Arista posts record 58.5% revenue growth on AI networking

Arista Networks reported $3.66 billion in third-quarter revenue, up 58.5% year over year, the company’s largest year-over-year revenue growth since it became a public company.

CEO and chairperson Jayshree Ullal told Wall Street analysts on the third-quarter call that AI and enterprise networking led the growth. The result gives a direct read on two infrastructure priorities: AI clusters and enterprise networks. Ullal’s account ties networking demand to distributed inference, while Nightingale ties enterprise growth to reliability and security.

Ullal offered a scale check on the revenue figure. Arista reported $2.9 billion in revenue for all of 2021, according to Ullal’s earlier commentary, a total larger than the company’s latest quarterly revenue.

What is changing in AI networking?

“We are witnessing the migration from highly centralized training compute to disaggregated and distributed inference, prefill, and decode, demanding much more intelligent networking and colocation of compute, storage, and KV cache resources rather than simply focusing on raw feeds and speeds and flops of compute. A well-designed network is becoming a foundational cornerstone for optimal token efficiency,” Ullal said.

Ullal tied that shift to the way AI systems now distribute work. “When we initially began this transformational AI journey, we spoke about four or five AI customers with approximately 50,000 to 100,000 Nvidia GPUs. And in fact, we were largely staring at a proprietary network cycle such as InfiniBand,” Ullal said.

Ullal extended the comparison: “In 2026, we are in a super cycle of AI, having already surpassed 100 Etherlink customers. By the end of ’27, we are forecasting greater than 10 gigawatts of deployment across 10 or more different types of compute accelerators.”

In Ullal’s account, networking has become a deployment constraint rather than a supporting product. The move from centralized training to distributed inference, prefill and decode requires compute, storage and KV cache resources to be connected across a much larger system.

How is Arista approaching AI security?

Ullal said the rise of generative AI attacks and increasing cyber threats make network security integral to customers’ plans. Arista’s Autonomous Virtual Assistant (AVA) gives administrators a direct way to query network conditions, obtain explanations and recommendations, and get help diagnosing problems.

Duda described AVA’s reach: “AVA is incredibly powerful. It seamlessly navigates networking and security infrastructure from campus to WAN, to data center, to cloud, continuously monitoring and troubleshooting, isolating root causes without manual labor-intensive tasks,” said Ken Duda, Arista co-founder, president, CTO & director.

Duda also framed AI as a security risk: “The bad news, of course, is the massive challenge AI poses for us and our customers by discovering and exploiting security vulnerabilities. This threat landscape demands secure and robust networking.”

He then named the collaboration: “Over the past few months, we’ve been collaborating with Anthropic, Google, OpenAI, and many others to integrate new AI-enabled security capabilities from frontier models into our existing software security pipeline,” Duda said.

Arista has maintained one of the industry’s lowest CVE counts, with fewer than 100 over the past two decades, Duda said. He also said vulnerability counts elsewhere are rising exponentially, partly because multiple networking operating systems tend to expose more vulnerabilities than a single operating system.

Why are enterprise campuses growing?

“We’re seeing phenomenal growth across our enterprise business, driven by customers building mission-critical networks, which benefit from Arista’s industry-leading quality and reliability,” said Todd Nightingale, president & COO of Arista.

“As the entire industry responds to AI-powered attack exposure, customers are motivated by the fact that keeping their infrastructure current with security updates will protect, segment, and defend against vulnerabilities beyond the network,” Nightingale said.

Over the past few years, Nightingale said, nearly every significant enterprise had reconsidered its campus and branch networks and now treats them as mission-critical, with reliability the highest need.

“This has only accelerated because of the Glasswing and Mythos effect where people are really focused deeply not only on providing the highest level of reliability but being able to do that as they rapidly patch and adjust their infrastructure over time,” Nightingale said.

“And the differentiation with EOS and the ability for Arista software to be able to do this with practically no downtime with extremely, extremely high uptime, it’s highly differentiated in the market. And I really think that’s the core of what’s driving market share gains here,” Nightingale added.

Arista has a campus revenue target of more than $1.25 billion for the year. The target gives the enterprise push a measurable financial marker, while Nightingale ties broader share gains to EOS software that supports rapid patching with minimal downtime.

Where is Arista putting R&D money?

Arista reported quarterly R&D spending of a little more than $130 million, up year over year. Duda said R&D normally runs at approximately 9% to 11% of revenue.

“We’re building so much new stuff. The opportunity is humongous in AI … and more engineering resources devoted to hyperscale partnering,” Duda said.

Duda also said Arista continues to work with Cerebras, Microsoft and Maya on incorporating accelerators, tuning buffers and optimizing software to maximize I/O throughput.

Duda tied the spending increase to a new liquid-cooling technology introduction and the major infrastructure and technology upgrades it requires: “In addition, there’s a new technology introduction around liquid cooling. This requires major infrastructure upgrades and technology upgrades on our part. So I think that’s what’s really driving the expansion in spending there,” Duda said.

In March, Arista announced the eXtra-dense Pluggable Optics (XPO) module, which offers 12.8Tbps of bandwidth across 64 electrical lanes. Its integrated liquid-cooled cold plate supports 400W+ module power consumption, Arista stated.

Arista also said it has assembled about 45 leading optics-module suppliers under a multi-source agreement (MSA) to build and support XPO. The module pairs 12.8 Tbps of bandwidth with a liquid-cooled cold plate capable of supporting 400W+, matching the infrastructure and technology upgrades Duda linked to liquid cooling.

Arista’s programs span the requirements Ullal and Nightingale described: distributed AI compute, rapid patching, reliable campuses, accelerator-specific I/O, liquid cooling and optical transport. The scope ties Ullal’s AI forecast to Nightingale’s enterprise target and Duda’s engineering spending.

What must scale next?

Ullal’s end-2027 forecast of more than 10 gigawatts across 10 or more accelerator types sets the next test for Arista: its networking systems, security software and optical modules must scale with a wider AI market while enterprise buyers remain focused on reliability and security.