Supply Was the Ceiling; Arista Just Raised It by $1.1 Billion

Results vs. Consensus

Arista reported Q2 2026 (quarter ended June 30, 2026) after the close on August 4. The scorecard is unusually one-sided: every line beat, the beats widened as they moved down the income statement, and the full-year outlook was raised for the third time in six months. Where Q1 was a good quarter the market rejected, Q2 was a very good quarter the market accepted for about three sessions.

Versus the company's own Q2 guidance (issued May 5)

The gap between the guide and the print is the most informative number in the release. Arista guided Q2 operating margin down from Q1's 47.8% to a 46-47% range, and delivered 49.9%. A company that had spent two quarters telling investors margin was the price of securing supply instead posted its best operating margin of the cycle.

Year-over-year comparison

Prior-period non-GAAP figures were recast this quarter. Arista changed its methodology for calculating the income tax effect of non-GAAP adjustments effective Q4 FY2025, and retrospective application raised Q2 2025 non-GAAP net income by $10.7M. The per-share impact on Q2 2025 was nil, so the $1.02 versus $0.73 comparison is clean; the six-month comparison carries a $0.03 recast benefit in the prior-year base.

Quarter-over-quarter comparison

Revenue

The headline is the first $3 billion quarter, but the more useful number is the shape of the year it implies. First-half revenue of $5,744.7M is up 36.5% against the prior-year first half. The $12.6B full-year guide, less that first half and less the ~$3.3B Q3 guide, implies a fourth quarter near $3.56B. That puts implied second-half growth at roughly 43%, accelerating from the first half, in a business that was doing $2.9B annually five years ago. Acceleration at this revenue base is the single most striking fact in the print.

Margins

Two quarters ago the framing was that memory cost would push operating margin to about 46%. One quarter ago it was that management would "hurt" gross margin to assure supply. This quarter operating margin printed 49.9% and the full-year target went to 48-49%. Gross margin has not recovered to prior-year levels and management is not claiming it will: the FY range stays 62-64%, and pricing power is still deliberately withheld. What changed is the denominator. Revenue is growing fast enough that a nearly flat opex dollar base converts directly into margin, and the mix headwind at the gross line is being more than offset below it.

EPS

Non-GAAP EPS of $1.02 grew 39.7% against revenue growth of 37.7%, so the model is compounding earnings slightly faster than sales even with gross margin down 220 bps YoY. That is the operating-leverage argument working in real time. The gap between GAAP EPS of $0.95 and non-GAAP $1.02 is $0.07, driven mostly by $120.4M of stock-based compensation (4.0% of revenue), which is disclosed, consistent with prior quarters, and unremarkable for the sector.

Segment & Mix Performance

Arista does not report operating segments in its earnings release, so the disclosed mix is product versus service and domestic versus international, supplemented by the annual customer-sector and product-category framing management gives on calls. The composition detail that mattered this quarter was what management would not break out: the $1.1B guidance raise was left deliberately unallocated.

Revenue composition

Non-GAAP operating expense detail

R&D dollars rose only 2.4% sequentially while revenue rose 12.1%, which is what produced most of the operating-margin beat. Management attributed the increase to "an additional investment in liquid cooling, high Radix switching and AI optimizing software," so the spend is going where the roadmap is. The pattern to watch is whether R&D at 9.2% of revenue is a durable new level or simply a quarter in which revenue outran hiring.

AI fabrics

The disclosed AI metric moved from customer count to installed breadth. Etherlink AI fabrics now exceed 100 cumulative customers, against the four to five that management first described in 2024. That is the clearest quantification yet that the AI business has broadened past the original titan cohort, and it reframes the concentration debate: the risk is no longer whether anyone besides the two 10% customers buys AI fabrics, but how much of the revenue those hundred-plus customers represent, which Arista still does not disclose.

"Our AI fabrics momentum with Etherlink switches now exceeds 100 cumulative customers from the initial 4 to 5 customers I spoke of in 2024." — Jayshree Ullal, Chairperson & CEO

Assessment: Customer count is the diversification evidence the concentration bear point has been waiting for, and it arrived without a corresponding revenue disclosure. The AI fabrics goal itself stayed at "at least $3.5 billion," so the count grew while the dollar target did not, which tells you the average new customer is still small relative to the titans.

Scale-across

Scale-across, connecting distributed data centers so that compute in one location can be pooled with another, was flagged at Q1 as emerging faster than expected. It is now sized: approximately 30% of the AI target for 2026, which management put at roughly $1.2 billion when pressed, against a market management sizes at $3-4B today and $15-20B by 2030.

"The Scale-Across switching and routing TAM is forecasted to be roughly $15 billion to $20 billion in 2030, and Arista is well poised in this segment." — Jayshree Ullal, Chairperson & CEO

Assessment: If Arista is capturing roughly $1.2B of a $3-4B market, it holds something like a third of a category it effectively defined, on its highest-value routing platform. The 2030 TAM is a management-cited third-party forecast and should be treated as directional rather than bankable, but the near-term share position is real revenue today. This is the most attractive line in the AI story: high-value hardware, structural rather than cyclical demand (power and space scarcity do not resolve), and a competitive field that has not formed yet.

