I'll research ANET before writing the preview. Let me gather recent filings, transcripts, news, and price data.Let me read the latest earnings materials and get price data.Let me get recent macro/sector context on hyperscaler capex and any ANET-specific news around the report.# Arista Networks (ANET) — Q2 Earnings Preview
Arista enters this print in an unusual position for a hyper-growth AI-networking name: the constraint is no longer demand — it's the company's ability to source parts. On the prior call, CEO Jayshree Ullal described current demand as "the best I've ever seen in my Arista tenure," while flagging that supply is "a slightly different and opposite tail," citing industry-wide shortages across wafers, silicon, CPUs, optics and especially memory. Management now frames this as a 1–2 year phenomenon, not 1–2 quarters, with 52-week lead times on chips and multi-year purchase commitments being signed to lock in capacity.
The key tension for investors: Arista is deliberately paying up to secure supply and largely eating the cost rather than passing it fully to customers, which pressures gross margin but protects share and customer relationships ("we don't want to keep GPUs idle… because Arista didn't supply the network"). Watch for any color on de-commits from suppliers, which management said "doesn't feel good" and could cap 2026/2027 growth.
Guidance for the quarter being reported (Q2): - Revenue ~$2.8B - Non-GAAP gross margin 62–63% - Non-GAAP operating margin 46–47% - Non-GAAP diluted EPS ~$0.88 (~1.27B diluted shares, ~21.5% tax rate)
Last quarter (Q1) — a strong beat/setup: - Revenue $2.709B, +35.1% y/y and +8.9% q/q, well above the $2.6B guide - Non-GAAP EPS $0.87 (+32% y/y); GAAP EPS $0.80 - Non-GAAP gross margin 62.4% (down from 64.1% a year ago — mix + supply costs) - Non-GAAP operating margin 47.8% - Record operating cash flow of $1.69B; cash & investments ~$12.35B
Arista has a long history of guiding conservatively and beating, so the bar tomorrow is really the FY raise and the AI/margin commentary, not the $2.8B headline.
Management raised FY2026 twice already, most recently to ~27.7% growth (~$11.5B), up from an initial ~20% and then ~25%. Within that: - AI fabric target raised to $3.5B (from $3.25B), more than doubling AI sales y/y - Campus target maintained at $1.25B - FY gross margin range held at 62–64%; operating margin ~46%
What to watch: another FY raise and/or a further bump to the AI target would be the clearest bullish signal. Conversely, management has repeatedly hedged that the raise is now as much about how much they can ship as about demand — so a guide that holds rather than climbs may be read as supply-capped rather than demand-soft.
Ullal's "three fabrics" narrative is now the organizing lens: - Scale-out (horizontal leaf-spine): Arista's heritage; 100+ cumulative customers on 800G; 1.6T at production scale expected in 2027. - Scale-across (interconnecting distributed data centers): the emerging differentiator, anchored by the high-end 7800R series. Management expects it to be at least ~1/3 of the AI number and potentially bigger — a high-margin routing-led opportunity that plays to Arista's strengths. - Scale-up (in-rack, vs. NVLink): essentially $0 in 2026, a 2027–2028 story tied to the ESUN Ethernet standard and 1.6T chips; 5–7 rack designs in active engineering.
Also listen for: diverse accelerator attach (explicit nods to AMD MI-series and Google TPUs), neocloud and sovereign cloud momentum (management thinks neocloud is underappreciated), and traction on the new XPO liquid-cooled optics MSA (100+ endorsers; pitched as the next OSFP for 1.6T/3.2T).
Microsoft and Meta remain the anchor >10% customers. The key incremental catalyst: management reiterated it expects one, possibly two new 10% customers in 2026 on demand — but caveated that actual recognition depends on shipments. Confirmation (or naming) of a new whale would be a meaningful positive. Note that large-customer mix carries lower gross margin, so faster titan/AI growth is a structural margin headwind.
Q1 non-GAAP GM of 62.4% sat at the low end and is down ~170bps y/y. Management attributes the pressure primarily to customer mix (large cloud/AI customers dilute margin) with a secondary hit from memory/silicon costs and tariffs. The bull case for stabilization/upside is a favorable enterprise/campus mix shift in 2H and price increases that flow through only as backlog burns down. Any GM guide below the 62–63% range, or commentary that supply costs are worsening, would be the most likely negative surprise.
Total deferred revenue jumped to $6.2B (from $5.37B), with product deferred up ~$643M q/q, driven by customer-specific acceptance clauses on new AI products. Qualification cycles have stretched from 2–4 quarters to 6–8 quarters. Management repeatedly warns that deferred can swing sharply quarter-to-quarter independent of underlying demand, and that acceptance timing "can create quarterly and sequential dynamics that do not follow prior-year trends." Translation: don't panic on a soft sequential quarter or a big deferred build — it's timing, not demand.
Related balance-sheet items to track: purchase commitments ballooned to $8.9B (from $6.8B) and inventory to $2.38B — both intentional bets to secure supply, but they warrant monitoring for any demand-air-pocket risk down the line.
Notably, Arista did not repurchase stock in Q1, leaving ~$818M of the $1.5B authorization. A resumption of buybacks — given the ~$12.35B cash pile and record cash flow — would be a modest positive; continued abstention may raise questions about valuation discipline or capital preservation for supply commitments.
ANET closed around $120 heading into the print, having staged a dramatic recovery from ~$64 April tariff-driven lows (~+85%), and roughly flat-to-modestly-up on the year. It slipped a few percent in the days just before the report. The setup is a stock that has re-rated hard off the lows into a high bar, moving broadly in sympathy with NVIDIA and the AI-capex complex — so the read-through from hyperscaler capex commentary (Microsoft/Meta/Google) matters as much as Arista's own guide.
One sentiment flag: a cluster of insider "intention to sell" (Form 144) filings in late July/early August from CEO Jayshree Ullal, CTO Ken Duda (a sizable ~30,000-share planned sale), and director Charles Giancarlo. These appear to be pre-planned/programmatic sales rather than fundamental signals, but they add to the "high bar into the print" narrative.
| Bullish tells | Bearish tells |
|---|---|
| Another FY2026 and/or AI-target raise above $11.5B / $3.5B | FY guide merely reaffirmed (read as supply-capped) |
| Confirmation of a new 10%+ customer | Gross margin guided below 62–63% / worsening supply costs |
| Gross margin holding or mix-driven upside | Rising supplier de-commits / lengthening lead times |
| Buyback resumption | Weak sequential revenue + large deferred build spooking the tape |
| Strong scale-across / 1.6T / 2027 setup commentary | Cautious hyperscaler capex read-through |
The most important thing to remember: with demand explicitly outstripping supply, the debate is shifting from "is there enough AI demand?" to "how much can Arista physically ship, and at what margin?" Guidance direction and gross-margin trajectory — not the in-line-looking $2.8B/$0.88 headline — will drive the reaction.
Note: figures above are drawn from Arista's most recently reported quarter and its official guidance; consensus estimates and any last-minute pre-announcements should be checked against the tape before the print.