Stock and Call
Arista Networks (ANET) is rated Bullish at $154.03. The stock has pulled back ~15% from its May 2026 high of $179.80, creating an attractive entry in a business that just printed its fastest growth quarter of the cycle at +35.1% revenue growth. Recommendation: Buy a position at current levels. Conviction: High. Suggested sizing: 2-4% of a concentrated portfolio — enough to matter, sized for a company with zero debt and fortress margins. Target: $188 (consensus, +22% upside). Stretch target: $200 (Barclays/TD Cowen, +30%). Stop-loss: $130 (-15%), below which the AI switching thesis would have a problem.
Why It Is Promising
Arista makes the high-speed Ethernet switches that connect every GPU inside AI clusters. Every $1B in NVIDIA GPU sales generates roughly $150-250M in adjacent switching revenue, and Arista captures the lion's share of the open-standards Ethernet segment. The thesis is near-linear: as long as hyperscalers buy NVIDIA GPUs, they buy Arista switches.
The company is a cash-printing machine. Q1'26 revenue of $2.7B grew 35.1% YoY — the fastest quarter in the current cycle. Gross margins of 61.9%, operating margins of 42.7%, free cash flow margins of 60.5%, and zero debt with $2.8B cash on the balance sheet. The EOS software platform creates lock-in at the protocol level — customers don't rip and replace network operating systems.
Ethernet is winning over InfiniBand for AI fabrics. Even NVIDIA's own Spectrum-X push validates Ethernet as the AI fabric of choice. But Arista's open-standards approach appeals to hyperscalers who don't want to be locked into NVIDIA's proprietary networking stack — a key competitive moat.
Evidence Quality
Catalyst Path
The primary catalyst is the hyperscaler AI capex supercycle — Meta $125B+, Microsoft, Amazon, and Google building data centers at unprecedented pace, each needing Arista switching fabric. AI inference scale-out is a secondary long-term catalyst: inference clusters will need more switches than training clusters due to geographic distribution. Near-term, the Raymond James upgrade to Outperform (May 15) and rising analyst targets (Barclays $195, TD Cowen $200) provide price catalysts.
Main Risks
Gross margin compression from the AI switch mix shift (61.9% vs 65% in 2024) is the primary risk. If margins trend toward 58%, the bull case weakens significantly. NVIDIA Spectrum-X is a competitive wildcard — if hyperscalers adopt NVIDIA's integrated networking stack, Arista loses share in the fastest-growing segment. Customer concentration in Meta, Microsoft, and AWS creates single-customer capex risk. At 52.9x trailing P/E, any growth deceleration below 25% would trigger multiple compression.
What Would Change the Call
Sell if revenue growth decelerates below 25% for two consecutive quarters — that signals the AI cluster buildout is peaking. Sell if gross margin drops sustainably below 58% (structural mix problem, not temporary). Sell if NVIDIA Spectrum-X wins a design at any of Arista's top 3 hyperscaler customers.
Watchlist Action
Next checkpoint: Q2'26 earnings (expected late July/early August 2026). Watch for revenue growth rate (needs to stay >25% YoY), gross margin trajectory (stabilization above 60% is bullish, below 58% is bearish), and any Spectrum-X competitive commentary. Entry at $154 is attractive; if the stock pulls back to $130-140 on no fundamental news, that is an add zone. If it breaks below $130, the thesis has a problem and the position should be reduced.