Core data center, campus, and routing

The non-AI business was the bear point coming out of Q1, where the implied core growth rate looked close to flat. Management addressed it directly and reversed the framing: the incremental $1.1B is expected to come substantially from core data center front-end, enterprise, campus, and routing rather than from AI. Campus stays at a goal of at least $1.25B for the year.

"So if you take out the campus and AI, which we believe will grow, then one of the things that our new software analysts tell us and questioning me on is why is your core data center front end so flat. And so I'd like to first make the observation that it isn't going to be flat, but we couldn't answer that question in Q1." — Jayshree Ullal, Chairperson & CEO

Assessment: This is the most consequential mix statement on the call and it resolves the standing Q1 bear point in the bulls' favor. Core and campus revenue carries higher gross margin than titan AI business, so a raise weighted toward those categories is margin-accretive, which is a coherent explanation for why the operating-margin target went up at the same time as the revenue target. The caveat is that this remains an assertion about the back half; it has not yet shipped.

Key Financial & Balance-Sheet KPIs

The cash-flow line deserves attention because it moved the wrong way on a much better quarter. Operating cash flow of roughly $1.1B was down from Q1's record $1.69B despite net income rising, and the six-month statement shows why: inventories consumed $288.2M, receivables $379.3M, and other assets $619.7M, against a $1,493.5M deferred-revenue inflow. Growth at this rate is working-capital hungry, and management said plainly that lumpiness should be expected: "Inventory level fluctuations are expected to continue as we work through the balancing of component timing and availability." Deferred revenue rising to $6.87B while operating cash flow fell is not a contradiction; it means the deferred build no longer outruns the inventory and receivables build the way it did in Q1.

Key Topics & Management Commentary

Overall Management Tone: Confident and, for the first time in three quarters, forward-leaning rather than defensive. Where Q1 was spent explaining a constraint, this call was spent demonstrating what had been done about it, with an operations executive brought onto the call to walk through the fixes in detail. The one place management was deliberately unhelpful was on composition: it repeatedly declined to say where the incremental billion dollars will come from or which customers will cross 10%, framing that opacity as shipping flexibility rather than uncertainty.

1. The Third Raise: $11.5B to $12.6B

The full-year guide has now moved three times, and this move was by far the largest. From a roughly $10.5B goal set at the Analyst Day, to $11.25B in February, to $11.5B in May, to $12.6B now: a 40% growth rate against FY2025 revenue of $9,005.7M. The $1.1B increment is larger than the two prior raises combined, and it arrived one quarter after management had characterized the guide as a shippability floor rather than a demand estimate.

"Given our improving stance in supply chain, we are excited to increase our guidance for the third time this year to $12.6 billion revenue in 2026. We are now projecting 40% annual growth, which is an incremental $2.1 billion over our Analyst Day goal of $10.5 billion and an incremental $1.1 billion over our recent projections of $11.5 billion in May of 2026." — Jayshree Ullal, Chairperson & CEO

Assessment: Our Q1 read was that the $11.5B number was a floor that would rise if supply eased at all. It rose by 9.6% in a single quarter. That validates the beat-and-raise pillar emphatically, and it also raises the bar: a guide that has moved three times in six months is no longer being treated by the market as conservative, which removes some of the cushion that made prior raises look like layups.

2. What Actually Changed in the Supply Chain

Rather than assert improvement, management gave the call over to its operations leadership for a specific inventory of what was fixed. The list is concrete: multiyear agreements with strategic component vendors, new qualified suppliers for risk reduction, memory secured for 2026 with visibility into 2027 across DDR4, DDR5 and NAND, capacity built in a 12-month window for PCBs and optics, better delivery terms, a newly established liquid-cooling supply chain, and geographic diversification to three contract manufacturers and three distribution facilities.

"Our memory supply has been secured for 2026, and we have extended visibility well into 2027 across DDR4, DDR5 and NAND memory." — Todd Nightingale, Chief Operating Officer & President

Arista also hired a global operations executive, Eugenia Corrales, and framed the response through the lens of prior supply crises in 2016 and 2021. Purchase commitments now stand at approximately $9.7B, from $3.6B a year ago.

Assessment: This is a company-specific fix inside an industry-wide problem, and the distinction matters. Arista has not solved the shortage; it has bought, qualified, and contracted its way to the front of the queue. That advantage is real but it is also purchased, and $9.7B of multiyear commitments, against a cost of revenue running near $4.6B annualized on the new full-year guide, is a substantial forward obligation if demand ever disappoints.

