Microsoft and Meta also write their own switch software and buy plain hardware to run it.
Is it a good business? · 6 · What does its record show?
Sales grew four times and profit per $1 widened
Sales, $B
2.9
4.4
5.9
7.0
9.0
11.9
Dec 2021Dec 2022Dec 2023Dec 2024Dec 2025last 12 mo
Sales$2.95B → $11.9Brevenue, 2021 to the past year
Profit31% → 44%operating profit per $1
Per $1Near 64%kept per $1 after making it
Research, selling and share pay all fell per sales dollar while research spending still rose.
Is it a good business? · 7 · How much room is there to grow?
A market of tens of billions within reach
$100Ba year on AI switches by 2030, per Dell'Oro
About $67Bthe part Arista sells into
$10.54BArista's sales over four quarters
The market is large enough; the question is the share Arista keeps.
Is it a good business? · 8 · Who is in charge?
A long-serving chief, little insider ownership
Since 2008Jayshree Ullal has run Arista
Under 3%of the shares held by insiders
521insider sales in a year, and no open-market buying
Co-founder Ken Duda runs the technology; the operations and finance chiefs are recent hires.
The story right now · 9 · Why do AI clusters need so many switches?
AI chips must constantly talk to each other
June 2026AI switch spending passed ordinary networks
58.5%Arista's sales growth last quarter
3rdArista's rank in AI switching
Training an AI model splits work across thousands of chips that must share results constantly. Every link runs through switches, so AI clusters need far more of them.
The story right now · 10 · Why is fast growth not enough?
The price pays for a leader; the count says third
AI switch spending soarsArista grows fast.
Rivals grow fasterCelestica, NVIDIA.
Arista ranks thirdPer Dell'Oro.
The price assumes firstYears of leadership.
Whether third place means lost ground or deliveries awaiting acceptance is not yet known.
The story right now · 11 · What changed lately?
Sales beat the forecast and the next quarter is higher
$3.66Bsales, up 58.5%
44.9%operating profit per $1
$6.84Bbilled but not yet earned, a small dip
38% to 40%2027 sales growth forecast
Solid quarter
Arista forecast 2027 sales of about $18.5B and called the dip timing.
The story right now · 12 · What is everyone arguing about?
Leader of AI networking, or third?
Many investors think
Arista is the pure winner of AI networking
We think
The growth is real, the leadership is not shown
Against us: sales outgrew every past doubt, and only the worry about profit per $1 came true.
The price and our view · 13 · What is a share worth?
Even on steep forecasts the price is a fifth above our value
$157what we think a share is worth · the last close is of that
$157our value
Attractive below $125
Neutral
Unattractive above $188
$70.46gloomy: growth halves and less kept per $1
$157our estimate: sales more than doubling by 2029
$230sunny: growth beats forecasts by a quarter
Value per share in three futures. None is a prediction of the share price.
We estimate what a share is worth from the cash the business should earn over many years. Our rule: buy only when the price is at least a fifth below our estimate, as a cushion for being wrong.
The price and our view · 14 · How could we be wrong?
Third place, and two customers carrying sales
Third, not firstBiggest
Dell'Oro ranks it behind two.
Two customers, 42% of salesShowing now
Bigger discounts as they grow.
Insiders only sellWatch
No open-market buying.
Third place's causeUnknown
Lost ground or awaiting acceptance.
Broadcom supplies the chips with no written guarantee.
The price and our view · 15 · How will we know?
February shows whether the dip was timing
Sales of $3.9B to $4.0BGood sign
Billed-but-unearned sales rising againGood sign
Another fall in billed-but-unearned salesBad sign
FEB 202717
Fourth-quarter 2026 results, 17 February. Companies report every three months; each report is a test of the story.
Our view also changes on price alone: a rise to more than a fifth above our estimate moves us to Unattractive, a fall to a fifth below it to Attractive.
The price and our view · 16 · So, what do we think?
A fast grower priced just inside our cushion
$157our value
Attractive below $125
Neutral
Unattractive above $188
NeutralMedium conviction
Arista's switches and software run AI data centres and sales are growing fast. But the price assumes it keeps leading, while an independent count ranks it third and two customers carry a large share. Fair, not cheap: Neutral.
↑ ↓ to move · C for contents
What it does
More about the company
Who it is
The chips in an AI data centre swap results with each other constantly, through switches: boxes with dozens to hundreds of high-speed ports. Arista sells those boxes, and the same software, EOS, runs on every one, from an office cupboard to an AI cluster.
It designs none of the chips and builds none of the boxes. Broadcom supplies the silicon, with no written guarantee, and contractors assemble them.
How it makes money
Arista is paid when it ships a switch, and again for multi-year support. Much of it must first pass the customer's acceptance test, so cash often arrives before the sale counts: $6.87B had been billed and not yet earned at the end of June. Two customers bought 42% of everything it sold in 2025.
