ANET: The Thesis Holds, but $205 Is Not an Entry
Arista’s AI-networking thesis is intact, but the price has outrun the evidence. ANET closed at $204.92 on 28 September 2026, 33.0% above the $154.03 price on 25 May 2026 at which the original work supported buying; that turns the call from BUY to HOLD. (Market data, Q1’26 filing)
01ANET — HOLD
Why now. This is a quiet-day re-underwrite, not a new catalyst call. The question is whether the operating case still compensates for the stock’s move from $154.03 as of 25 May 2026 to a $204.92 close as of 28 September 2026; at today’s evidence level, it does not justify fresh buying. (Q1’26 filing, Market data)
The evidence. The core thesis was built on Q1’26 revenue of $2.709bn, up 35.1% year over year, with 61.9% gross margin, 42.7% operating margin, $1.694bn operating cash flow, $2.790bn cash at 31 March 2026, and no debt. Those figures support the quality case: AI and cloud build-outs are pulling through high-speed Ethernet demand, while EOS gives Arista a sticky software layer around the hardware. But the same filing showed gross margin down from 63.7% in Q1’25, and the original memo’s valuation support—40.7x forward earnings as of 22 May 2026—was already demanding before the subsequent share-price rise. (Q1’26 filing, Arista investor relations)
Levels and triggers. Do not chase $204.92 as of 28 September 2026. I would reconsider ACCUMULATE below $175, provided reported year-over-year revenue growth remains at least 25% and gross margin remains at least 58%; those are the operating thresholds in the original underwrite, not forecasts. Confirmation would be another filing showing that AI-network growth remains above that revenue threshold without a structural margin break. The thesis is killed by revenue growth below 25% for two consecutive reported quarters, gross margin sustainably below 58%, or a disclosed major design loss at a top hyperscale customer. (Q1’26 filing, Arista investor relations)
Horizon. This HOLD is for the next 12–18 months from 29 September 2026: long enough for several quarterly filings to test AI-network demand and product-mix economics, but not an endorsement at any price. (Arista investor relations)
The bear case. NVIDIA’s Spectrum-X can turn an expanding Ethernet market into a share-loss problem for Arista, while customer concentration means a hyperscaler capex pause can hit growth abruptly. If AI switching also carries structurally lower hardware margins, revenue can keep rising while the earnings quality that supports the premium multiple deteriorates; the fall from 63.7% gross margin in Q1’25 to 61.9% in Q1’26 is the warning to watch. (Q1’26 filing)
What I’m watching: the next Arista 10-Q—specifically year-over-year revenue growth, gross margin, and any disclosure pointing to a hyperscale design loss. (SEC company filings)
02Sources
Independent equity analysis, for information only, not investment advice.
Originally published in the Axelrod Research newsletter.