Catch up on the essentials
  • Arista Networks reported third-quarter 2026 revenue of $3.66 billion, up 58.5% year over year, and raised its full-year revenue outlook for the fourth time in 2026 to approximately $13.3 billion.
  • Inventory increased to $3.2 billion from $2.5 billion in the preceding quarter, while purchase commitments rose to $11.4 billion from $9.7 billion.
  • Scale-out has represented the largest portion of demand, while scale-across generated roughly 30% of the company’s AI-related mix in 2026 and could represent a larger share in 2027.

Selected from this article · 2026-11-05

Read on for the full picture

Arista’s growth is increasingly tied to AI networking

Focused detail of a modern server rack with blue LED indicators in a data center.

Arista Networks reported third-quarter 2026 revenue of $3.66 billion, up 58.5% year over year, and raised its full-year revenue outlook for the fourth time in 2026 to approximately $13.3 billion. The quarter exceeded management’s prior $3.3 billion guidance, which CFO Chantelle Breithaupt attributed to continued demand from AI and enterprise customers.

The scale of the forecast suggests AI networking has become a central growth engine rather than a supplementary product line. Management expects revenue growth of 38% to 40% in 2027, or about $18.5 billion at the midpoint, with AI fabrics forecast to double to approximately $7 billion. CEO Jayshree Ullal said the company had surpassed 100 Etherlink customers and expects deployments exceeding 10 gigawatts across more than 10 types of compute accelerators by the end of 2027.

These are company expectations, not evidence of completed 2027 deployments or bookings. The reviewed reporting also does not provide an open sourcing process, named supplier allocation or customer-level order schedule.

Supply commitments rise with deployment expectations

Inventory increased to $3.2 billion from $2.5 billion in the preceding quarter, while purchase commitments rose to $11.4 billion from $9.7 billion. Breithaupt said Arista is investing ahead of customer demand and securing components for new products and AI deployments.

The commitments include chip-related spending and a multiyear memory agreement. This supports a direct link between Arista’s deployment outlook and upstream component requirements, but it does not identify individual memory or chip suppliers, purchase volumes or confirmed orders for particular customers. Capital expenditures were $100.7 million, while inventory and purchase commitments also increased during the quarter.

Profitability remains high despite margin pressure

Arista generated $1.8 billion in operating income, equal to 49.1% of revenue, and $1.53 billion in net income, or 41.9% of revenue. Diluted earnings per share were $1.20, up 55.8% from $0.77 in the prior-year period.

Gross margin was 64%, below 65.2% a year earlier but above 63.4% in the prior quarter. The sequential improvement suggests near-term profitability remained intact despite a lower year-over-year comparison, while the year-over-year decline leaves the durability of margins under heavier AI deployment demand an open question.

Mix is broadening beyond the largest AI use case

Management divides its AI-fabric activity into scale-out, scale-across and scale-up networking. Scale-out has represented the largest portion of demand, while scale-across generated roughly 30% of the company’s AI-related mix in 2026 and could represent a larger share in 2027. Scale-up deployments are expected to remain in trials during 2027, with production anticipated in 2028.

This is a product-development timetable rather than a procurement timeline. The key industry implication is that Arista’s growth opportunity spans several networking architectures, although each has a different maturity: scale-out is already the largest demand contributor, scale-across is expanding within the current mix, and scale-up remains at the trial stage.

The next test is whether rising component commitments convert into the forecast AI deployments without further margin dilution. Observable indicators include quarterly AI-fabric revenue, deployment volumes, inventory and purchase-commitment trends, and updates on scale-up trials.

Industry and product implications

Industry / product

Disclosed fact

Demand interpretation

Verification signal

Main risk

AI fabrics (downstream)

Management forecast that Arista's AI fabrics business would double to approximately $7 billion in 2027.

[1]

Qualified inference; planned. Scale-out has represented the largest portion of demand, while scale-across represented roughly 30% of the AI-related mix in 2026.

Whether 2027 AI-fabric revenue and deployment disclosures support the approximately $7 billion forecast.

The forecast is exposed to AI deployment demand; scale-up remains in trials.

Scale-up networking (enabling)

Arista expected scale-up deployments to remain in trial stages during 2027, with production anticipated in 2028.

[1]

Research watchlist; planned. Scale-up is one of three AI-fabric categories supported by Arista.

Look for disclosed production timing and customer deployment evidence beyond trials.

No production scale-up revenue or customer procurement is established in the reviewed reporting.

Etherlink (downstream)

Ullal said Arista had surpassed 100 Etherlink customers as demand increased for networking infrastructure supporting AI training and inference.

[1]

Disclosed fact; operating. Etherlink supports AI training and inference networking workloads.

Monitor future customer-count disclosures and associated deployment volumes.

Customer count does not disclose order value, deployment size or customer concentration.

Chip-related spending and multiyear memory agreement (upstream)

Arista's $11.4 billion in purchase commitments included chip-related spending and a multiyear memory agreement.

[1]

Disclosed fact; unknown. Management said inventory and purchase commitments rose as Arista secured components for new products and AI deployments.

Monitor inventory, purchase commitments and supplier-capacity disclosures.

The reviewed reporting did not identify individual suppliers, component volumes or customer-specific orders.

Share this article

FacebookX

4 sources

Sources

Table sources