3. Operating Margin: The Raise Nobody Guided For

The FY2026 non-GAAP operating margin target moved to 48-49% from approximately 46%, a raise of roughly 250 bps at the midpoint, while the gross-margin range was left untouched at 62-64%. The entire increase therefore comes from operating leverage, not from cost relief.

"We have increased our fiscal 2026 operating margin target now at a range of 48% to 49%, while maintaining an expected tax rate of 21.5%." — Chantelle Breithaupt, CFO

The mechanics are visible in the quarter. Non-GAAP opex of $411.0M was 13.5% of revenue against 14.6% in Q1, and the 210 bps of sequential operating-margin expansion splits roughly evenly between the 100 bps gross-margin recovery and 110 bps of opex leverage.

Assessment: This is the most underrated disclosure of the quarter, and it is worth more to the earnings model than the revenue raise. A 250 bps margin target increase on a $12.6B revenue base is roughly $315M of incremental operating income, before any revenue upside. It also quietly reverses the Q4 2025 narrative that memory cost had structurally reset margins lower.

4. Gross Margin: Tariff Refund, Mix, and Still No Price Increase

Gross margin of 63.4% recovered 100 bps sequentially and beat the top of guidance by 40 bps. Management partitioned the beat carefully: 20-30 bps was a tariff refund, and the balance was customer mix. Notably, none of it was pricing, and management reiterated that price increases remain a later-cycle event.

"I would say the kind of the -- I think versus guide is probably the best way to do it versus guide, you'll see there's probably 20 to 30 bps of tariff refund and the remainder more of a mix equation." — Chantelle Breithaupt, CFO

Assessment: Strip the tariff refund and the underlying gross margin beat is 10-20 bps, which is a rounding error. The gross-margin story is unchanged from Q1: mix-driven, range-bound at 62-64%, with a pricing lever still deliberately holstered. Investors should model the margin improvement below the gross line, not at it.

5. The First $3 Billion Quarter

Management framed the milestone against the company's own history rather than against peers, and the comparison is genuinely arresting: a single quarter now exceeds what the entire company earned in revenue in 2021. Arista also entered the Fortune 500 and the Russell 50, and was named a Leader in the 2026 Gartner Magic Quadrant for Enterprise Wired and Wireless LAN.

"Arista is experiencing significant demand to achieve our first $3 billion quarter in revenue. To put this in perspective, just 5 years ago, our entire year was $2.9 billion in 2021." — Jayshree Ullal, Chairperson & CEO

Assessment: Scale milestones are usually decoration, but the Gartner recognition in enterprise LAN is substantive for the campus thesis: it is third-party validation in the one category where Arista is the challenger rather than the incumbent, and campus is where the diversification argument has to be won.

6. Scale-Across Gets a Number

Last quarter scale-across was described as "at least 1/3" of the AI number without a market context. This quarter it has both a share (approximately 30% of the AI target) and a market frame: roughly $3-4B today, growing to $15-20B by 2030. Management also argued the demand is structural rather than a temporary workaround.

"And the answer is no, they won't because we're going to be in a constant state of scarcity for power, space and compute." — Jayshree Ullal, Chairperson & CEO

Scale-across is currently dominated by cloud and AI titans, with early adoption spreading to neoclouds and little presence in enterprise.

Assessment: The scarcity argument is the load-bearing claim, and it is a good one: if power and land remain the binding constraints on AI buildout, distributing clusters is permanent architecture rather than a stopgap. That converts scale-across from a cyclical revenue line into a structural one. The risk is that it is currently concentrated in exactly the customers the concentration bear point worries about.

7. 1.6T Enters Trials; Production Is a 2027 Event

Arista introduced the 7060XE7, a 1.6-terabit platform delivering up to 100 Tbps of system bandwidth with Linear Pluggable Optics that cut interconnect power roughly 60% versus traditional pluggable optics, plus its first liquid-cooling options. Management was disciplined about the timeline, drawing the analogy to the 800G cycle: trials in the second half of 2026 with a single-digit number of very large customers, and real production in 2027.

"In terms of customers, it will be single digits, but very large customers. And I think the real production will still be in 2027." — Jayshree Ullal, Chairperson & CEO

Assessment: Correctly excluded from the 2026 numbers, which makes the $12.6B guide a pure 800G-and-below figure. That is a meaningful point for the 2027 setup: next year's growth does not depend on 1.6T, it gets 1.6T on top. The 60% interconnect power reduction is also a competitively relevant claim in an environment where power is the scarce input.

8. The Unallocated $1.1 Billion

The most-pressed question of the call was where the incremental revenue comes from, since neither the AI fabrics goal (at least $3.5B, with the CEO referencing "at least $3.6 billion" earlier in the call) nor the campus goal (at least $1.25B) was raised alongside the total. Management declined to allocate it, and was explicit that the refusal was deliberate.