Growth and profit
What growth costs
Growth has made Arista more profitable: research, selling and running costs rise more slowly than sales, so each new dollar of sales costs less to win. The squeeze is elsewhere, because selling more to the biggest buyers lowers what each box earns, so the gains show up below the gross margin, not in it.
Its record
The numbers
Is it growing?Fast
Sales grew 41% over the twelve months to Sep 2026, faster than the 29% in the year to Dec 2025.
Is it profitable?Strong
Operating margin is 44%, up from 39% in the year to Dec 2023.
Does growth pay off?Pays off
Running costs take 19¢ of each sales dollar, down from 32¢ in the year to Dec 2021.
Does it earn good returns?Strong
Each dollar put into the business earns 30% a year after tax, up from 23% in the year to Dec 2021.
Is it making cash?Strong
The business brought in $5.24B of cash from operations, 44% of sales, before capital spending, which the filer does not tag.
How steady is it?Uneven
Yearly sales growth has ranged from +20% to +49% over the last 5 years.
Are shares being diluted?Stable
The share count rose 1.0% in 3.8 years, and pay in shares costs 4.3% of sales.
Why it wins
The evidence
A customer does not adopt a switch; it adopts a way of running its network. Its automation, monitoring and staff training are written once for EOS, so leaving means rewriting them and retesting new hardware.
It solves a real problem. Training one AI model, Meta counted 35 of 419 unplanned stoppages in 54 days on switches and cables; one fault can restart the whole job.
The lock-in stops at the two biggest customers, who wrote their own switch software a decade ago and still buy Arista hardware to run it.
Who could beat it
The rivals in full
Dell'Oro, an independent research firm, ranked Arista third in switches for AI networks in both of the last two quarters. First was Celestica, which sells unbranded boxes to the biggest buyers, including Meta. Second was NVIDIA, which sells its network with its chips, so the buyer never holds a switch contest. Cisco, fourth, gained the most ground.
The whole switch. A hyperscaler buys a fully designed white box from Celestica and runs its own or an open-source network operating system on it, so the buyer stops paying for EOS and does the integration itself.
The back-end AI fabric, bought as part of the accelerator purchase rather than as a separate networking decision - so the buyer never runs a switch evaluation at all.
The same Ethernet switch sold to the same hyperscaler, bundled with security and observability the buyer would otherwise source separately.
Hewlett Packard Enterprise, with the Juniper portfolioHPE
The campus, routing and data-centre switch lines Arista sells as its adjacencies, now from one vendor across all three after the Juniper acquisition Arista's own 10-K records.
EOS itself, on Arista's own hardware. The customer runs its own operating system and buys the box, so the software Arista's moat rests on is not part of the sale.
Meta, with FBOSS and its own OCP switch designsMETA
Both the operating system and, on the Wedge line, the box - those are designed by Meta and manufactured by Celestica, not bought from Arista at all.
How big it could get
The market and who has signed
The market
Dell'Oro forecasts spending on switches for AI networks to pass $100B a year by 2030, and about two-thirds of AI switch sales today are the Ethernet kind Arista sells. That is a market of about $67B, against Arista's $10.54B of sales over the last four quarters.
Who has signed
Customers have committed $8.4B of future sales, 91% of it due within two years. But only $875M is described as binding. Most of the rest is the $6.87B already billed and waiting for customers to accept the equipment. Beyond that, nothing is ordered.
Who runs it
Leaders, pay and the cash
Who runs it
Jayshree Ullal has run Arista since 2008 and took it to $9 billion of sales; co-founder Ken Duda runs the technology, and two recent hires from Fastly and Aspen Technology run operations and finance. Insiders own under 3%, mostly Ullal's; a family trust outside management holds about 15%. Bonuses and performance shares pay on sales and operating profit for the year, none on the share price or on cash, and last year paid the maximum.
What they did with the money
It holds $13.3B of cash and securities, no debt, and bought back no shares this year. Instead it has committed $9.7B to buying parts, nearly triple a year earlier, to secure supply in a market short of it.
The last report
The latest report
Arista's sales rose 58.5% to $3.66B and the next quarter is guided higher
What it did to the thesis Solid quarter
It confirmed the growth and kept margins inside our range, and the call explained the dip in billed-but-unearned sales. It answered the quarter's test without settling whether third place in AI switching is timing or lost ground.
The report
The third quarter of 2026 was a record: sales of $3.66B, up 58.5% on a year earlier and 11% above the $3.3B Arista had guided, with gross margin of 63.6% and operating margin of 44.9%. Arista guided the next quarter to $3.9B to $4.0B.
The one soft spot was billed-but-unearned sales, down 0.4% to $6.84B. On the call Arista called it timing and expects the year-end balance above last year's, and guided 2027 up 38% to 40%, about $18.5B.
What's happening now
The background
The share price does not make it urgent: Arista returned 44.4% over the past year, against 149.8% for the median company in its AI networking theme.
