Arista Networks, Inc. (ANET) Business Profile

What it does

Arista Networks is a Santa Clara-based networking company founded in 2004. It sells the equipment and software that move data between computing systems in AI centers, cloud and enterprise data centers, offices and wide-area networks.

Its platform combines Ethernet switches and routing platforms with EOS, its network operating system; NetDL, its network data lake; and CloudVision, its management platform. In plain terms, the hardware connects servers, accelerators, storage and sites, while the software helps customers configure, monitor, automate and troubleshoot those networks. The company’s Core offerings cover AI, cloud and data-center networking; Cognitive Adjacencies cover campus networks and routing; and Cognitive Networks cover software and services.

For AI systems, Arista’s Etherlink portfolio addresses scale-out connections between accelerators in multiple racks and scale-across connections between geographically distributed clusters. Its campus portfolio includes wired switches, Wi-Fi access points and the VeloCloud SD-WAN portfolio. Its routing platforms address multi-cloud connections, data-center interconnects, VPNs, core routing and enterprise-edge routing.

Customers include cloud and AI companies, service providers, financial-services organizations, government agencies and enterprises in sectors including healthcare, education, manufacturing, media and oil and gas. The company provides support through Arista A-Care Services, including product replacement options, software fixes, releases and case management.

How it makes money

Arista earns product revenue mainly from switching and routing platforms, related network applications, EOS software and accessories such as cables and optics. It earns service revenue primarily from post-contract support contracts, generally sold with products and renewed later. Hardware and post-contract support are usually sold in one contract, but are separate performance obligations when the customer receives distinct products and services.

The company sells directly and through distributors, value-added resellers, systems integrators, OEM partners and technology partners. Channel partners normally receive an order from an end customer before ordering from Arista, and generally do not hold Arista inventory. Arista uses direct sales teams and technical systems engineers for major accounts, territories and verticals.

Pricing reflects the product category, expected volume, customer vertical and size, discounting policies, market conditions and gross-margin objectives. Large-scale orders commonly receive discounts, which can reduce gross margin in the periods when those sales occur. Revenue timing can also be affected by product trials and contracts with acceptance terms: products may be invoiced only when trials finish and customers accept them.

FY2025 product revenue was $7.6 billion; Q2 2026 total revenue was just over $3 billion. Service revenue increased with initial and renewal support contracts as the installed base expanded. In Q2 2026, management said its deferred revenue balance was approximately $6.9 billion, with the majority product related; it also said acceptance clauses and new-product deployments can make the balance move substantially between quarters.

Customers and how it sells

Arista groups its customers into Cloud and AI Titans, AI and Specialty Providers, and Enterprise customers. Its enterprise base includes financial services, government agencies, media and entertainment, healthcare, oil and gas, education, manufacturing and industrial customers. It sells globally through direct sales and channel partners, with sales teams organized by geography, major end-customer accounts and vertical markets.

Customer concentration is material. Reseller B represented 26.0% of FY2025 revenue and Reseller A represented 16.0%. The company expects the timing and concentration of sales to vary quarterly and annually because large orders are complex, customer spending and deployment cycles differ, and products can require long evaluation, testing and qualification periods. Management said Microsoft and Meta are its two longest-standing partners, and expects one or two additional customers to account for more than 10% of sales.

International revenue was $697.8 million in Q2 2026, or 23% of total revenue. Arista’s products are shipped through direct fulfillment facilities in the United States, the Netherlands and Singapore, while spare parts are stocked through third-party logistics suppliers in more than 200 locations worldwide.

Suppliers and costs

Arista outsources most manufacturing and supply-chain management to contract manufacturers. Its primary manufacturing partners are Jabil Inc., Sanmina Corporation and Foxconn Hon Hai, which manufacture products in Malaysia, Vietnam, Mexico and other countries. Contract manufacturers procure components and assemble products to Arista’s specifications, while Arista controls bills of materials, approved component suppliers, test procedures and quality-assurance programs.

The company relies on merchant silicon, integrated circuits and power supplies from a limited number of suppliers, including sole-source providers. Broadcom is its predominant merchant-silicon vendor for switching chips. Arista and its contract manufacturers generally do not have written supply agreements guaranteeing key-component availability.

Cost of product revenue includes inventory paid to contract manufacturers and merchant-silicon vendors, manufacturing overhead, freight and inventory-management costs. Service costs mainly comprise personnel and other costs for global support and services. Gross margin can be affected by customer and product mix, discounts on large orders, competition, component costs, materials scarcity, freight, tariffs, and excess or obsolete inventory and supplier-liability charges.