"And that I would like to reserve that $1.1 billion question to, well, it depends on how we ship. If we ship more front-end AI or we ship more WiFi or wired or Etherlink switches or routing. So I'd like to give our customers the priority and Todd's team the flexibility to ship what we can. And that's why we're not holding ourselves to a number. But I think at this point, we can most certainly say all numbers are going up." — Jayshree Ullal, Chairperson & CEO

Assessment: Defensible operationally and unsatisfying analytically. The mix matters a great deal to margin, since core and campus carry better gross margin than titan AI business, and the raised operating-margin target implicitly encodes an assumption about that mix which investors cannot verify. We read the combination of an unraised AI target and a raised total as evidence the increment skews non-AI, which is thesis-positive, but it is an inference rather than a disclosure.

9. Blue Box Versus White Box: The Moat Question

The competitive question came from the opposite direction than usual: as AI networks get more complex across denser clusters, multiple data centers, and distributed fabrics, does system-level value rise relative to commodity hardware? Management argued yes for complex use cases while conceding the white-box market is permanent at the low end, citing operational simplicity, EOS feature depth, and reliability as the three reasons customers migrate to blue box.

"So I would say white box is certainly a tactical solution that we tend to see more in use cases that are simple, scale-up or scale-out where the actual amount of software and system requirements are low. But when you look at traditional network topologies, you can slow down the job completion significantly if you go with a box-by-box approach." — Jayshree Ullal, Chairperson & CEO

The technical substance behind the claim was unusually specific this quarter: Smart System Upgrade for non-disruptive software upgrades, Multipath Reliable Connection for spraying a single flow across many fabric paths, and SRv6 segment routing used for AI fabric load balancing across both scale-out and scale-across.

Assessment: Job-completion time is the right frame for this argument because it converts a software feature into an XPU-utilization dollar figure the customer already measures. The concession that white box persists at the commodity end is honest and does not damage the thesis; Arista has never competed there.

10. Scale-Up, NVLink, and the Non-NVIDIA Opening

On the third AI fabric category, management was candid that scale-up inside NVIDIA's vertical stack is largely closed to Arista and everyone else, and that the opportunity lies with non-NVIDIA accelerators: AMD's MI series, Google TPUs described as an increasingly formidable training processor, inference accelerators, and customer in-house silicon.

"I feel a little bit like 2, 3 years ago when we were talking about InfiniBand and now we don't mention InfiniBand, but it took 2, 3 years to move to Ethernet. So it will take time to go from a proprietary scale-up that's been around a long time with NVLink to these other alternatives, even if Ethernet is really good. But I expect us to do much better there." — Jayshree Ullal, Chairperson & CEO

Assessment: The InfiniBand analogy is the strongest form of the bull argument on scale-up, and management earned the right to use it by having been correct on the last one. It is also a multi-year option with zero embedded in the 2026 guide. The honest framing ("we live in an NVIDIA world") is more credible than a claim of imminent displacement.

11. Deferred Revenue and the Working Capital of Growth

Total deferred revenue reached approximately $6.87B, up from $6.2B, with product deferred up roughly $600M sequentially. Management repeated its caution that the balance is volatile and driven by customer-specific acceptance clauses on new products and use cases rather than by underlying demand.

"As mentioned in prior quarters, the deferred balance can move significantly on a quarterly basis, independent of underlying business drivers." — Chantelle Breithaupt, CFO

Alongside it, DSO rose to 68 days from 64 and inventory rose to $2.54B with turns flat at 1.7x. Operating cash flow of roughly $1.1B was well below Q1's $1.69B.

Assessment: The deferred build continuing was our number-two commitment to watch from Q1, and it was met: a sequential decline would have been the clearest bear signal, and it did not happen. The offset is that the working-capital intensity of this growth rate is now visible in the cash-flow statement. Cash conversion below net income for a quarter or two is a fair price for 40% growth, but it is the line where a demand disappointment would show up first.

12. Capital Allocation: A Second Quarter With No Buyback

Arista repurchased no stock for a second consecutive quarter, leaving $817.9M of the $1.5B May 2025 authorization unused, while cash and marketable securities grew to $13.34B.

"In the quarter, we did not repurchase our common stock. Of the $1.5 billion repurchase program approved in May 2025, $817.9 million remain available for repurchase in future quarters." — Chantelle Breithaupt, CFO

Assessment: This was our sixth commitment to watch and it was not met. We flagged after Q1 that a repurchase into the $147 level would have signalled management's own valuation view; none came, and the stock has since risen 28%. With $13.34B of cash, no debt, and $9.7B of purchase commitments to fund, the conservatism is explicable, but two quarters of a dormant authorization while the stock swung 30% is a passive stance from a management team that is otherwise acting decisively.

Guidance & Outlook

The interpretive point from Q1 has inverted. Three months ago the guide was explicitly a shippability number, capped by supply. This quarter management said it is guiding to what it is "confident we can get the supply for," with upside if supply loosens further, which is a subtly but importantly different statement: the ceiling is no longer the binding constraint, it is a confidence threshold.

