The Merchant-Silicon + Software ModelNarrow moat

Arista Networks (ANET) — moat facet

Buy the best chips, win on systems and software — capital-light by design.

Arista's business model is a deliberate, clever choice: it is fabless and uses merchant silicon — buying the best available networking chips rather than designing its own — and concentrates its engineering on system design and the EOS software that runs on top. This 'buy the best chips, add the best software and systems' approach is a major source of Arista's efficiency and agility. It avoids the enormous cost and risk of designing custom networking silicon (which runs to billions and requires deep chip-design expertise and fab relationships); it lets Arista adopt the newest, best merchant silicon quickly as it becomes available; and it focuses the company's resources on exactly where its advantages lie — software and system engineering — rather than on competing in silicon. It is a major reason Arista is so profitable, so capital-light, and so nimble at the performance frontier.

Where 2025 revenue went (% of revenue)Cost of product33.1%Operating income42.8%Research & development13.7%Sales & marketing5.9%Cost of service2.9%General & admin1.6%Arista Form 10-K FY2025
Buying the chips costs a third of revenue; building nothing else leaves 43% as operating income.

The model works because the value Arista adds is genuinely in the software and systems, not the chips: two vendors using the same Broadcom silicon can produce very different products depending on their system design, software, and integration, and Arista's are among the best. The merchant-silicon approach also aligns with the industry's direction and keeps Arista's costs and roadmap efficient. The flip side is the dependence it creates: Arista relies heavily on Broadcom for its most critical components, Broadcom sells the same silicon to Arista's competitors and to white-box makers and the hyperscalers themselves, and Arista's supply, costs, and roadmap are therefore partly in the hands of a supplier that arms the whole market. The merchant-silicon-plus-software model is a real, capital-efficient strength that plays to Arista's core advantages and drives its profitability and agility; but it rests on a dependence on Broadcom's widely-available silicon, which means the hardware itself confers no exclusive advantage and Arista's edge must come entirely from the software and systems it builds on top — a narrow, skill-based moat rather than a proprietary, structural one — the value-add over the shared Broadcom silicon is all software1.

Moat trajectory: Holding steady

Stable. Buy-the-best-chips-add-the-best-software is capital-efficient and plays to Arista's strengths — but it depends on Broadcom (which arms everyone), so the hardware confers no exclusive edge; the moat lives in software and skill.

The number that tests this moat
Moat Explorer calc
Capital spending as a share of revenue, first half
About 1.5% in H1 2026 ($84.2M)

Renting silicon from Broadcom keeps capital spending near zero; a sharp rise would mean Arista has begun building what it used to buy.

How it's calculated: 84.2 / 5,744.7, purchases of property and equipment over revenue.
Source: Arista Networks Q2 2026 results release (Exhibit 99.1, 4 August 2026) ↗
⚠ Threats to the moat
References
  1. ReportedThe value-add over shared Broadcom silicon is all software.
    Arista Networks Form 10-K, fiscal 2025 — revenue $9.01B (+29%), net income $3.51B, diluted EPS $2.75, gross margin ~64%; customer concentration disclosed (Microsoft ~26%, Meta ~16% of revenue) — FY2025 · publ. February 2026 · source ↗
Sources
Generated September 23, 2026