Arista Networks

Arista Networks' second-quarter results handily beat guidance, and management raised its full-year revenue guide by 10%. Sales rose 38% year over year to $3.06 billion, and 2026 guidance now sits at $12.6 billion (up from $11.5 billion), representing 40% growth.

Why it matters: We see Arista as best-of-breed for high-speed connectivity for artificial intelligence and non-AI, and network speed is increasingly a driver of AI model performance. As a result, demand is robust, and we credit the firm for navigating supply constraints to generate such a significant increase in guidance. Supply constraints abound across the AI supply chain, including with Arista's peers. We don't see any worsening in its competitive position as a result. Still, we like that this has become a strategic focus for management, headlined by rising purchase commitments for components. Management pegs Arista's "scale across" AI revenue, for connections between data centers, at $1.2 billion for 2026, and expects the market to rise above a 60% annualized clip through 2030. We expect Arista to maintain a strong share near 40% in this market, approaching $7 billion in 2030.

The bottom line: We raise our fair value estimate for wide-moat Arista to $230 from $190, behind a higher medium-term growth forecast. Shares rose 10% after-hours, on top of a longer rally since June. After recent strength, the market now looks closer in line with our bullish valuation. Arista faced pressure after its first-quarter results amid bearish market narratives about the stock's optics and market share. We're focused on strong demand and execution, and see Arista agnostic to optics architectures. The market now appears to agree with us. We model above management's 40% growth guidance for 2026 (the firm guides conservatively) and expect strong growth to endure, at close to 30% in 2027 and 20% in 2028. Pending supply availability, we could see further upside in 2027, as we expect demand to be just as strong as in 2026.

We view the revenue guidance raise for 2026 as extremely positive, and we interpret it as an ongoing strong-demand environment with upside from Arista. Nevertheless, management refrained from raising its 2026 AI revenue guide ($3.5 billion) or campus revenue guide ($1.25 billion). In short, management cited higher supply, giving confidence in increased revenue, and noted that incremental revenue can be spread across multiple markets. It'll be up to its go-to-market motion to determine where that revenue shakes out, but Arista expects upside to both numbers (even if neither got officially raised). We model $3.75 billion in AI revenue in 2026, but see plausible upside to $4 billion or more based on new guidance.

Arista's stock has faced trading pressure this year amid a bearish narrative that it wasn't well positioned for a world transitioning to co-packaged optics. We disagree with this bearish view, and believe that Arista sits one level above the optics paradigm as an agnostic switch provider. We see Arista as well-positioned to sell CPO switches. Furthermore, we don't believe CPO will take over the market anytime soon—we expect pluggable and copper architectures to remain long-term complements. Either way, we expect Arista to perform well no matter how optics architectures shake out over the next five years.

Our fair value estimate for Arista Networks is $230 per share. Our valuation implies a 2026 enterprise value to sales multiple of 21 times and a 2026 adjusted price/earnings multiple of 54 times. Against our estimate of 2027 and 2028 non-GAAP earnings, our valuation implies multiples of 43 times and 35 times, respectively. The greatest drivers to our valuation are growth in high-speed data center switching, across both AI and non-AI deployments.

We forecast 24% compound annual sales growth through 2030 for Arista. Sales into data centers are the biggest contributor to sales over our forecast, and we model a 27% compound annual growth rate. We expect Arista’s data center sales to increase faster than the market after 2026, with continued market share gains alongside Nvidia’s networking portfolio, and against all other peers. In particular, we see Arista leading a shift to Ethernet connectivity in artificial intelligence cluster networking, and we expect Arista to take share in this rapidly growing market. We expect Arista to outperform its 2026 target of $3.5 billion in total AI revenue, which we see as conservative.

We expect significant share gains for Arista’s campus portfolio as well, with the firm’s campus revenue increasing 20% annually through 2030 in our model. We expect Arista to surpass management’s $1.25 billion goal in 2026, inclusive of inorganic contribution from the firm’s 2025 acquisition of VeloCloud from Broadcom. We think heightened product development and go-to-market investment will be key to Arista’s penetration of these markets.

We forecast Arista’s non-GAAP gross margin to remain in its target range of 62%-64%. While in prior years the firm eclipsed the top end of this range, we believe a higher mix of large cloud customers will keep gross margin at this relatively lower, but very healthy range. We model Arista’s non-GAAP operating margin to remain close to 50% through 2030. Arista’s longer-term margin target is 40%, which appears conservative to us. While we expect Arista to increase its expenses rapidly to pursue market share gains in enterprise and campus markets, we also expect strong revenue that will prevent any operating margin deterioration. We don’t see our forecast for high operating margins implying underinvestment by the firm.