Implied ramp: First-half revenue of $5,744.7M plus the ~$3.3B Q3 guide leaves roughly $3.56B implied for Q4 against the ~$12.6B full year. That is +43.0% YoY in Q3 and +42.9% in Q4, versus +36.5% in the first half. The guide therefore embeds an acceleration in the back half, not a fade. Sequentially it requires +8.7% in Q3 and +7.7% in Q4, which is below the +12.1% just delivered, so the shape is demanding on a YoY basis and undemanding on a sequential one.

Street at: The Q3 guide landed roughly 12% above consensus on revenue and roughly 18% above on EPS, which is a far wider gap than the "roughly in line" Q2 guide that triggered the Q1 sell-off. Full-year growth of 40% is well clear of any bar the Street was carrying into the print.

Guidance style: Historically conservative and now demonstrably so, having moved three times in six months by a cumulative $2.1B against the Analyst Day frame. The change worth noting is that the conservatism now rests on demand judgment rather than on supply capacity, and visibility is still only about two quarters.

Analyst Q&A Highlights

Why the total went up but the AI and campus targets did not

The dominant line of questioning on the call went straight at the composition of the raise: if the full year moved $1.1B while both named category targets were reiterated, what exactly grew? Management confirmed both components will rise but refused to say by how much, framing the refusal as preserving shipping flexibility.

Q: "So Jayshree, you took up the calendar year '26 revenue guidance substantially from last quarter. But if I remember correctly, you got the AI target unchanged and only touch the campus. Can you kind of explain kind of the thought process there?"
— David Vogt, UBS

A: "Yes. No, if you had to ask me whether our AI number or campus number will go up, I think Chantelle, Todd, Ken and I absolutely believe it will. The question is not whether it will go up. The question is what is that number? And that I would like to reserve that $1.1 billion question to, well, it depends on how we ship."
— Jayshree Ullal, Chairperson & CEO

Assessment: A dodge, but an operationally coherent one. The useful inference is what management chose not to say: if the increment were predominantly AI, raising the AI target would have been the easiest and most market-friendly disclosure available. Leaving it flat while raising the total points the increment toward core and campus, which is better for margin and better for the diversification argument.

Where the incremental revenue actually comes from

A follow-on question pressed the same issue from the other side, asking for the source of the incremental billion if not AI or campus. This produced the most directly thesis-relevant answer of the call: an explicit reversal of the flat-core framing that had hung over the stock since Q1.

Q: "I wanted to follow up on the incremental $1 billion for the year. In that, I think you've reiterated the campus and AI. So I'd like to try to get a better understanding of the composition or the source of the incremental revenue if it's not the AI or campus portion?"
— Simon Leopold, Raymond James

A: "Okay. Actually, that's a great question, Simon, because in a nutshell, the answer is this is our core Arista product line we're talking about. Campus and AI are the newcomers to it. So if you take out the campus and AI, which we believe will grow, then one of the things that our new software analysts tell us and questioning me on is why is your core data center front end so flat. And so I'd like to first make the observation that it isn't going to be flat, but we couldn't answer that question in Q1."
— Jayshree Ullal, Chairperson & CEO

Assessment: This retires the single most durable bear point from the Q4 and Q1 reports. The core data center front end was the part of the business the guide had implied would be roughly flat, and management now says it will not be, having spent two quarters declining to commit. Combined with the previous exchange, it means the raise is weighted toward the higher-margin, more diversified part of the business.

What actually improved in the supply chain

Questioning probed whether the supply improvement reflected genuine capacity gains or simply Arista paying up, and asked specifically what had changed since the prior quarter's decommit disclosure. The answer resisted the temptation to declare victory.

Q: "Maybe I'll ask about the supply chain improvements. Last quarter, you talked about a supplier that was decommitted upstream. Maybe you can just expand a little bit on some of the efforts this quarter... Are they getting more capacity? Are you offering more pricing on your side? Are you getting more customer advocacy?"
— Michael Ng, Goldman Sachs

A: "Absolutely, Michael. So listen, I don't want you to believe that suddenly, we waved the magic wand and all our problems have gone away. The industry is going to have a 2-year problem. I don't think we get out of it as an industry until 2028. But Arista is taking individually and specifically steps in the first half of this year that we believe will have results in the back half of this year."
— Jayshree Ullal, Chairperson & CEO

Assessment: The most credible answer on the call precisely because it refused to over-claim. The distinction drawn between an industry problem lasting to 2028 and company-specific mitigation landing in the back half of 2026 is exactly the right frame, and it is consistent with a guide that rose without the shortage ending. It also implies the relative advantage over less-resourced competitors persists for two more years.

How much of the gross-margin recovery was durable

With gross margin up 100 bps sequentially despite rising component costs, questioning sought to separate one-time benefits from mix and from any pricing action. The CFO partitioned it precisely and ruled pricing out entirely.

