Arista Networks (ANET) Blitzscaling Score: 73/100 (B) — Moat, Speed & Risk

Arista Networks scores 73/B on the BlitzLab 100-point model, eight points above the 65-point success line. It is the only company in this batch whose growth is accelerating at the top of the scale: third-quarter 2026 revenue grew 58.5% year on year, the fastest since its IPO, on the networking inside AI data centres. It is a hardware company with software margins, no debt, and no equity raised since its 2014 IPO. Its risk is the mirror of its strength — two customers are about 42% of revenue, and its largest ecosystem partner is now its direct competitor.

The business

Arista builds the networking that connects servers inside data centres — switches running a single operating system, EOS, across the whole product line — and sells it to the largest buyers on earth: cloud providers and hyperscalers building AI clusters. The growth record is the score’s foundation. Twelve consecutive quarters of year-on-year growth, from the third quarter of 2023 through the second quarter of 2026, at rates from 15.9% to 37.7%, and the chief executive declared a nineteenth consecutive record quarter in the third quarter of 2025. The most recent four quarters before the rescore accelerated in sequence: 27.5%, 28.9%, 35.1%, 37.7%. The second quarter of 2026 was the first $3 billion quarter, at $3.036 billion, and fiscal 2025 revenue of $9.006 billion grew 28.6%. Trailing twelve-month revenue stands at $10.541 billion, and fiscal 2026 guidance of about $12.6 billion — roughly 40% growth — was raised three times during the year.

The third quarter of 2026, reported November 3, 2026, broke the pattern upward: revenue of $3.66 billion, up 58.5% year on year, the highest growth rate since the IPO and above the company’s own guidance of about $3.3 billion. Guidance for the fourth quarter of $3.9–4.0 billion implies quarter-on-quarter growth of 7–9%, a normalisation the score file records explicitly: the full mark for growth is a judgement on this quarter, not a promise about the next one.

The gates

BlitzLab scores only companies that pass two gates. G1 asks whether growth continues or a second growth curve exists. Arista passes on both: the twelve-quarter growth streak, and two second curves stated as numeric targets — AI-centre networking from $1.5 billion in 2025 to a 2026 target of $3.5–3.6 billion, more than doubling, and campus networking from more than $0.8 billion in 2025 to a 2026 target of $1.25 billion, reinforced by the $300 million VeloCloud acquisition for software-defined wide-area networking.

G2 asks whether the ten-year ceiling of the new market is at least ten times current revenue, a bar of about $105.4 billion against trailing revenue of $10.541 billion. Arista’s G2 is the only conditional pass in this batch, and the score file says so plainly. The chief executive’s stated total addressable market of “$100 billion-plus” is 9.5 times trailing revenue; data-centre networking of $103 billion by 2030 is 9.8 times. Either estimate alone falls just short of the bar. The pass rests on the breadth of the company’s market frame — data centre, campus, routing and AI centre together — and on Crehan’s estimate of a five-year cumulative $250 billion Ethernet switch market: on a ten-year view the ceiling has room to exceed the bar. The conditionality is the honest part of the verdict, and it is why market size scores 10 rather than 12.

Why 73: speed

Speed scores 20 of 25, and growth, A1, earns the full 10 of 10 — the only full mark on growth in this batch. The acceleration at scale is the reason: five consecutive quarters of quickening growth ending at 58.5%, guidance raised three times, and a $10 billion-plus trailing base growing at that rate. The score file’s own caveat is that the fourth-quarter guidance implies normalisation, so the 10 is a reading of this quarter’s evidence.

Capital velocity, A2, earns 4 of 5. Arista has raised no equity since its 2014 IPO, carries no borrowings, and returned capital instead: $1.6 billion of buybacks in fiscal 2025, plus an additional $1.5 billion authorisation approved in May 2025, with $787 million executed in the first quarter of 2025 — a record. The market capitalisation of about $238 billion in August 2026, rising to about $273 billion by October 10, puts capital access at the top tier. The point held back is narrow: the twelve-quarter market capitalisation trend was unconfirmed at scoring. Hiring and productivity, A3, earn 3 of 5: about 5,115 employees at the end of fiscal 2025, producing revenue per employee of $1.76 million — high productivity for a hardware company — but the historical headcount series was unconfirmed, so the conservative principle caps the mark. Share velocity, A4, earns 3 of 5: IDC puts Arista’s data-centre Ethernet switch share at 19.2% in the third quarter of 2025, on revenue of $1.9 billion up 29.1% year on year, against Cisco at 29.8% in first place. The discount is competitive: Nvidia has penetrated data-centre Ethernet to 11.6%, up 167.7% year on year, so Arista’s share is solid but the competitive picture is worsening.

Why 73: moat

Moat scores 42 of 60 — the lowest moat axis in this batch, and the reason a company with a perfect growth score lands at 73 rather than higher. Network effects, B1, earn 12 of 18: the moat is not a social-graph network effect but switching cost built from a software stack — one operating system, EOS, across all products, more than 150 million cumulative ports shipped, the CloudVision, NetDL and AVA stack, and a net promoter score of 89, with 94% strongly positive. Large customers have made what the score file quotes as “decisive standardizing” declarations, and joint engineering relationships run ten to fifteen years. Gross margin, B4, earns 11 of 12: GAAP gross margin held a 61.9–65.2% band for twelve quarters, 64.1% in fiscal 2025, with an operating margin of 42.8% — up from 38.5% in fiscal 2023 — and 45.4% in the second quarter of 2026, plus a net margin of 39.0%. For a hardware company that is top-tier; it stops short of software’s 70%-plus margins. Market size, B3, earns 10 of 12 on the conditional G2 logic, with the full mark reserved for the moment AI-centre revenue proves the $3.5 billion target in reported results.

