Arista grew revenue 58.5% and the stock is 21% below its October high; at $170.80 it still pays for my Bull case, so the verdict is hold

Arista's third quarter revenue was $3.66 billion, up 58.5% from a year earlier, and management told investors to expect about $18.5 billion next year, another 38% to 40%. The stock went the other way: from a closing high of $216.74 on October 9 it slid into the results, jumped 5% the day after, fell 10.8% on November 5 in a broad sell-off of AI infrastructure names and closed at $170.80 on November 6. That is 21% off the top, and it sounds like a discount until the numbers go into a model.

My Base value is $129.15, 24.4% below the price. The Bull value is $170.38, almost exactly where the stock trades, and the Bear value is $62.71, 63.3% below. I assign no probabilities to the cases. The verdict is Hold: the business is excellent, the price already assumes my optimistic case, and new money waits for $74.26 or less.

Valuation route: a fast-growing, cash-rich maker of networking switches and software with a few very large customers. The primary method is a ten-year free cash flow model with stock pay counted as a cost; the cross-check is a forward multiple of enterprise value to 2028 operating profit. Both use the same net cash and diluted share count. Values are present fair-value estimates in US dollars, not 12-month price targets.

What Arista is

Arista, based in Santa Clara, California, sells the Ethernet switches and the operating system, EOS, that move data between servers in large data centers. Its biggest buyers are the cloud and AI giants; in 2025 two unnamed customers took 26% and 16% of revenue, 42% together, up from 35% a year earlier. In the 2024 annual report the two were named as Microsoft and Meta. The newer products are Etherlink switches for AI clusters, where Arista now counts more than 100 customers, and campus networking for offices.

The interesting detail is what Arista does not do. It does not make the chips inside its switches and it does not sell the GPUs; it wins because its software lets one operating system run everything from a top-of-rack switch to a 100,000-GPU cluster. That is a real advantage, and it shows in a 64% gross margin. It is also a dependence: the competition includes Cisco, Nvidia's own networking, and white-box switches that the largest customers can design themselves.

The latest results

For 2026 as a whole the guide adds up to about $13.3 billion to $13.4 billion of revenue, roughly 48% growth. In February the company had pointed to about $11.25 billion and it raised the figure every quarter since. For 2027 management guided to about $18.5 billion, with AI networking doubling to about $7 billion and campus near $2 billion, an adjusted gross margin of 62% to 64% and an adjusted operating margin of about 49%.

Two lines on the balance sheet deserve a look. Inventory rose $674.5 million in the quarter to $3.21 billion, and purchase commitments reached $11.4 billion, up from $6.8 billion at the end of 2025, as Arista locked in memory and chip supply. Deferred revenue, $6.84 billion, slipped slightly in the quarter, which was reported as the first decline since 2024. None of that is a problem while orders keep coming; all of it becomes one if they pause.

Which numbers to trust

• Adjusted profit leaves out stock pay. Stock-based compensation was $379.4 million in the first nine months, about 4% of revenue. I count it as a cost, so my operating margin starts at 45.5% in 2027 instead of the guided 49%.
• Free cash flow trails profit. Inventory and receivables grow with sales. I charge 12% of each extra dollar of revenue as working capital and 1% of revenue as net capital spending.
• Interest income is not operating profit. The $14.36 billion of cash earns interest; I add the cash to the value once and leave its income out of the cash flows.
• Tax rises. The rate was 18.5% in the third quarter; management guides about 21% for 2027, and I use 21% throughout.

Cash, shares and capital returns

Arista has no debt and $14.36 billion of cash and marketable securities, about $11.23 a share. I value 1,278.2 million diluted shares, the third quarter count, which already includes in-the-money options and restricted stock. The share count has barely moved in three years: 1,281.1 million in 2024, 1,275.7 million in 2025. Arista bought back $1.6 billion of stock in 2025 and nothing in the first nine months of 2026, with $817.9 million left on its authorization. It pays no dividend.

Primary valuation: ten years of free cash flow

The Base takes the 2027 guide at face value, $18.5 billion of revenue, then slows growth to 22% in 2028, 16% in 2029 and 13% in 2030, fading to 5% by 2036. The operating margin after stock pay starts at 45.5% and drifts down to 42% as the largest customers push for discounts, close to Arista's own history. Tax is 21%. I discount at 10% a year with 3% growth after 2036.

The present value of the rest of 2026 is about $780 million, of 2027 to 2036 $63,195.5 million and of everything after 2036 $85,947.8 million, so the terminal value is 57.3% of the total. Enterprise value is $149,925.1 million. Add $14,359.0 million of cash, that is $128.53 a share.