Q: "A question on gross margin. So they improved about 100 bps sequentially despite some higher costs. So a, maybe Chantelle, can you help us think about the tariff refund benefit in the quarter? And then just sequentially, was the lift more mix related? Or was there any sort of pricing relief that's flowing through to offset some of the higher costs?"
— Matt Niknam, Truist

A: "So I would put that 20 to 30 for the quarter versus the guide. Probably in the range of maybe 30 to 40 bps if you look versus prior year, but it's in that kind of range. And the rest will be customer mix oriented. I wouldn't put anything in there yet due to a price increase in that equation."
— Chantelle Breithaupt, CFO

Assessment: Clean, quantified, and appropriately deflationary about its own good news. Netting out the tariff refund leaves a gross-margin beat of 10-20 bps versus guide, so the margin story remains a below-the-gross-line story. The unused pricing lever stays intact as later-cycle optionality rather than 2026 earnings.

Whether a new 10% customer is finally arriving

A recurring line of questioning returned to the promise of one or two additional 10% customers, now that the shipment constraint management had blamed for the delay has eased. The answer was the least forthcoming of the call.

Q: "Given all the supply chain work that you've done and the improvements on lead times and such, Jayshree, maybe can you revisit your thoughts on adding a 10%? I think in the past, you've talked about 1 or maybe even 2 additional 10% plus customers. And in that, does the 2 Ms remain 10% plus customers as we think about the updated guide for 2026?"
— Aaron Rakers, Wells Fargo

A: "But look, we're going to increase the number by $1 billion or more. And undoubtedly, we remain very committed to our 2 longest partners, Microsoft and Meta. And I fully expect there to be 1, maybe 2 10% customers. I'll leave it at that just because we want all the flexibility on what's in the product sectors and who are the 10% customers as we ship."
— Jayshree Ullal, Chairperson & CEO

Assessment: Third consecutive quarter of the same unfulfilled answer, and the supply excuse no longer covers it. "I'll leave it at that" is a deliberate stop. The concentration bear point neither improved nor worsened on this exchange, but the reason for the delay is now less clear than it was, and the 100-plus AI fabric customer count is the diversification evidence to lean on instead.

Sizing scale-across against the 2030 market

Questioning sought to convert the "approximately 30% of the AI target" framing into a dollar figure and a share position, and to test the credibility of the 2030 market forecast.

Q: "Maybe a quick follow-up actually on the scale-across. So like how large is the overall opportunity today relative to the $1.2 billion, like the 30% of the $3.6 billion that you referred to? And how much do you think you can capture of the $15 billion to $20 billion in 2030?"
— Antoine Chkaiban, New Street Research

A: "The market research is suggesting that it's going to be about a $15 billion, if you add some of the optics, maybe $20 billion in 2030. So I would say in 2026, as we capture $1.2 billion, this market is still maybe $3 billion to $4 billion, but quickly going to $15 billion in the next 4 years."
— Jayshree Ullal, Chairperson & CEO

Assessment: The implied share of roughly a third in a category Arista effectively created is the most concrete competitive-position datapoint disclosed this quarter. The 2030 market figure is a management-cited third-party forecast and carries the usual discount, but the growth arithmetic it implies is the strongest single support for the multi-year durability of the AI thesis.

How far forward visibility extends

The final question of the call went to the foundation under the raised guide: with supply loosening and demand described as exceptional, how much of the year does management actually see? The answer was unchanged from prior quarters and is the most important qualifier on the whole print.

Q: "With the tightness in the overall supply environment and your demand that you're seeing, I'm wondering how far out your visibility with customers is extending?"
— Adrienne Colby, Citi

A: "I think with the -- we still have the same kind of 2 quarters of visibility that we referred to. But what we would say is given where we are at this time of year, Jayshree and I are guiding based on what we're confident we can get the supply for. And if the supply was to release a little more, there is an opportunity to do better in the year."
— Chantelle Breithaupt, CFO

Assessment: The single most important caveat on the call and the one the market paid least attention to. Two quarters of visibility covers Q3 and Q4, so the 2026 guide is well underwritten; 2027 is not. It also confirms the framing shift: the guide is now bounded by what management is confident of sourcing rather than by hard capacity, which is a judgment call rather than a physical limit.

Whether rising complexity favors integrated systems over white box

The opening question inverted the usual competitive framing, asking whether AI networking growing more complex across denser clusters and distributed sites raises the value of an integrated platform and changes the build-versus-buy calculus.

Q: "As the networking problem seems to be getting more complicated, more system level, does that increase the value of an integrated networking platform versus white box approaches? And then are you seeing customers rethink the build versus buy decision at this point when it comes to AI infrastructure, both across hyperscalers and frontier labs?"
— Amit Daryanani, Evercore

A: "So I would say white box is certainly a tactical solution that we tend to see more in use cases that are simple, scale-up or scale-out where the actual amount of software and system requirements are low. But when you look at traditional network topologies, you can slow down the job completion significantly if you go with a box-by-box approach."
— Jayshree Ullal, Chairperson & CEO

Assessment: Management answered the moat question in the customer's own currency, job completion time, rather than in feature-list terms. That is the right argument and it is getting stronger as clusters distribute, because the coordination problem scale-across creates is precisely the kind that a box-by-box approach handles badly. The frontier-lab half of the question went unanswered.