We assign Arista Networks a wide economic moat rating based on intangible assets in high-speed networking and customer switching costs. We view Arista’s high-speed switches and software-led approach as significantly differentiated from other networking competitors and very difficult to replicate. We expect strength in high-speed switching to generate economic profits for Arista, more likely than not, over the next 20 years.

Arista’s networking switches for high-speed applications are best of breed, in our view, resulting from a software-led approach over its networking hardware. Arista’s specialty within networking is high-speed switches, at speeds of 100 gigabits or more, that are designed for data centers. These switches create a local network to then connect to a wider network and the internet. Data traffic continues to explode, increasing the need for Arista’s gear, in our view.

We believe Arista’s high-speed switches are the preferred option of both public cloud providers and enterprises building private clouds, and it occupies roughly one third of the market for 100-gigabit ports and faster. It has gained this share steadily since its founding in 2004 and has dethroned networking colossus Cisco at high speeds. We believe Arista’s share gains are the result of fundamentally better performance at higher speeds and a software-led approach that outsources semiconductor development to merchant silicon from the likes of Broadcom and Marvell.

Arista’s extensible operating system sets it apart from Cisco and other networking players and represents considerable intangible assets, in our view. We see Arista’s differentiation clearly in its first-class gross profit margins—rivalled only by Cisco (with which it is neck-and-neck) in networking equipment providers.

Arista’s decentralized software is the product of nearly two decades of development—moving from system-based software initially to a newer database and data lake approach. EOS presents a unique value proposition: its decentralized and open structure makes it appealing to hyperscalers with large IT teams that want to fine-tune software for mass deployment, while its robust array of features and modular APIs makes it easy to deploy for smaller-scale enterprise data centers. For all customers, Arista’s single image and fault-containment offer the possibility of easier network upgrades and lower downtime.

This contrasts with a networking competitor like Cisco, which focuses on features but locks customers into a closed ecosystem of chips, hardware, and software, all of which generate significant switching costs at their respective customers. We don’t foresee Cisco abandoning its walled garden in favor of Arista’s open approach. An existing networking competitor would have to rebuild its software platform to rival EOS’ decentralization and single image, and likely risk an inferior product lacking over a decade of evolution and fine-tuning based on customer feedback. On the other hand, an aspirational new entrant would find it difficult to build out an equally rich feature set of code and match Arista’s efficiency and security, even if it were to achieve hardware parity using merchant silicon.

We also see switching costs for Arista. We think networking equipment broadly carries switching costs for well-positioned providers. Arista’s proprietary software and hardware configurations require training and certifications, requiring weeks of class, hours of testing, and thousands of dollars in expense. Furthermore, networking setups take months to implement and years to optimize—Forrester claims that a typical business doesn’t fully utilize the value of new networking technology for two years. If a company were to switch away from Arista, it would require wholesale retraining of its workforce under a new vendor, inclusive of the time and expense to do so. Furthermore, an IT department would spend the time and expense of getting the new network off the ground and have to redesign workflows and processes to accommodate different software and optimize operations. Even cloud customers, which we typically think of as conducive to lower switching costs due to their multisourcing and disaggregated buying patterns, experience the stickiness of Arista equipment. Even if Arista isn’t installed in an entire network, the portions in which it operates become sticky, as cloud customers develop workflows for, and write software on, EOS. We also view EOS as complementary to base software like Microsoft’s SONiC or Meta’s FBOSS. Microsoft, as an example, uses EOS working on top of SONiC for portions of its network, which gives it the programmability and features of EOS while maintaining consistency with other non-Arista portions of its network.

We view the combination of Arista’s strength in networking hardware and software design with inherent switching costs to networking equipment as creating a wide economic moat to defend against competitive threats to economic returns for 20 years.

Arista Networks has gained a top market share position in high-speed switching, resulting, in our view, from technology leadership, and it continues to gain share.

Arista holds best-in-class profit margins and earns robust economic profits, reflecting its strong value proposition and wide economic moat.

Arista earns heady free cash flow, which it can use for organic investment and shareholder returns.

Arista Networks has a weaker position in areas of networking outside high-speed switching and may struggle to expand into adjacent markets.

We see higher competition for Arista within generative AI spending, with Nvidia holding a large incumbency via bundles of its proprietary networking equipment with its GPUs.

Arista’s acquisition history is small and new, and it could risk destroying shareholder value with ill-advised deals.