The test came on 3 November: the September quarter beat the guide and the next is guided higher, while the billed-but-unearned balance dipped. Underneath, spending on switches for AI networks passed the ordinary data-centre network for the first time in June, and both Arista and Dell'Oro say supply, not demand, limits it.
Where we disagree
Both sides in full
The market
Arista is the pure-play winner of AI networking, and the price is what that leadership costs.
We believe
The growth is real and limited only by supply. The leadership the price pays for is not what the independent count shows.
The argument
Most analysts rate the shares a buy and none a sell. But two public facts sit outside that story: Dell'Oro's ranking, and the filing's own reason the margin is falling.
The market has doubted before, selling three strong reports in fifteen months over concentration, margin and supply. Sales outgrew all three; only the margin worry came true.
The evidence against us is real. Dell'Oro says Arista would rank higher counting its billed-but-unearned sales, and gross margin holding at 64% while sales grow above 30% would bound the discounts.
What it's worth
How we value it
Valued on the analysts' revenue path for three years, each later year closing 30% of the gap to 4%, on official operating profit with stock pay a cost and tax paid in full; the thesis's own evidence is the check on that path.
What the price assumes
Sales growth in the first year, then slowing as ours does, at the same profit per sale.
The price assumed
47.4%
Our estimate
41%
Measured at the close of
16 Nov 2026
The three cases
Bear case$70.46
Growth runs at half the analysts' path, 20%, 14% and 10%, the margin settles a fifth lower at 33.4%, and new capital earns only the 8% rate: the downturn test.
Our value$157
Revenue grows 41%, 28% and 21% over three years, the analysts' path (C74), then fades toward 4%, closing 30% of the gap each year; the margin moves to 41.7%, the analysts' year-three figure after stock pay.
Bull case$230
Growth runs a quarter above the analysts' path, 51%, 35% and 26%, the margin reaches 45.9%, and new capital earns 15%.
What could go wrong
Every risk
The risks
Third, not first
Dell'Oro ranks Arista behind Celestica and NVIDIA in AI switching for two quarters running, with Cisco gaining the most.
Two customers, 42% of sales
One was 26% of 2025 sales, up from 18% in 2023. The chief executive expects only one, maybe two, above 10% this year.
Insiders only sell
Insiders sold 521 times in the past year, the chief executive and the co-founder among them, and never once bought on the open market.
What we still do not know
Whether third place is lost ground or equipment still awaiting acceptance. Nobody publishes the figure that would settle it.
How much of the $6.84B billed but not yet earned is AI equipment, and whether September's 0.4% dip is timing. The company says only that most of it is product.
What the AI business earns. Arista reports as one segment, so no quarterly AI figure is filed.
What changes our view
Every route to a new view
Every route to a new view
→ Attractive
The share price falls to about $125, 80% of what we think the business is worth.
→ Unattractive
The share price rises to about $188, 20% above what we think the business is worth.
→ Unattractive
Either of the two largest customers falls below 10% of sales because it bought less. Dell'Oro ranks Arista fourth or lower, or gross margin stays below 62% for two quarters running on customer mix.
The next test
It supports the case if
Revenue at or above $3.9 billion with gross margin at or above 62.5% and deferred revenue back above the $6.84 billion of September - the September dip was timing, supply is releasing, and the acceptance pipeline is still filling.
It pressures the case if
Revenue below $3.9 billion, or gross margin below 62.5%, or deferred revenue below the September balance for a second quarter - the first says supply did not release, the second that the discount to the largest customers is widening, the third that the acceptance pipeline has stopped filling.
We would also be wrong if
Billed-but-unearned sales fall for two quarters running while growth slows.
Sales growth settles below 20% while operating margin falls under 45%.
The verdict
Scores and how we weighed it
The scores
Growth
3 of 4
Business quality
3 of 4
Financial position
4 of 4
Execution record
4 of 4
Per share capture
3 of 4
Potential
3 of 4
Risk profile
2 of 4
Valuation
3 of 4
Evidence quality
3 of 4
How we weighed it
ValueNegative
The price has fallen back just inside the line a fifth above our value, with no company news; it still sits above a value that already takes the analysts' steep path, and the bear case sits far below.
BusinessPositive
Network switches for AI data centres on one software system, with a strong balance sheet and record of delivering.
GrowthNegative
Growth is fast, but the price now assumes more of it than even the analysts' steep path.
RisksNegative
Two customers are a large share, an outside count ranks it third in AI switching and insiders only cash out.
RivalsNegative
The dearest of its filed rivals on operating profit, by a wide margin.
The groupMixed
Server and network builders move together on AI spending news.
Latest newsPositive
The last report confirmed the growth and kept the margin inside our range.
In plain words
Our estimate of value
We forecast how much cash the business can earn over the next ten years and beyond, then ask what that stream is worth today to an investor who wants an 8% yearly return. We do it three times: a gloomy future, an expected one and a sunny one. We call a share Attractive only when its price is at least 20% below the expected estimate, and Unattractive when it is 20% above.