ANET
Catalysts
Key Risks
The Opportunity
Arista Networks builds the high-speed networking equipment that connects servers inside the massive data centers run by companies like Microsoft and Meta. Think of them as the company that builds the highways inside the internet's biggest buildings. They make the switches and the software that lets thousands of computers talk to each other at incredible speeds - and right now, the explosion of artificial intelligence is causing those buildings to get much, much bigger.
The business itself is exceptional. Revenue has grown from about $1 billion in 2016 to over $10 billion today, and the company keeps nearly 40 cents of every dollar as pure profit. They carry zero debt, generate over $5 billion in free cash per year, and have beaten Wall Street's earnings expectations every single quarter in recent memory. The CEO has been in charge for 17 years and has executed flawlessly. The co-founder still holds a massive personal stake. This is a well-run machine.
The problem is the price tag. At roughly $199 per share, the stock is trading at about 62 times its annual earnings. The market is essentially betting that Arista will grow at 50% per year to justify this price - a pace that even the company's own exceptional recent performance (37-40% growth) cannot sustain indefinitely. Our analysis, using multiple valuation approaches weighted toward growth-friendly methods, arrives at a mid-point intrinsic value of approximately $117 per share. That is about 41% below where the stock trades today.
What could go right? AI infrastructure spending could last longer and grow faster than anyone expects. Arista's new push into enterprise campus networking (through its VeloCloud acquisition) could open a massive new market. The upgrade cycle to faster switching speeds could sustain above-average growth for years. In a bull scenario, the stock could eventually grow into its valuation.
What could go wrong? Arista depends heavily on just two customers for a large share of revenue. If either Microsoft or Meta cuts back on data center spending, shifts to a competing technology, or starts building its own networking equipment, Arista's growth story changes overnight. More broadly, if AI spending cools even modestly, the premium valuation could compress sharply - and with a stock this expensive, a 30-40% correction from a growth scare is not unusual. The business is outstanding, but the price already reflects a near-perfect future.
Breakdown
Arista's balance sheet is exceptionally clean. As of Q2 2026, total assets stand at $23.72B against total liabilities of $8.92B, yielding total equity of $14.80B. The company carries zero long-term debt and zero current debt - a rarity in the technology sector.
Cash on hand is $2.29B ($9.81/share), down from $2.79B in Q1 2026, likely reflecting the $300M VeloCloud acquisition from Broadcom [Arista Press Release, July 2025] and normal working capital swings. The current ratio of 2.96 and quick ratio of 2.59 indicate robust short-term liquidity. The key balance sheet consideration for a networking equipment company is inventory and intangible asset valuation.
While granular breakdowns are unavailable from the data provided, the absence of goodwill-heavy M&A history (VeloCloud at $300M is modest relative to $23.7B in assets) suggests book values are reasonably close to fair values. The investment portfolio (money market funds, commercial paper, government and corporate bonds, agency securities) is Level 1 and Level 2 fair value instruments per the 10-K and 10-Q XBRL data, meaning these are marked to market with minimal valuation risk. Book value per share of $10.71 dramatically understates the economic value of the business, which is driven by intellectual property in EOS (Extensible Operating System) and customer relationships - assets that do not appear on the balance sheet at their economic worth.
NAV of $11.73 is essentially irrelevant for a company whose value derives almost entirely from its earnings power and growth runway rather than its tangible asset base.
Arista generates substantial free cash flow - $5.16B on a TTM basis, representing an FCF margin of roughly 49% on TTM revenue of approximately $10.55B. This is extraordinary capital efficiency, supported by an asset-light business model that outsources manufacturing. The company pays no dividend and has no debt to service, so virtually all cash flow is retained.
Arista has historically used excess cash for share repurchases: the December 2024 $1.2B buyback authorization is referenced in the SEC filings. However, with shares outstanding at 1.26B and the stock trading at nearly $200, buybacks at these levels are not value-accretive from a per-share basis. The VeloCloud acquisition for $300M [Network World, 2025] is Arista's most significant M&A in years - a disciplined, bolt-on deal representing less than 6% of one year's FCF.