What They're NOT Saying

1. Where the $1.1B raise lands: Neither the AI fabrics nor the campus target moved with the total, and management explicitly declined to allocate the increment. Since mix drives gross margin, the raised operating-margin target implicitly assumes a composition investors cannot verify.
2. Any framing of 2027: Management said the industry supply problem runs to 2028 and that visibility is two quarters, yet offered no growth framework for next year despite $9.7B of multiyear purchase commitments that imply a specific internal forecast.
3. Which customers approach 10%: Third consecutive quarter of "1, maybe 2" without names or timing, and this quarter without the supply constraint to explain the delay. The explicit "I'll leave it at that" was a decision, not an oversight.
4. What share of revenue the 100-plus AI customers represent: The customer count was disclosed as diversification evidence, but not the revenue concentration behind it, which is the number that would actually settle the concentration debate.
5. Why the buyback stayed dormant: Second consecutive quarter of zero repurchases with $817.9M authorized and $13.34B of cash. No rationale was offered; the working-capital and purchase-commitment explanation is our inference.
6. Backlog or book-to-bill: Arista does not disclose either, and the deferred-revenue balance was again characterized as volatile and "independent of underlying business drivers," which removes the one proxy investors had been using.
7. A gross-margin path within the range: The FY range stayed 62-64% and Q3 was guided to ~63%, but management gave no view on where Q4 or the exit rate lands, and pricing power remains explicitly deferred with no trigger date.
8. The tariff refund in dollars: Quantified only in basis points versus guide and versus prior year, never as an absolute figure or as a statement of whether more refunds are expected.

Market Reaction

• Pre-print setup: ANET closed at $190.51 on August 4, up 45.4% year to date, up 58.3% over the trailing twelve months, and up 19.1% in the prior thirty days, at the very top of its 52-week closing range of $116.13 to $190.51. The stock had risen 3.0% in the print session itself. The bar was high again, though the composition was different from Q1: this run was a recovery from the post-Q1 de-rating rather than an extension of an uninterrupted advance.
• After-hours move: Shares rose approximately 13% following the release and call, with pre-market quotes near $212.50 the next morning.
• Reaction session (August 5): The stock gapped up 10.3% to open at $210.19, traded as high as $214.89 (+12.8%), then faded through the session to close at $197.31, up 3.6%. Volume of 20.2M shares was 2.5x the 30-day average. The S&P 500 fell 0.2% on the day, so the move was entirely company-specific.
• Subsequent sessions: The stock closed at $192.32 on August 6 and $188.67 on August 7. That last level is 1.0% below the pre-print close and 12.2% below the reaction-day high.

The reaction is more interesting than the reaction-day number suggests. On the print itself, the market did what the fundamentals warranted: a 13% after-hours move and a 10.3% opening gap for a company that beat by 7% on revenue, 15% on EPS, raised the year by $1.1B, and raised the margin target by roughly 250 bps. What happened next is the analytically useful part. Within one session the gain had two-thirds unwound, and within three the entire post-print move, including the 10.3% opening gap, was gone.

No company-specific news explains the fade. No guidance was revised, and the 10-Q filed August 5 contained nothing that contradicted the release. The most plausible reading is positioning rather than fundamentals: a stock that had already advanced 19.1% in thirty days into the print had substantially discounted a good result, and the opening gap gave holders the exit they had been waiting for. There is a secondary explanation worth weighing, which is that some of the selling reflects the same conclusion we draw in the thesis section: with supply no longer the binding constraint, the guide is exposed to demand in a way it was not three months ago, and investors who owned the stock for a supply-constrained backlog story now own a different one.

For a fundamental owner the practical consequence is straightforward. Estimates went up roughly 14% and the share price ended the week where it started, so the stock is cheaper on forward earnings than it was on August 4 despite the beat. That is the opposite of the Q1 setup, where the multiple compressed because the price fell; here it compresses because the earnings rose.

Street Perspective

Debate: Is the supply fix a durable competitive advantage or a purchased one?

Bull view: The bull case on the Street holds that Arista has converted an industry-wide shortage into a share-gain opportunity, using its balance sheet and vendor relationships to secure memory through 2026 with visibility into 2027, triple its purchase commitments, add contract manufacturers, and stand up a liquid-cooling supply chain, all while smaller competitors remain rationed. The reward is a guide that rose 9.6% in one quarter.

Bear view: The bear camp counters that this advantage was bought rather than built: $9.7B of multiyear purchase commitments, up from $3.6B a year ago, is a large fixed forward obligation against a demand outlook management can only see two quarters into, and inventory and receivables are already consuming cash faster than earnings are producing it.