Distribution scalability, B2, earns 9 of 18 — the lowest single moat item in this batch, and the score’s most deliberate judgement. The efficiency is extreme: sales and marketing of 3.6% of revenue, $109.8 million in the second quarter of 2026. But the distribution runs through two customers: Microsoft and Meta together account for about 42% of 2025 revenue, roughly 26% and 16% respectively, and the larger the customer the larger the price discount, diluting margin — a trade-off the 10-Q states explicitly. There is little viral element. The score places Arista at the top of the 5–9 band: distribution is extraordinarily efficient and extraordinarily concentrated at the same time.

Why 73: creative destruction and demerits

Creative destruction scores 11 of 15. C1, the order being displaced, earns 4 of 5: Arista is at the front of Ethernet displacing InfiniBand in the AI back-end fabric, with EtherLink AI, the 7800 spine and 1.6T products. The point held back is that Nvidia’s Spectrum-X is penetrating from inside the same Ethernet camp, so the benefit of the displacement is shared rather than captured. C2, irreversible behaviour change, earns 4 of 5: hyperscaler AI clusters are standardising on Ethernet, and deferred revenue rose from $6.2 billion in the first quarter of 2026 to $6.9 billion in the second — $5.1 billion short-term plus $1.8 billion long-term — a signal of long-dated order structure. It is still in progress, which is why the fifth point waits. C3, cost-structure change, earns 3 of 5: the 1.6T AI fabric launched, and the XPO multi-source agreement claims rack reductions of up to 75% and footprint reductions of up to 44% — but both are company claims without independent measurement, so the conservative principle caps the mark.

Demerits total zero — the cleanest demerit sheet in this batch. D1, debt, is 0: no borrowings, cash of $13.3 billion against purchase commitments of $9.7 billion. The commitments are 92% of trailing revenue, large in absolute terms, but within the normal range for securing hardware supply. D2, revenue quality, is 0: no restatement history was confirmed, though the auditor and audit opinion were not directly checked against the full 10-K, which the score file records as an open item. D3, governance, is 0: a single class of common stock, one share one vote, with the Bechtolsheim Family Trust at 14.6% holding the same one-vote shares; insider selling is entirely under pre-arranged 10b5-1 plans. D4, regulatory and social cost, is 0: no material litigation was confirmed in disclosures, tariffs are a 10-Q risk factor but a second-quarter 2026 tariff refund restored 100 basis points of gross margin quarter on quarter, and short interest sits at about 1.0% after falling 35.6% in July.

The blitzscaling lens

Through the blitzscaling lens, Arista is a platform-plus-performance blitzer. The platform is EOS: one operating system, one data lake, one management plane, and standardisation is the blitz itself — the product is not the switch, it is the standard the customer adopts. The performance lever is the 800G-to-1.6T speed leadership that makes hyperscalers choose Arista first. The weak lever is classic network effects: customers getting more numerous does not directly make the product better for other customers, and the closest analogue is the indirect network effect of standardisation — the more the industry standardises, the higher the cost of leaving.

On the speed-versus-efficiency trade-off, Arista breaks the blitzscaling stereotype. The textbook says efficiency comes later; Arista grew 40% while holding an operating margin of 42.8%, thanks to the 3.6% sales-and-marketing ratio and direct sales to large customers. Its version of “efficiency later” is supply-chain pre-investment: purchase commitments rising from $8.9 billion to $9.7 billion and inventory of $2.38 billion, securing components before demand is fully confirmed, with $13.3 billion of cash as the buffer. And it buys speed with margin: large customers get price discounts in exchange for share, the trade-off the 10-Q names directly. The stage is the middle of the first blitz — the opposite of a company between blitzes. The core data-centre business is in high-speed growth with a 19.2% share chasing Cisco, and the AI-centre and campus second blitzes are at the target stage. The current 73/B is a mid-blitz score, and if the second curves prove out in reported revenue, the B3 full mark is the named upside.

The single largest risk is the one the score file states as a sentence: the biggest customer becomes the biggest competitor. Nvidia is Arista’s GPU-ecosystem partner and, through Spectrum-X, a direct rival that has taken 11.6% of data-centre Ethernet. With Microsoft and Meta at 42% of revenue, if those customers begin choosing Nvidia fabric instead of Arista, the fuel of the blitz — large-customer standardisation — turns into the thing that ends it. The source of the speed and the end of the speed are in the same place.

What would change the score

Upward, the levers are named. AI-centre revenue reaching the $3.5–3.6 billion target in reported results would lift B3 toward its full mark and convert the conditional G2 pass into a clean one. Campus revenue reaching $1.25 billion would do the same for the second curve. A data-centre Ethernet share that rises toward Cisco’s 29.8% while Nvidia’s penetration stalls would lift A4. Downward, the triggers are concentration and margin: Microsoft’s share rising further from 26%, a large customer shifting volume to Spectrum-X, or the large-customer mix pushing gross margin down toward the first quarter of 2026’s 61.9% trough, when tariffs compounded the discount effect. The valuation note is separate from the score, as the model requires: a market capitalisation of about $238 billion in August 2026, about $273 billion by October 10, and a trailing price-to-sales ratio of about 23 times — the business score does not look at price.

---

Score table

Sources

• BlitzLab scoring file scores/2026Q4/ANET_2026Q4.md (T = 2026-11-03; Q3 2026 rescore; model v1.0 frozen 2026-10-10) and datapack datapacks/2026Q4/ANET_2026Q4_datapack.md. Figures from company press releases, 10-K/10-Q filings as cited second-hand in the score file (SEC EDGAR JSON API was blocked with 403 in the collection environment; items not directly checked against primary text are labelled in the score file), IDC share data, and earnings calls. No investment advice, price target, or buy/sell/hold opinion is made in this article.