Arista Networks Sees AI Demand Fueling 40% Growth, $9.6B in Supply Commitments

Key Points

• Arista raised its revenue outlook to 40% growth, or $12.6 billion, as AI networking demand accelerates. Executives said the current AI infrastructure cycle could continue for years across training, inference, scale-up and scale-across architectures.
• Purchase commitments surged to $9.6 billion from roughly $3.6 billion three quarters earlier, reflecting strong demand and efforts to secure chip supply through 2026 and 2027, though near-term component constraints remain.
• Arista is expanding beyond its largest hyperscale customers into neo-cloud providers, enterprises and international markets, while developing next-generation networking technologies. The company expects gross margins of 62%–64% and operating margins near 48%.

Arista Networks NYSE: ANET executives outlined continued demand tied to artificial intelligence networking, a strengthening supply position and expanding opportunities beyond its largest hyperscale customers during Citi’s Global TMT Conference.

Chief Financial Officer Chantelle Breithaupt said the company had raised its revenue outlook in its latest earnings report to 40% growth, or $12.6 billion. She also addressed investor attention on product deferred revenue, which has increased as AI deployments become more complex.

Product deferred revenue represents equipment that has been shipped, invoiced and paid for but remains subject to customer acceptance criteria, Breithaupt said. Such deployments can take 18 to 24 months before the company recognizes the associated revenue on its income statement.

“I would keep my eye on the P&L growth, what’s happening in deferred revenue,” Breithaupt said, while cautioning against viewing the metric on a quarter-to-quarter basis. Instead, she said investors should evaluate trends over four to six quarters along with purchase commitments and remaining performance obligations.

AI Cycle and Supply Commitments

Breithaupt characterized the current AI buildout as Arista’s second major growth cycle after the cloud deployment cycle of 2019 through 2021. She said the AI cycle is now roughly two-and-a-half to three years old from a materiality standpoint, but could extend for years as networking needs expand across scale-up, scale-out and scale-across architectures, as well as AI training and inference workloads.

On supply, Breithaupt said industry constraints remain, particularly around fab capacity, despite the company’s progress in securing vendor arrangements for capacity and memory through 2026 and 2027. She said Arista is still encountering shorter-term component constraints involving items such as printed circuit boards and capacitors.

Arista’s purchase commitments rose to $9.6 billion at the end of the second quarter, from approximately $3.6 billion three quarters earlier. Breithaupt said the increase reflects demand the company is seeing and supports chip supply extending into the second half of next year.

“We feel very good that the supply chain team learned a lot during COVID, learned a lot during the beginning of AI,” she said.

Networking Requirements for AI

Tyson Lamoreaux, Arista’s senior vice president of cloud and AI networking, said AI infrastructure creates more demanding requirements than conventional cloud networking. AI traffic is highly coordinated, while reliability is especially important because current AI systems have fewer ways to work around network failures.

Lamoreaux said Arista’s heritage in high-performance hardware engineering, software reliability, security and scalability positions the company for the AI era. He highlighted the transition to 1.6-terabit link speeds, which brings greater heat, power and signal-integrity challenges.

He also pointed to Arista’s ability to patch software and automate deployment workflows without interrupting data-plane traffic. That reliability can be important for customers seeking to keep costly AI accelerators highly utilized, he said.

Breithaupt said the company expects two hyperscale customers to account for more than 10% of revenue, while it anticipates a third and potentially a fourth customer reaching that threshold. She also cited opportunities among neo-cloud providers, which are seeking lower token costs through optimized AI infrastructure.

In the enterprise market, she said Arista sees opportunities both in data center share gains and in demand for simplified operations, security and total-cost-of-ownership benefits. The company has about 20% to 22% data center market share, according to Breithaupt, while its campus share is about 5%.

Scale-Across and Scale-Up Roadmaps

Executives said scale-across AI networking could be supported by power constraints, resiliency requirements and government policies that encourage geographically distributed data centers. Lamoreaux cited South Korea as an example, saying the country’s policies supporting distributed power generation and local consumption could lead to more distributed data center infrastructure.

Breithaupt said Arista is revisiting its total addressable market estimate after presenting a $105 billion TAM at its October 2025 analyst day. The company expects to provide an updated view later this year encompassing scale-up and scale-across opportunities.

For scale-up Ethernet, Breithaupt said Arista has included no scale-up revenue in its 2026 guidance. The company expects potential trials and pilots late next year, with revenue beginning in 2028 as the ecosystem and standards develop.

Lamoreaux said Arista’s XPO architecture is aimed at network-powered optics, or NPO, as an intermediate option between copper connections and co-packaged optics, or CPO. He said the company expects early XPO deployments in the second half of next year and a more substantial ramp in 2028, while CPO adoption could begin ramping roughly 12 months later.

Margins, Pricing and International Growth

Breithaupt said Arista is guiding for gross margin of approximately 62% to 64% and operating margin in the 48% range. She said the company expects operating leverage because its operating costs do not need to grow at the same rate as revenue, while continuing to invest in innovation.

On component-cost inflation, including memory, Breithaupt said the company seeks to apply targeted and transparent price increases based on the bill of materials rather than imposing uniform increases across customers. She said Arista does not seek to reprice existing backlog and aims to remain margin-neutral when passing through applicable cost increases.

Outside AI and hyperscale markets, Breithaupt said international expansion remains a long-term growth avenue. She said Europe, the Middle East and Africa posted 36% growth in the prior quarter, supported by large customers as well as regional enterprise and neo-cloud demand. Arista is also investing in channel relationships, particularly internationally, as it expands its enterprise go-to-market capabilities.

Software revenue is expected to remain roughly 18% to 20% of total revenue, executives said, as software licensing and support are generally delivered alongside the company’s networking systems rather than as a separately driven business.

About Arista Networks (NYSE:ANET)

Arista Networks, Inc designs, develops and sells cloud networking solutions for large data centers, cloud service providers, internet companies, enterprises and other organizations. Its portfolio includes high-performance Ethernet switches, routers and wireless networking products used to connect servers, storage systems, users and applications across data center and campus environments.

The company's software offerings include Arista EOS, its Linux-based network operating system, and CloudVision, a cloud-based platform for network management, automation, monitoring and analytics.

This instant news alert was generated by narrative science technology and financial data from MarketBeat in order to provide readers with the fastest reporting and unbiased coverage. Please send any questions or comments about this story to contact@marketbeat.com.