Revenue per employee is approximately $1.76M ($9.01B FY2025 / 5,115 employees), which is exceptional and far above the networking equipment peer group. Capital allocation has been conservative and shareholder-friendly: no dilutive acquisitions, no excessive stock compensation visible in the share count trajectory (shares have remained relatively stable around 1.26B), and a growing cash hoard. The primary concern is that with $2.29B in cash and $5B+ annual FCF, the company is accumulating capital without a clear return-maximizing deployment strategy beyond buybacks at elevated multiples.
Arista's financial track record over the past decade is among the most impressive in the technology sector. Revenue has compounded from $1.13B in 2016 to $9.01B in 2025 - a nearly 8x increase in nine years, representing a CAGR of approximately 26%. Growth has accelerated rather than decelerated: 2024 revenue grew 19.5% to $7.0B, and 2025 revenue grew 28.7% to $9.01B.
TTM revenue through Q2 2026 is approximately $10.55B, and management has guided FY2026 to $12.6B [Business Wire, August 2026], implying roughly 40% growth. Gross margins have been remarkably stable: 63.0% in TTM versus 64.2% in 2024, 61.9% in 2023, and 63.9% in 2018. The slight Q2 2026 margin of 62.8% versus 65.6% prior year is worth monitoring as a potential mix shift signal [Business Wire, August 2026].
Operating margins have expanded from 21.5% in 2016 to 43.1% TTM, demonstrating massive operating leverage. Net margins have expanded similarly from 16.3% to 38.4%. EPS has grown from $0.16 in 2016 to $3.21 TTM.
Critically, Arista has beaten consensus EPS estimates in every reported quarter in the dataset: Q1 2025 through Q3 2026 show consistent beats, with actual EPS exceeding estimates by 5-15% each quarter. This consistency of execution and under-promise/over-deliver pattern is a hallmark of strong management.
The consensus analyst estimate projects 21.8% EPS growth over the next five years, which appears conservative given current trajectory. FY2026 guidance of $12.6B revenue [Business Wire, August 2026] implies approximately 40% growth over FY2025's $9.01B - well above the 5-year average. However, sustainability is the critical question.
Arista's AI-driven revenue target was raised to $3.5B for FY2026 [Arista Q1 2026 Earnings, May 2026], representing roughly 28% of total guided revenue. The AI infrastructure spending cycle is currently in a hyper-growth phase with Microsoft and Meta (Arista's two largest customers) planning massive data center buildouts. The enterprise network equipment market is projected to grow at 11.8% CAGR through 2031 [Mordor Intelligence, 2025], providing a solid baseline growth floor.
My base-case assumption is that Arista can sustain 20-22% EPS growth for the next 3-5 years, driven by: (1) continued AI infrastructure buildout at hyperscalers, (2) market share gains in campus networking via VeloCloud, and (3) the 400G/800G/1.6T upgrade cycle. The reverse DCF implies the market is pricing in 50% annual growth, which is unsustainable beyond the near term. Even a deceleration from 40% to 20% growth - still excellent by any measure - would represent a significant headwind to the current valuation multiple.
Forward P/E of 38.5x is rich even for a premium growth compounder; it implies the market is pricing in both high growth AND multiple maintenance, a combination that historically does not persist.
Arista possesses a narrow-to-wide moat built on several reinforcing advantages. First, EOS (Extensible Operating System) runs a single software image across all hardware platforms, creating significant switching costs - once a data center is built on EOS, the operational tooling, automation scripts, and network management workflows are deeply integrated. Second, Arista has achieved technology leadership in high-speed Ethernet switching, with the recent 1.6 Tbps Etherlink platform [Simply Wall St, 2026] and XPO optics technology positioning it at the frontier of AI networking.
Third, deep customer relationships with hyperscalers (Microsoft and Meta as top-two customers) create incumbency advantages - these customers value proven reliability at scale over cost savings from switching. Arista has surpassed Cisco in market share for high-speed data center switching [Network World, 2025; Oreate AI Blog, 2025], a remarkable achievement against a deeply entrenched incumbent. The moat trend is strengthening: the VeloCloud acquisition extends Arista's reach from data center core into enterprise branch/campus [Network World, 2025], and AI workload requirements favor Arista's Ethernet-based approach over NVIDIA's proprietary InfiniBand.