Management said Q2 2026 purchase commitments were $9.7 billion, mainly for chips supporting new products and AI deployments.

Competitive position

Arista competes in data-center and campus networking with Cisco, Dell/EMC, Extreme Networks, Hewlett Packard Enterprise, Huawei, Juniper Networks, Nvidia and white-box networking vendors using open-source operating systems. In network detection and response, its offerings compete with Cisco, Darktrace and ExtraHop. In network packet brokers, its DANZ Monitoring Fabric competes with Cisco, Gigamon, Keysight, Netscout and other network-monitoring software providers.

For back-end AI clusters, Ethernet competes with InfiniBand and NVLink interconnects. InfiniBand and NVLink are often sold with Nvidia GPUs as part of vertical solutions. The company also faces customers and partners that may build switches or cloud-service solutions for internal use, and vendors offering commodity hardware with open-source software. Competitive factors include breadth of offerings, reliability, ease of use, pricing, total cost of ownership, performance, scale, programmability, interoperability, bundling, availability, support and fulfillment.

Our reading: EOS, NetDL and CloudVision provide a meaningful software and operational moat because they are used across the company’s hardware and network domains, but customers can still choose white-box hardware and open-source operating systems. Our reading: engineering collaboration, diagnostics, support and the ability to address scale-out and scale-across AI networks are meaningful differentiators, although Cisco, Nvidia and other large vendors have broad product portfolios and bundling capacity.

Growth opportunities

Arista’s growth plan centers on increasing sales in cloud, AI and enterprise data-center Ethernet switching and routing, as well as campus workspace markets. It plans to expand research and development, add features and software applications, broaden its product portfolio, expand its global sales force and deepen channel partnerships.

AI networking is a central product vector. The company’s Etherlink portfolio supports scale-out and scale-across AI uses today and scale-up uses in the future. Management said Etherlink AI fabrics exceeded 100 cumulative customers in Q2 2026. It also said 1.6T products would enter trials in the second half of 2026 and that production is expected to ramp in 2027. The company forecasts the scale-across switching and routing market at roughly $15 to $20 billion in 2030.

Campus, routing, core data-center networking, software and support are additional vectors. The company is developing campus switching, Wi-Fi, SD-WAN, routing, network detection and response, network access control, microsegmentation, CloudVision automation and security capabilities.

Management raised its 2026 outlook to approximately $12.6 billion of revenue and 40% annual growth. It set a 2026 campus revenue goal of at least $1.25 billion and an AI Fabrics goal of at least $3.5 billion. Management also expects the incremental demand it is seeking to fulfill to span back-end AI fabrics, core data-center front end, campus and routing.

Structural risks

• Customer concentration and order timing: Large purchases by a limited number of customers represent a substantial portion of revenue. Large-customer orders can shift with capital-expenditure budgets, cost-reduction efforts, deployment cycles, product qualification and acceptance periods; their discounted pricing can also reduce gross margin.

• Component availability and commitments: Key components come from sole or limited sources, including Broadcom for merchant silicon. Supply shortages, long lead times, supplier changes, memory scarcity and shipment interruptions can affect product availability and cost. Non-cancellable purchase commitments can create excess-inventory and supplier-liability exposure if demand forecasts change.

• Demand and revenue volatility: AI infrastructure demand may be difficult to forecast, and customer trials or acceptance clauses can delay revenue recognition or require returns. Revenue and growth rates may be volatile because order size, timing, complexity and customer acceptance vary.

• Competitive and technology change: The networking market changes rapidly and faces intense competition, consolidation, new entrants, white-box offerings and alternative AI interconnects. Arista must continue to develop products and services that interoperate with third-party operating systems, software and hardware.

• Macroeconomic, trade and international exposure: Inflation, recession risk, tariffs, export controls, currency movements, geopolitical pressures and reduced IT or network-infrastructure spending could affect demand, supply continuity, product costs and margins.

• Quality, security and intellectual property: Product defects, errors, vulnerabilities or cybersecurity incidents could harm the business. Intellectual-property claims could restrict product distribution or services, require redesigns, or lead to damages, royalties or other fees.

Bottom line

Our reading

Our reading: Arista is a networking-platform business whose hardware, EOS software, management tools and support serve AI, cloud, data-center, campus and routing workloads. Its business is exposed to large-customer order timing and to a supply chain centered on limited component sources, while its purchase commitments have risen with AI deployments. Our reading: the breadth of the platform and its software-led operating model are central to its competitive position, but it operates against large systems vendors, vertically integrated AI interconnects and white-box alternatives.