Even at 9% and 3.5% growth the model stays below today's price. With the 10-year Treasury at 5.13% in October and a stock that moves about 1.6 times as much as the market, 10% is not a demanding rate.

Secondary valuation: a forward multiple

The cross-check applies 16 times to 2028 operating profit of $10,156 million, as a value at the end of 2027, discounted back. That gives an enterprise value of $151,901.4 million and $130.07 a share. Sixteen times is close to what my cash flow model itself implies for the business at the end of 2027 (15.8 times), a company then growing in the mid teens. The market pays 20.1 times the same profit today. I did not build a dated peer comparison for this analysis.

The two methods are 1.2% apart and rest on the same revenue forecast, so the agreement is not independent; it only confirms the arithmetic between them.

History and earnings check

Arista has earned a premium for a decade, and earned it. Revenue went from $7.0 billion in 2024 to about $13.35 billion this year, and free cash flow per share from $2.87 in 2024 to $3.33 in 2025; my Base has $4.57 for 2027. On adjusted earnings of about $4.37 a share for 2026 (nine months plus the fourth quarter guide), the price is 39 times this year's profit and about 28.5 times my rough estimate of $5.99 for 2027. The growth-adjusted price to earnings ratio on 2027 is about 0.8, which looks cheap, but it divides by a single year of 37% growth that comes from a spending boom by a handful of customers; I treat it as a flag, not a value. The free cash flow yield is 2.3% on the last 12 months and 2.7% on my 2027 figure.

Valuation triangulation

• Cash flow model (weight 60%): $128.53.
• Forward multiple (weight 40%): $130.07.
• Base value: $129.15, inside the method range of $128.53 to $130.07.
• The cash flow model gets more weight because it carries the slowdown, the working capital and the tax rate explicitly.

What is the market already pricing in?

The 2027 figure makes the gap concrete. Management guided to $18.5 billion; the price fits $25.3 billion with my margins. Put differently, the market is not paying for the guide but for the guide being beaten by a wide margin, year after year, with no pause in AI data center spending. That is exactly my Bull case, and the Bull comes out at $170.38.

If the market keeps paying 20.1 times forward operating profit, the Base gives about 10.7% a year to the end of 2030 and the Bear -6.4% a year. Most of that Base return would come from earnings growth while the multiple holds, which is the part I would not count on after a 38% to 40% growth year.

Bear, base and bull scenarios

• Bear: the large AI customers pause to digest what they bought; 2027 revenue lands at $16.5 billion, below the guide, and 2028 falls 10%. Price pressure takes the margin to 38% and then 36%. Cash flow model $62.49, multiple $63.04.
• Base: as described above. Cash flow model $128.53, multiple $130.07.
• Bull: Arista beats the guide with $19.5 billion in 2027, grows 28% in 2028 and 20% in 2029, and holds a 44% to 46% margin as Ethernet takes over AI back-end networks. Cash flow model $171.10, multiple $169.31.

Each case blends the two methods 60/40. The discount rate is 10% in all three; growth after 2036 is 2.5%, 3% and 3.5%. The multiples are close to what each case's own cash flows imply at the end of 2027 (11.9, 15.8 and 19.2 times).

How the buy price is set

Confidence is C, with a margin of safety of 42.5%. The reporting is clean and the cash is real; the grade is lower because everything after 2027 depends on how long a few customers keep spending on AI data centers, and my cross-check is not independent of the cash flow model. That gives a maximum buy price of $74.26, a level the stock last traded near in the spring of 2025.

15% more or less revenue in every year moves the Base to $146.56 or $111.73, the largest single effect at 13.5% of the value. A point of discount rate gives $142.14 or $119.33, four points of operating margin $140.31 or $117.98, three turns of multiple $137.69 or $120.60, and working capital at 6% or 18% of extra revenue $130.08 or $128.21. Each extra billion of cash adds $0.78 a share.

For scale, the expected return under my Base is about 15.5% a year at $74.26, 10.0% at $129.15, 8.2% at $170.80 and 7.5% at $200. At today's price the stock sits at my Bull value, which is where I would review trimming rather than adding.

Dividend

Arista pays no dividend. It returns cash through buybacks: $1.6 billion in 2025, none so far in 2026, with $817.9 million authorized. At this price a buyback would retire shares above my value, so I do not count future buybacks as a source of return.

Reasons to own Arista

• Revenue up 58.5% in the third quarter with an adjusted operating margin of 49.1%, among the highest in hardware.
• A 2027 guide of about $18.5 billion, with AI networking doubling to about $7 billion.
• $14.36 billion of net cash, no debt and a share count that has not grown.
• One operating system across data center, AI cluster and campus, which keeps customers on Arista as they scale.
• Management that raised its 2026 revenue outlook every quarter, from about $11.25 billion to about $13.35 billion.

