[ANET] Arista Networks, Inc.: Scaling AI Through a $9.7 Billion Supply Bet

Arista crossed $3 billion of quarterly revenue in Q2 2026, passed 100 EtherLink AI fabric customers and maintained an objective of at least $3.5 billion of FY2026 AI Fabrics revenue. [3][6] At the same time, purchase commitments reached about $9.7 billion and first-half deferred revenue increased roughly $1.5 billion. [4][5] The opportunity and risk share one source: Arista secured AI-network supply before deployment and customer acceptance convert hardware into revenue.

Company Background and Business Structure

Arista develops high-speed Ethernet switching, routing and network applications on its EOS operating system for cloud operators, AI clusters, enterprise campuses, service providers and specialist AI customers. Systems produce product revenue; installed equipment supports technical support, repair and software-update service revenue.

FY2025 product revenue was $7.5769 billion, or 84.1% of total revenue, and service revenue was $1.4288 billion, or 15.9%. Cloud and AI Titans represented 48% of revenue, Enterprise 32%, and AI and Specialty Providers 20%. [2] AI Fabrics and Campus are management categories rather than GAAP segments.

Financial History and Current Position

FY2025 revenue was $9.0057 billion, gross profit $5.7687 billion, operating income $3.8561 billion and net income $3.5114 billion. [1] Product revenue grew 28.8% and service revenue 27.7%, showing similar expansion in systems and the installed support base. [2]

Q2 2026 revenue was $3.0357 billion, up 37.7%. Non-GAAP gross margin was 63.4%, up 100 basis points sequentially but down 220 basis points year over year. Non-GAAP operating income was about $1.5 billion at a 49.9% margin. [3]

First-half operating cash flow was $2.8 billion. Inventory increased $288.2 million, receivables $379.3 million and deferred revenue about $1.5 billion. [4][5] Purchase commitments were $9.7 billion, versus $3.6 billion a year earlier. [4] Commitments improve supply visibility but add cash exposure if demand, acceptance or product timing changes.

Operating Model

Product revenue equals customer deployments multiplied by ports and system content and realized price. AI customer count becomes revenue only when trials scale, products ship and acceptance occurs. Cloud and AI titan capacity plans are therefore both a growth driver and a concentration risk.

Campus is the second product lane. System shipments generate product revenue, followed by support revenue across the installed base. Campus can reduce hyperscaler concentration if it broadens enterprise exposure, although channel, competition and mix can alter margin.

Supply is the operational constraint. Arista relies on merchant switch silicon, memory and contract manufacturing, and uses multiyear commitments to secure capacity. Those commitments are not customer orders. Components enable shipments when demand arrives; unused components become inventory, prepayment or cancellation exposure.

Acceptance connects supply to reporting. Some large-customer contracts defer revenue until acceptance even after shipment. Acceptance releases revenue and associated cost, while collections, receivables, inventory and deferred cost determine operating cash flow.

Core Debates

Can more than 100 AI customers produce at least $3.5 billion of AI Fabrics revenue?

Arista had more than 100 EtherLink AI customers, a $3.5 billion-plus FY2026 AI Fabrics objective and $3.0357 billion of Q2 total revenue, up 37.7%. [3][6] Customer count demonstrates reach, but cluster size, ports, pricing and acceptance vary widely.

Confirmation requires customer count, objective delivery and total revenue to progress together. The 1.6 Tbps platform is scheduled for H2 2026 trials and 2027 production. [6] More customer logos with a lower revenue objective, or delayed trials and production, would weaken the scale case.

Can Campus become a material second growth engine?

Cloud and AI Titans were 48% of FY2025 revenue, Enterprise 32%, while product and service revenue were $7.5769 billion and $1.4288 billion. [2] Management's FY2026 Campus objective is at least $1.25 billion. [6]

The second-engine test is objective delivery, stable or higher Enterprise mix, and service revenue that follows the installed base. A lower Campus objective, declining Enterprise mix or material margin dilution would leave diversification unproved.

Can multiyear supply commitments avoid inventory and acceptance risk?

Purchase commitments reached $9.7 billion, about 2.7 times the prior-year level. First-half inventory rose $288.2 million, deferred revenue about $1.5 billion, and operating cash flow was $2.8 billion. [4][5] Management said 2026 memory was secured with visibility into 2027. [6]

Q2 non-GAAP gross margin remained within the 62% to 64% full-year range. [3][6] Stable supply, inventory release, deferred-revenue conversion and margin inside that range would validate the commitment strategy. Rising commitments and inventory with slower revenue, or prolonged acceptance delays, would turn supply assurance into balance-sheet risk.

Industry and Competitive Position

Arista uses a consistent EOS software layer and merchant silicon across cloud data centers, AI networking and Campus. It competes with Cisco, Nvidia's InfiniBand and Ethernet offerings, and other switching vendors. Software consistency, automation and hyperscaler operating experience are its primary advantages.

Constraints include large-customer concentration, reliance on critical suppliers such as Broadcom, rapid product transitions and long purchase commitments. Stronger AI demand raises the value of secured supply while increasing the cost of a forecasting error.

Risks and Falsifiers

• More AI customers with a lower $3.5 billion objective or slower total revenue would show weak deployment depth.
• Higher Cloud and AI Titan mix alongside a lower Campus objective would undermine diversification.
• Rising commitments and inventory with two quarters of slower revenue, or margin below 62%, would reveal demand or cost mismatch.
• Deferred revenue accumulating without recognition, or related returns and write-downs, would weaken cash conversion.

What to Watch Next

Connect the 100-plus AI customers to the $3.5 billion objective and 1.6 Tbps product milestones. Test the $1.25 billion Campus objective against Enterprise mix and service growth. For supply, read $9.7 billion of commitments, inventory, deferred revenue, 62% to 64% gross-margin guidance and operating cash flow as one chain.

Conclusion

Arista's model links AI demand, Campus diversification and supply assurance, so they cannot be assessed separately. More than 100 AI customers become economic only through larger accepted deployments; Campus improves mix only through enterprise and service growth; and $9.7 billion of commitments create advantage only when inventory releases, margin holds and cash remains strong. [2][4][6]

Morningstar described Arista as a leading high-speed connectivity supplier across AI and non-AI workloads and linked better supply to the higher outlook. [7] Aardvark noted that the total revenue increase was not allocated among AI, Campus and core, leaving composition unresolved. [8] The next proof is therefore below consolidated revenue: AI objective delivery, independent Campus contribution, and commitments and deferred revenue converting into gross profit and operating cash.