54%).
Street Perspective
Debate: Does the 13.6% sell-off create a genuine entry point, or is it a valuation reset with further to go?
Bull view: The bull case on the Street holds that the drop is pure expectations-resetting on a business that beat every metric, generated record cash flow, and is accelerating on billings; at ~40x forward for a 27%-plus grower with 46% operating margins and net cash, the risk/reward improved.
Bear view: The bear camp argues the stock was priced for 30%-plus and flawless execution, that supply constraints and margin pressure are now structural for one-to-two years, and that ~40x forward is still rich enough to keep de-rating if the macro or AI-capex narrative wobbles.
Our take: The bulls have the better argument. The problems are supply and expectations, not demand, and Arista's history of raising a conservative guide through the year is intact. We would add on the pullback rather than wait for a cleaner tape.
Debate: Is supply-constrained growth a warning sign or a high-class problem?
Bull view: Some sell-side desks frame demand outstripping supply as the best problem a hardware company can have: it implies backlog, pricing optionality later, and multi-year visibility, and Arista's $8.9B of purchase commitments locks the supply that competitors may not secure.
Bear view: Skeptics counter that a supply ceiling means the reported growth understates lost revenue, that de-commits could worsen, and that competitors facing the same shortages could win share if Arista cannot ship.
Our take: It is a high-class problem, with a caveat. Arista's scale, balance sheet, and vendor relationships make it a priority buyer, so it is better positioned than most to ship through the shortage, but the constraint does convert some 2026 upside into 2027, which is why the stock will trade on lead-time headlines.
Debate: Does gross-margin compression threaten the earnings-compounding story?
Bull view: A growing consensus view is that margin pressure is mostly mix (winning the biggest AI deals) rather than cost, that opex discipline is holding operating margin near 46–48%, and that Arista has withheld pricing power it can deploy later.
Bear view: The bear camp notes gross margin has fallen from ~65% to ~62% and management is openly willing to "hurt" it further, so the model's historical ~48% operating margin may be structurally lower, capping EPS growth below revenue growth.
Our take: Both are partly right, but the mix framing wins. Margin at the low end of a growing revenue base still compounds earnings at a high absolute rate, and the withheld pricing power is a real, if later-cycle, lever. We model operating margin near 46% and EPS compounding in the mid-to-high 20s.
Model Update Needed
| Item | Prior model | Suggested change | Reason |
|---|---|---|---|
| FY2026 revenue | ~$11.25B | ~$11.5B, bias higher | Guide raised to 27.7%; supply-capped, so upside is timing-dependent |
| FY2026 gross margin | 62.5–63.5% | 62–63% | Q1 at 62.4%; mix + deliberate cost absorption to secure supply |
| FY2026 operating margin | ~46% | ~46% | Opex discipline holding; unchanged |
| AI fabric revenue | $3.25B | $3.5B | Target raised; scale-across now ~1/3 of the number |
| FY2026 non-GAAP EPS | ~$3.55 | ~$3.55–3.65 | Q1 $0.87 + Q2 guide $0.88; H2 sequential build |
| Buyback | Steady | Watch for resumption | Zero repurchase in Q1; stock now 13.6% lower |
Valuation impact: Estimates move modestly higher on the revenue raise, but the de-rating does most of the work: at the $147.01 reaction close, ANET trades near 40x forward and ~47x trailing, down from ~48x forward pre-print. On unchanged fundamentals and a better entry, the risk/reward improves. We frame fair value as tracking earnings higher through the year, with the supply chain the swing factor; the Outperform remains an earnings-compounding call.
Thesis Scorecard Post-Earnings
Grading the standing thesis established at the Q4 initiation.
| Thesis Point | Status | Notes |
|---|---|---|
| Bull #1: AI networking secular leadership | Confirmed (strengthened) | #1 >10GbE share; 4th customer fully on Ethernet; TPU attach added; billings +54%; AI target to $3.5B |
| Bull #2: Conservative beat-and-raise cadence | Confirmed with nuance | Raised again (25%→27.7%), but below Street's 28–30%; the "floor" held, the bar did not |
| Bull #3: Software/EOS stickiness & operating leverage | Confirmed | Op margin 47.8%; record $1.69B OCF; opex flat sequentially; single-EOS front/back-end moat |
| Bear #1: Cost/supply pressure on margin | Confirmed (escalated) | Now a broad multi-year supply ceiling + willing to "hurt" GM; 52-week lead times, de-commits |
| Bear #2: Customer concentration | Neutral (improving) | Microsoft & Meta named; neocloud/enterprise diversifying; but no new 10% customer confirmed yet |
| Bear #3: Full valuation | Challenged (relieved) | -13.6% de-rating cuts forward multiple from ~48x to ~40x; premium partly worked off |
Overall: Thesis strengthened on demand and partly de-risked on valuation, with the key debate migrating from "is demand real" to "how fast can supply convert it." The margin/supply bear point escalated but remains a governor on pace, not a threat to the secular case.
Action: Maintain Outperform; use the pullback to add. The two variables that would move us to Hold are a sequential decline in deferred revenue (backlog rolling over) or evidence that supply constraints are ceding share to competitors rather than merely delaying Arista's own revenue.