ANET — Arista Networks, Inc.

The August 4, 2026 Q2 beat drove the mood: revenue of $3.04B beat by roughly $200M, non-GAAP EPS of $1.02 beat the $0.86-$0.89 consensus, and management raised full-year guidance for the third straight time. Etherlink AI fabric passed 100 customers with a $3.5B+ AI-fabric revenue target for 2026, and several banks raised targets to $215-$259. Insider sales were pre-planned 10b5-1 sales, not a red flag.

Growth is accelerating, not just elevated. YoY revenue growth went from roughly 30% in Q2'25 to 28% in Q4'25, 35% in Q1'26, then 38% in Q2'26, with Q3'26 guidance implying roughly 43%. Quarterly revenue stepped from $1.51B in Sep-2023 to $3.04B in Jun-2026. Operating margins stayed in the low-to-mid 40s, reaching 45.4% in Q2'26, and GAAP EPS rose from $0.48 in Dec-2023 to $0.95 in Jun-2026. The Q3'25 EPS dip to $0.67 was a one-off blip, not a reversal.

Expensive versus its own history. As of 2026-08-27, TTM P/E was 62.6x versus the 5-year average of 42.0x, and TTM P/S was 24.0x versus the 5-year average of 14.8x. The forward P/E is roughly 49x on FMP's next-fiscal-year consensus EPS of $4.107, which is still above the historical average but more reasonable given ~40% revenue growth, a 63% gross margin, no debt, and a 22.6% ROIC.

At $201.09, the market is pricing in continuation of the AI data-center capex boom rather than a turnaround or trough. A 49x forward multiple against ~40% revenue growth and roughly 49% forward EPS growth gives a PEG around 1.0, so the stock is not absurd if the growth holds. The vulnerability is multiple compression if hyperscaler AI spending eventually normalizes to 15-20% growth; the current price needs earnings to keep catching up.

As of 2026-08-27, RSI(14) was 59.2, firmly in the bullish but not overbought zone. Price was 3.85% above the 20-day SMA, 11.8% above the 50-day SMA, and 33.9% above the 200-day SMA, so the intermediate trend is clearly up. The large gap above the 200-day means a consolidation or pullback would not be unusual, and the stock is also within about 6% of its 52-week high, so chasing here carries more mean-reversion risk than buying after a deeper pullback.

• AI infrastructure capex cycle: hyperscalers are aggressively expanding data-center footprints, directly boosting high-speed Ethernet switching demand.
• Component supply constraints: $9.7B in multiyear purchase obligations and memory/silicon/optics shortages are expected to persist into 2028, creating cost and availability risk.
• Customer concentration: Microsoft and Meta dominate recent AI-fabric demand, so any pause in their capex plans would hit results.
• Higher-for-longer rates and multiples: a 24x P/S stock is sensitive to discount-rate shocks and tech rotation.
• Competition in AI networking: Nvidia, Broadcom, Cisco, and white-box vendors are all targeting the same data-center switching opportunity.

• Q3 2026 earnings, expected early November 2026: confirmation of the ~$3.3B revenue guide and a possible fourth upward revision to FY2026.
• Etherlink AI fabric adoption: further customer additions beyond 100 and progress toward the $3.5B+ dedicated AI-fabric revenue target.
• 1.6T liquid-cooled platform ramp and enterprise/tier-2 cloud diversification beyond the core hyperscaler customers.
• Any easing or worsening of component supply conditions, which could swing margin and revenue visibility.

• Valuation reset: at ~62.6x TTM P/E and ~24x TTM P/S, even a modest growth slowdown could trigger sharp multiple compression.
• Customer concentration: a meaningful share of AI revenue is tied to Microsoft and Meta; a capex pause, inventory digestion, or in-house silicon substitution would hurt growth.
• Supply chain and component inflation: the jump in purchase obligations to $9.7B reduces flexibility, and memory/optics cost inflation could pressure the exceptional operating margin.
• Competitive encroachment: Nvidia networking, Broadcom silicon, Cisco, and white-box Ethernet vendors are all competing for AI data-center sockets.

Arista is a best-in-class franchise riding the strongest data-center capital-spending cycle in years, with accelerating revenue, elite margins, zero debt, and substantial free cash flow. But the price already reflects a lot of that success: this is a great business at a rich price rather than a clear bargain, so the expected returns are more dependent on flawless execution continuing than on valuation re-rating.

I applied a 54x justified P/E to FMP's next-fiscal-year consensus EPS estimate of $4.107. That multiple is a growth-adjusted premium to Arista's 5-year average P/E of 42x, justified by ~40% revenue growth, ~49% operating margins, zero debt, and 22.6% ROIC, but deliberately below the current 62.6x TTM multiple to avoid extrapolating the AI cycle at peak. 54 x 4.107 = 221.78.