However, the moat faces a genuine test: NVIDIA's networking capabilities are expanding, HPE/Juniper combined is a credible challenger [Hudson Labs, 2026], and Broadcom supplies critical merchant silicon to Arista, creating supply chain dependency. I assess this as a narrow moat trending toward wide, but acknowledge that 10+ year moat durability predictions in fast-moving technology sectors carry inherent uncertainty.
CEO Jayshree Ullal has led Arista for 17+ years, presiding over one of the most successful growth stories in networking history. Her track record on capital allocation is strong: disciplined M&A (VeloCloud at $300M is characteristic), consistent execution against guidance (every quarter in the dataset shows beats), and no empire-building acquisitions. Insider ownership at 17.01% is exceptionally high, providing strong alignment - co-founder Andreas Bechtolsheim alone holds 4.66% (58.57M shares) [WallStreetZen, 2026].
Recent insider transactions show modest net selling (-1.75%), primarily from Ullal ($10.65M in April 2026 sales), which is routine portfolio management for a long-tenured executive with substantial holdings. The elevation of Kenneth Duda to President & CTO [SEC Form 8-K, September 2025] and hiring of Todd Nightingale (ex-Cisco) as COO [Simply Wall St, 2025] signals thoughtful succession planning and competitive intelligence acquisition. The one governance blemish is Bechtolsheim's SEC insider trading settlement (~$1M civil penalty) related to trading ahead of the Acacia Communications acquisition [SEC Litigation Release LR-25955].
This was a personal matter, not corporate, but it reflects negatively on an individual who remains in a senior role. Overall, management quality is a clear positive, though I acknowledge that AI-based assessment cannot capture interpersonal dynamics, board independence in practice, or the quality of internal debates.
Customer concentration is the single most material risk. Microsoft and Meta (via resellers) account for a significant portion of revenue - the 10-K references two resellers each exceeding 10% of revenue across 2023-2025. If either hyperscaler reduces capital spending, brings networking in-house, or shifts to a competitor (particularly NVIDIA for AI networking), the revenue impact would be severe.
Competitive risk is moderate but intensifying: HPE's acquisition of Juniper creates a formidable combined entity targeting the same campus/enterprise segment Arista is entering [Hudson Labs, 2026], and NVIDIA's Ethernet ambitions (via Spectrum-X and related platforms) could erode Arista's AI networking advantage. Legal exposure appears manageable: the Orckit patent suit [CourtListener, 2023-ongoing] and WSOU Federal Circuit case [CAFC Opinion, March 2025] are typical patent disputes for a technology company of this size, with no indications of material financial exposure. Technology risk centers on the Ethernet-vs-InfiniBand debate for AI clusters - if InfiniBand proves definitively superior for large-scale AI training, Arista's core thesis weakens.
Valuation risk is the elephant in the room: at 62x TTM earnings and a reverse DCF implying 50% growth, any growth disappointment or multiple compression would result in outsized stock price declines. Beta of 1.6 confirms elevated systematic risk exposure. Supply chain risk from Broadcom merchant silicon dependency adds another dimension.
Arista operates in a sector experiencing a rare confluence of secular tailwinds. The enterprise network equipment market is projected to grow from $81.75B to $163B by 2031 at 11.8% CAGR [Mordor Intelligence, 2025], with AI infrastructure spending providing an acceleration layer above baseline growth. Arista has entered the Fortune 500 for the first time in 2026 [Simply Wall St, 2026] and was named a Leader in Gartner's Magic Quadrant for Enterprise Wired and Wireless LAN [Simply Wall St, 2026] - clear markers of industry leadership.
Institutional positioning is strongly bullish: T. Rowe Price increased its stake by 845% [HoldingsChannel, 2025-2026], and overall institutional ownership is 73%. The analyst consensus of 1.17 (where 1 = strong buy) is near-unanimously bullish.
Social sentiment scores of 7-8/10 across platforms confirm positive market perception. No activist investor campaigns or takeover interest has been identified - Arista is an acquirer, not a target, as evidenced by VeloCloud. The competitive landscape is shifting favorably for Arista as hyperscaler AI spending continues to accelerate, with five major companies expected to spend over $700B on data center infrastructure in 2026 [Arista Networks Stock Analysis, April 2026].
However, this overwhelmingly positive sentiment and positioning itself constitutes a risk: when expectations are universally high, even modest disappointments can trigger severe corrections.