Our take: Both descriptions are accurate and the bulls have the better of it on a two-year view. A purchased advantage is still an advantage when the purchasing power is not replicable, and Arista's net cash position of $13.34B makes the commitment coverable even in a bad scenario. The legitimate concern is not solvency, it is that the commitments encode a demand forecast that extends well past the two quarters of visibility management admits to.

Debate: Does the operating-margin raise reset the earnings model, or does it fade?

Bull view: A growing consensus view is that the move to a 48-49% full-year operating-margin target on an unchanged 62-64% gross-margin range proves the model has structural operating leverage, since it came entirely from opex efficiency on a business now large enough that flat spending translates directly into margin. Roughly 250 bps on a $12.6B base is material incremental operating income before any revenue upside.

Bear view: Skeptics note that R&D at 9.2% of revenue is the lowest in recent memory for a company that competes on software depth, that management has already flagged incremental investment in liquid cooling and high-Radix switching, and that a single quarter of hiring lagging revenue is not a permanent structural shift. On this view margin normalizes back toward the high forties as spending catches up.

Our take: The bears are directionally right that 9.2% R&D is not a steady state, and we would not extrapolate it. But the guide already accommodates that: 48-49% for the full year against 48.9% delivered in the first half implies management expects roughly flat-to-slightly-lower margin in the back half, not further expansion. The raise is credible precisely because it is not extrapolating the peak quarter.

Debate: Does 46x forward earnings leave anything for shareholders?

Bull view: Some sell-side desks argue that a 40% grower with roughly 49% operating margins, no debt, $13.34B of net cash, and a category-defining position in the fastest-growing part of networking is not expensive at 46x, particularly with 1.6T, scale-up, and unused pricing power all sitting outside the current numbers. Post-print price targets moved sharply and uniformly higher.

Bear view: The bear camp contends the stock has already re-rated from roughly 40x after the Q1 sell-off back to the mid-forties, that the trailing GAAP multiple is near 59x, and that this is a hardware business whose revenue is levered to the capital budgets of a handful of hyperscalers, so any deceleration in AI capex compresses both the estimate and the multiple simultaneously.

Our take: The bears are right that valuation is the live risk and it is why we hold conviction at 8 rather than higher. But the multiple is doing less work than it appears: the price is flat versus the pre-print level while estimates rose roughly 14%, and the wide dispersion in post-print targets suggests the Street has not settled on the post-supply-fix earnings power. We would rather own a 40% grower at 46x with 2027 estimates still forming than wait for a multiple that a business compounding at this rate is unlikely to offer.

Model Update Needed

Valuation impact: Our FY2026 non-GAAP EPS estimate moves to roughly $4.10 from roughly $3.60, an increase of about 14%. The build: $1.89 delivered in the first half, the $1.07 midpoint of the Q3 guide, and an implied Q4 of $1.12 to $1.15 (roughly $3.56B of revenue across the 48-49% operating-margin range, plus other income near the $120M to $130M recent run-rate, taxed at the guided 21.5% on approximately 1.28B diluted shares). At the August 7 close of $188.67 that is about 46x FY2026 earnings, against roughly 40x at the post-Q1 low and roughly 48x entering the Q1 print. We are not publishing a formal price target; the honest framing is that fair value is tracking earnings higher and the swing factor has changed from supply capacity to demand durability. The stock is up 28.3% from the Q1 reaction close of $147.01 and 44.0% year to date, so the easy part of the post-Q1 recovery has been captured.

Thesis Scorecard Post-Earnings

Grading the standing thesis carried from the Q1 2026 update.

Grading last quarter's commitments

Overall: Thesis strengthened. Three of six commitments were met or exceeded and a fourth partially, and the one that mattered most, whether supply would cap how fast demand converts to revenue, resolved decisively in the bulls' favor. Two pillars strengthened within their existing status and the dominant bear point was de-risked from an escalating constraint to a managed one.

One newly established risk. The constraint has moved from supply to demand, and that is a genuine change to the risk profile rather than a restatement of an existing pillar. For three quarters the guide was a shippability floor, which meant a demand wobble would have been absorbed by backlog before it reached revenue. With the ceiling lifted, roughly two quarters of visibility, and a guide management describes as what it is confident of sourcing, a demand disappointment would now pass straight through. We are adding this to the standing thesis as a fourth bear point rather than folding it into the supply pillar, because the two now point in opposite directions.

Action: Maintain Outperform; conviction from 7 to 8. The fundamental improvement is larger than the multiple expansion: FY2026 earnings power rose roughly 14% while the share price is flat against the pre-print level. We would continue to add, with the caveat that the easy recovery from the Q1 de-rating is behind us. The two developments that would move us to Hold are a sequential decline in deferred revenue, unchanged from last quarter, or a Q3 print in which the 43% implied second-half growth rate slips, which would confirm that the demand exposure created by the supply fix is real rather than theoretical.