ANET — Arista Networks, Inc.

The tape and the analyst community are aligned bullish as of 2026-09-13. Q2 2026 (reported 2026-08-04) delivered the company's first $3B+ quarter ($3.036B, +37.7% YoY) and non-GAAP EPS of $1.02 that crushed the ~$0.86-0.88 consensus, and management raised the FY2026 revenue outlook to ~$12.6B (~40% growth) on better component visibility. The analyst breakdown shows 41 buys and 12 holds with zero sells; consensus PT sits near $224.85 and recent PTs average $238.91. The main counterweight is valuation (trailing P/E >62x), and the stock's -5.18% one-month return as of 2026-09-13 shows the market is quick to trim when execution risk is the only thing priced. Net: bullish, but a bull case that depends on flawless follow-through.

This is a clean, still-accelerating growth story, not a blip. Revenue grew from $1.510B (Q3 2023) to $3.036B (Q2 2026), with YoY growth climbing steadily from ~20% (Q3 2024) to 25.3% (Q4 2024), 27.5-30.4% through 2025, then 35.1% (Q1 2026) and 37.7% (Q2 2026) - acceleration, not deceleration. Diluted EPS roughly doubled from $0.43 to $0.95 over the same window (with one small sequential dip in Q3 2025 to $0.67 that immediately re-accelerated). GAAP operating margin has held in a tight 41-45% band and printed its best level of the series in Q2 2026 at 45.4% (non-GAAP ~49.9%). The one soft spot worth watching: gross margin has slipped from ~65.2% (Q2 2025) to 62.9% (Q2 2026), consistent with the component-price-inflation warnings in the research brief - a mild compression, not yet a structural break. FCF is powerful and lumpy ($5.16B TTM), though Q2 2026 operating cash flow of $1,083M was actually below Q2 2025's $1,201M on working-capital/purchase-commitment build - not a red flag at this growth rate, but the reason the deterministic 'growth' sub-score reads softer than the top line suggests.

Expensive, both versus its own history and versus its growth. As of 2026-09-13, ANET's trailing P/E was 62.2x against a 5-year average of 42.0x, and P/S was 23.8x against a 5-year average of 14.8x - roughly a 48% and 61% premium to its own long-run multiples, respectively. On the $4.106 next-FY EPS estimate, the forward P/E is ~48.6x, with a TTM PEG of 2.65. That is a rich multiple for a hardware-levered name, but it is a genuinely premium franchise: 63% gross margin, 43% operating margin, 38% net margin, 22.6% ROIC, zero debt, and 2.96x current ratio. The premium is 'earned' but leaves very little cushion - the stock is discounting the raised $12.6B guide and no margin slippage.

At ~48.6x forward earnings, the market is not pricing in a turnaround or a depressed cycle - it is pricing in near-perfect execution of a raised guidance number. Revenue and EPS still have to arrive; Q2 2026 beat comfortably and management raised, so the estimate direction is up, but the multiple means the stock is borrowing heavily against future quarters. Practical read: a beat-and-raise in Q3 2026 likely holds or extends the multiple, while any guidance trim (or gross-margin guide-down from component inflation) would compress a 50x multiple violently. This is a 'prove-it-every-quarter' stock, not a value-with-catalyst setup.

Trend is intact but extended as of 2026-09-13. RSI(14) stood at 57.06 - neutral, neither overbought nor oversold, so there is room to run before momentum stretches. Price sat +3.33% above the 20-day, +7.71% above the 50-day, and +30.11% above the 200-day SMA, a classic strong-uptrend stack. The stock is inside its 52-week range of $114.52-$214.89, up 22.3% over three months and 49.4% over six, but off 5.18% over the past month - a normal shakeout within an uptrend, and the recent +3.0% week / +5.6% day shows buyers stepping back in. With the 200-day premium at +30%, this is not a spot to chase aggressively; pullbacks toward the 50-day (~$185 area) are the more sensible add zones.

• AI infrastructure capex cycle: hyperscaler (Microsoft, Meta) spend on generative-AI compute is the dominant demand driver for Arista's high-speed Ethernet switching and AI fabrics - powerful tailwind, but a cyclical capital-budget line, not a subscription.
• Component supply constraints: advanced silicon, memory, and optics remain tight; CEO Ullal flagged industry tightness potentially lasting through 2027, and Arista's multiyear purchase commitments ballooned to ~$9.7B from ~$3.6B a year earlier - an input-cost and working-capital risk.
• High-multiple rate sensitivity: with a 1.617 beta and a 62x trailing P/E, ANET is a long-duration asset whose multiple is exposed to any rise in rates or risk-off rotation out of growth.
• Customer concentration: demand is heavily weighted toward a handful of mega-cap hyperscalers, so a single capex pause at Microsoft or Meta would hit estimates disproportionately.
• Competitive intensity: Nvidia's Spectrum-X, Broadcom's merchant silicon push, Cisco, and whitebox/ODM alternatives all contest the AI-networking fabric, a structural margin and share risk as the category scales.

• Q3 2026 earnings (expected early-to-mid November 2026) - the next hard checkpoint on whether the raised ~$12.6B FY2026 guide is tracking and whether gross margin holds.
• Customer adoption ramp of newly introduced 1.6 Tbps AI fabric platforms and liquid-cooled configurations, plus enterprise/security expansion (AI-Driven Zero Trust Branch, Palo Alto Networks integration).
• Further guidance raises or analyst PT revisions - recent PTs already average $238.91 versus a $224.85 consensus, so upward revisions are live.
• Easing of component supply constraints, which would de-risk the gross margin trajectory and the $9.7B purchase-commitment overhang.

• Valuation risk is the dominant one: 62x trailing / ~48.6x forward and 23.8x sales leave no room for a miss, and a multiple de-rate can hurt far more than a modest fundamental stumble.
• Gross margin compression from component price inflation (already 65.2% -> 62.9% YoY) and the risk that the $9.7B of purchase commitments locks in elevated input costs.
• Customer concentration in Microsoft and Meta - any macro-driven slowdown in their AI capex is an outsized hit.
• Competitive share loss to Nvidia Spectrum-X, Broadcom, or Cisco in AI networking fabrics.
• Working-capital/lumpiness risk: Q2 2026 operating cash flow of $1,083M came in below the prior-year $1,201M despite record net income, and the accruals ratio turned modestly positive (0.55%), so earnings are running slightly ahead of cash.
• Macro/rate sensitivity given a 1.617 beta and a very high earnings multiple.

Arista is a genuinely high-quality, still-accelerating AI-networking franchise - 38% revenue growth, ~45% operating margin, 22.6% ROIC, zero debt, and a $12.6B raised guide - trading at a price that already reflects most of that promise. As of 2026-09-13 at $199.59, the ~48.6x forward multiple and 62x trailing P/E mean you are paying up for execution that has not yet happened. This is a good business at a full-to-rich price: own it for the AI capex cycle with position-sizing discipline and an eye on gross margin, not because it is a bargain.

Forward-P/E approach: I applied a justified forward multiple of 50x (a premium to ANET's 42x five-year average trailing P/E, warranted by the step-up in growth to ~38% revenue / ~28% forward EPS growth, 45% operating margins, and a debt-free balance sheet, but below the ~55x implied by the analyst consensus) to the next-FY EPS estimate of $4.106, giving ~$205.3. Cross-checked against a simplified 10-year DCF (22% FCF growth years 1-5, 12% years 6-10, 3% terminal, 10% discount) which returns roughly $177 - a reminder that the DCF is sensitive to assumptions and that the market is paying for a longer growth runway than a conservative DCF credits.