What could go wrong?

• Concentration. Two customers bought 42% of 2025 revenue, and management expects three or four above 10% this year.
• A spending pause. AI data center budgets set the pace; a digestion year would hit revenue and margins together, as in my Bear.
• Competition. Nvidia sells its own networking, Cisco expects about $7.5 billion of AI infrastructure revenue in its fiscal 2027, and the largest customers can build white-box switches.
• Supply bets. $11.4 billion of purchase commitments and $3.21 billion of inventory are fine in a boom and expensive in a pause.
• Margin. Bigger customers get bigger discounts; the adjusted gross margin is already down from 65.2% to 64.0% in a year.

The options market expects movement. At-the-money implied volatility averaged about 56% in October, mid-range for the past year, which suggests a typical one-month move of about 16%, or $27.60, either way. I could not get a reliable read on how much more downside protection costs than upside, so I draw no tail-risk conclusion. Short interest is small at about 1% of shares. None of this changes the value.

Management execution

• The 2026 revenue outlook was raised from about $11.25 billion in February to about $13.35 billion now.
• Operating margins held near 49% while revenue grew almost 50%.
• The balance sheet stayed debt-free and the share count flat despite steady stock pay.
• Since the 2025 annual report the largest customers are no longer named, which makes concentration harder to track.
• Buybacks were heavy in 2025 and stopped in 2026, without a stated reason.

On my ten-point management check (two points each), Arista scores 16 of 20.

Stock Analyza scorecard

The valuation score is 5 of 100.

Earnings quality is not the full 15 because free cash flow trails profit as inventory and receivables grow, and growth runway is not full because it rests on a few customers. On the economic score the company reaches at least 7 of 12, with one criterion unavailable; that is a high-quality growth business, but I cannot certify it as a compounder on the evidence I have. Business quality and valuation are shown separately and are not averaged.

Confidence score

The grade measures how well the model is identified, not the chance of a loss.

What I would watch from here

The full-year report, expected in February 2027, has to show fourth quarter revenue inside the $3.9 billion to $4.0 billion guide, a confirmed 2027 outlook of about $18.5 billion, an adjusted gross margin of at least 62%, and inventory and purchase commitments growing no faster than revenue.

Option prices are worth a glance before the February report: a sharp rise in the cost of downside protection would tell me the market expects a guide cut, though it would not change the value on its own.

The three most important thesis breakers

1. A pause in AI data center spending by two or three customers, which leads toward the Bear.
2. Margin pressure from bigger discounts or competition from Nvidia networking and white-box switches.
3. Revenue beating the 2027 guide by $1 billion or more with margins intact, which leads toward the Bull and would make today's price fair.

Stress test and independent check

I rebuilt the cash flow model, the multiple, the blend, all three scenarios and the reverse valuation in a second, separately written script, and the results match to the cent. That checks the arithmetic, not the forecasts.

A broker quote shows $170.15; the official close was $170.80, which I use. My conservative choices are stock pay as a cost, a margin that fades to 42% and a 21% tax rate; my generous one is taking the 2027 guide in full. The weakest point of the model is everything after 2027, because a handful of customers decide it. The strongest bull argument is a management team that has beaten its own numbers all year. The strongest bear argument is that the price already needs the guide beaten by about a third.

Final verdict: hold at $170.80, do not add

Arista is one of the best hardware businesses in the market: 64% gross margins, almost half of revenue left as operating profit, $14 billion of cash and a product that the largest AI builders keep buying. The 21% fall from the October high took out the froth, not the optimism. At $170.80 you are buying my Bull case at full price.

Verdict: HOLD at $170.80, do not add. Base $129.15, 24.4% below the price, Bear $62.71, Bull $170.38, Confidence C (17.5/25). Required margin of safety 42.5%, maximum buy price $74.26.

New money waits for $74.26 or less, or for evidence that the 2027 guide will be beaten by a wide margin, which would lift the Base. For an existing holder the original cost is irrelevant; the price already sits at my Bull value, so this is a level to review trimming, not to add.

Sources

Primary sources are Arista's third quarter 2026 results (SEC filing, November 3, 2026), the full-year 2025 release, and a filing summary for the 2027 outlook, purchase commitments and customer concentration.

The reference price is $170.80, the November 6, 2026 NYSE close (StockAnalysis). Price moves are from Trefis, the 10-year Treasury yield from FRED, option data from Options Analysis Suite and short interest and beta from StockAnalysis.