Arista’s beat-and-raise lands with a thud as investors fixate on growth cadence and margin mix; $136 shapes up as the line in the sand

Arista Networks delivered another clean quarter on the numbers, but the stock didn’t get the memo. Shares slid from roughly $155 to $134 before stabilizing, as investors focused less on Q3’s modest beat and more on signs of decelerating sequential revenue and an outlook that, while solid, didn’t match the market’s freshly upgraded hyperscale CapEx dreams. Technically, $136—the October low—now marks a key support level to watch intraday.

Against expectations , Arista posted Q3 revenue of $2.308 billion versus $2.257 billion consensus, and non-GAAP EPS of $0.75 versus $0.72. Non-GAAP gross margin printed 65.2% (a touch above the Street), with operating margin leverage still enviable for a hardware business. The company guided Q4 revenue to $2.3–$2.4 billion (midpoint just above consensus) and forecast non-GAAP gross margin of 62–63% and operating margin of 47–48%. Fundamentally, that’s a beat-and-raise quarter; optically, it wasn’t enough to satisfy a market that had just reset its AI spending expectations higher.

Under the hood, drivers remain broad. Total revenue grew 27.5% year over year, with software and services contributing roughly 18.7% of sales. Data center and AI projects continued to lead, while enterprise (campus) momentum supported the year-on-year gross-margin improvement. Cash generation was strong—$1.3 billion from operations—and the balance sheet remains a fortress with $10.1 billion in cash and investments. Deferred revenue climbed to $4.7 billion (from $4.1 billion in Q2), reflecting both demand and acceptance clauses tied to new products and use cases.

So why the selloff? Two words: cadence and mix. First, cadence. Sequential growth has downshifted: Q2 grew ~10% sequentially, Q3 managed ~1.6%, and the Q4 outlook implies a flatter profile than bulls wanted. On the call, CEO Jayshree Ullal pushed back on the notion that demand is slowing, characterizing the quarterly lumpiness as supply-driven: lead times on standard components, merchant silicon and memory remain long—38 to 52 weeks—so shipments, not orders, are gating revenue. She reiterated confidence in demand and in Arista’s ability to sustain its longer-term 20% growth algorithm.

Second, mix. The Q4 gross-margin guide (62–63%) implies product margins dipping below 60% when cloud and AI titans dominate the ship mix. Management framed this as normal mix variability, not a reset to the model; services/software still carry north-of-product margins, and enterprise remains margin-accretive, but AI/cloud volume swings will tug the blended number quarter to quarter. The guide also bakes in “known tariff scenarios,” another ingredient in near-term GM pressure.

Strategically, Arista’s AI narrative is intact and arguably strengthening. The company says three of its “big four” cloud AI customers have already surpassed the 100,000-accelerator cluster size, with the fourth on pace; beyond hyperscalers, Arista cited momentum with 30–40 non-hyperscaler AI customers. The firm continues to champion open, Ethernet-based scale-up and scale-out fabrics (Ultra Ethernet Consortium and the new ESUN workstream) and highlighted collaboration with Oracle on Acceleron, plus ecosystem ties with NVIDIA, AMD, Broadcom, Anthropic, OpenAI, Pure Storage and VAST Data. Management is also leaning into full-rack solutions: multiple projects (5–7 at any time) spanning cabling, co-packaging, power, cooling and XPU interconnects, with revenue recognition potentially mixing classic OEM and JDM-style models over time.

Headwinds aren’t imaginary. Apart from supply constraints and long lead times, Arista faces competitive pressure from bundled solutions and white-box offerings in certain use cases, which can influence price/mix. International revenue slipped to 20.2% of total (from 21.8% in Q2), a small but notable wobble for those looking for broader geographic contribution. And while deferred revenue is rising, the acceptance-clause mechanics can drive quarterly volatility unrelated to underlying demand—something investors need to stomach in a build-cycle this intense.

The outlook bridges near-term realism and longer-term ambition. For Q4, management expects $2.3–$2.4 billion of revenue, 62–63% non-GAAP GM and 47–48% non-GAAP OM. For full-year 2025, Arista reiterated 26–27% growth (~$8.87 billion midpoint), including $750–800 million from campus and at least $1.5 billion from AI centers. For 2026, the company maintained about 20% growth on a higher dollar base (now ~$10.65 billion), with an AI center target of $2.75 billion and campus at ~$1.25 billion, and a 62–64% gross-margin range depending on mix. Citi stuck with a Buy and $176 target, calling out accelerating AI sales in 2026; however, the sell-side and the tape are plainly wrestling with a gap between Arista’s steady 20%-growth message and the market’s wish for a steeper near-term ramp in the wake of heightened hyperscale CapEx chatter.

What analysts probed on the call maps to those same pressure points. Bank of America pressed on the sequential deceleration and what it means for 2026; Ullal reiterated demand strength and pointed to supply as the gating factor. Wells Fargo dug into the gross-margin math and whether sub-60% product margins implied a structural change; management said it’s mix, not a new model, and highlighted manufacturing discipline. Goldman Sachs asked about the move toward full-rack solutions and how Arista will participate alongside compute; management detailed early design wins across multiple accelerator options and a blend of OEM/JDM approaches as Ethernet standards evolve. Citi asked who makes the networking decisions in those AI builds—LLM providers or cloud titans—and Ullal described a joint-decision process with stable share inside the “circle” of core cloud customers. JPMorgan queried whether any of the 100K-cluster plans had changed; Ullal again pointed to supply variability, not demand, and noted three titans already above the 100K mark with the fourth approaching.

Bottom line: this was a high-quality quarter overshadowed by timing optics. Demand is not the problem; shipping windows are. Mix will keep near-term margins choppy as AI/cloud volumes surge, but the long-term model remains compelling, and Arista’s AI fabric strategy—open Ethernet, observability/automation with CloudVision AI and AVA, and deeper rack-scale roles—puts it in the right swim lane as networks catch up to the compute build-out. For traders, $136 is the immediate pivot; for investors, the bigger pivot is 2026, when today’s supply-constrained orders are more likely to convert into a revenue curve steep enough to satisfy AI-era expectations.