The earnings number isn't the story. The supply chain is.
Arista Networks posted Q2 adjusted EPS of $1.02 against a consensus estimate of roughly $0.89 - a 15% miss in the other direction, which in analyst-speak means a decisive beat. Revenue also exceeded consensus estimates. The stock responded by sprinting up 12.3% over the week, touching a new 52-week high of $194.35 and closing at $190.51 on $2.7 billion in turnover. Year-to-date, ANETANET-- is up 45.4%.
The headline treatment calls this a trounce. The pattern is the exact inverse of last quarter: in May, AristaANET-- beat Q1 EPS ($0.87 vs. $0.81 consensus) and revenue ($2.71 billion vs. $2.62 billion expected), and the stock collapsed 13.6% the next day. Same playbook - earnings above estimates, revenue above estimates - opposite market reaction.
Quick Backtesting Tool
What changed between Q1's dump and Q2's surge is not the product. It's the supply chain narrative.
The Q1 Hangover
In Q1, Arista delivered what should have been a celebration. Revenue grew 35.1% year over year. And yet the stock cratered.
The reason was supply. Management flagged wafer fab shortages, semiconductor supply de-commits, and lead times stretching beyond 52 weeks. Deferred product revenue - money already collected but not yet recognized because the hardware hadn't shipped - ballooned to $3.63 billion, up approximately $643 million in one quarter. That deferred revenue build was the market's fear: Arista was collecting cash but couldn't ship enough product to recognize it. The AI revenue target had been raised to $3.5 billion, which looked aggressive when supply couldn't keep pace with demand.
Any astute infrastructure investor understands the dynamic. Deferred revenue is a double-edged sword. It proves demand is real - customers are prepaying. But it also means revenue recognition is bottlenecked, not by sales, but by the factory. If supply constraints persist, those deferred dollars sit on the balance sheet indefinitely, and gross margins take a hit from the mix of delayed higher-margin AI products.
Q2: The Beat Has Two Layers
The $1.02 EPS print versus $0.89 consensus is not a rounding-error beat. For a company with a 42.8% operating margin, that's meaningful - it signals either more volume than expected, better margins than expected, or both.
The TTM picture confirms the quality. Gross margin stands at 63.5%, operating margin at 42.8%, and free cash flow margin at an extraordinary 54.4%. Return on invested capital is 28.0%. Trailing free cash flow is $5.28 billion, growing 39.4% year over year. Capex is trivial - $146 million TTM - because Arista is a software-optimized switch designer, not a foundry. The capital structure is clean: $2.79 billion in cash, $8.17 billion in total debt, a current ratio of 283%, and a net debt position that the market reads as functionally neutral.
This is one of the most efficient capital structures in all of hardware. The question isn't whether Arista prints money. It's whether the supply bottleneck that threatened to derail the AI revenue pipeline has actually eased.
Here's the data gap: without the full Q2 earnings release, it's impossible to say whether the Q2 revenue beat came from newly resolved supply constraints or simply from converting that $3.63 billion deferred revenue backlog into recognized sales. If it's the latter - a backlog drawdown - the next quarter could look weaker unless supply has genuinely improved. If supply has improved, the trajectory steepens.
That distinction matters because Arista's $3.5 billion AI revenue target for fiscal 2026 now rests on three quarters to deliver a significant step-up in AI sales. The step function is real. Supply is the only thing that can stop it.
The Cisco Comparison Nobody Is Making
Cisco is up 5.1% on the same trading day, riding the same AI infrastructure wave at a $480 billion market cap and 40.1x trailing P/E. Cisco's quarterly revenue growth sits at roughly 12% year over year - respectable for a company of its size, but a fraction of Arista's 30.6% TTM growth rate.
The contrast is structural, not cyclical. Cisco is a diversified networking giant spanning security, collaboration, and observability, and its AI infrastructure orders are accelerating but from a smaller base share. Arista is pure high-performance data center switching, and its EOS software stack has become the default operating system for large cloud operators. The switching cost - literally - is enormous. Cloud providers that have standardized on EOS across thousands of racks aren't switching for incremental price improvements.
But Cisco's size is also Cisco's insulation. Arista's focus is its vulnerability.
Customer Concentration Is The Shadow Risk
Two customers each drive 16% to 20% of Arista's revenue. Nearly half of the company's annual revenue flows from a handful of cloud and AI titans. That concentration is the engine of the growth story and the single biggest risk.
Hyperscaler capex cycles are lumpy. When the big spenders pull back - whether for margin pressure, model efficiency gains, or a simple rebalancing of build-vs-buy decisions - Arista takes the hit first and hardest. There's no enterprise diversification cushion like Cisco enjoys. The 42.8% operating margin is impressive, but it's built on a revenue base that a single customer's guidance cut could disrupt.
That said, the current AI buildout is not a discretionary expense for these companies. It's existential. The concentration risk is real but deferred - it matters more when the AI capex growth rate decelerates, which is not this quarter or likely not the next. For now, concentration is working in Arista's favor, feeding the revenue pipeline and driving the backlog.
The Cross-Currents
Supply chain resolution: Likely positive. The magnitude of the Q2 EPS beat suggests either volume acceleration or margin recovery - both of which point to supply easing. But without the deferred revenue number for Q2, the full picture is incomplete. Directionally bullish, but the evidence is partial.
AI revenue target execution: On track but unproven at the required pace. The $3.5 billion target demands a step function in quarterly AI revenue that requires sustained supply delivery. One quarter of beats doesn't confirm the trajectory; three does. Directionally constructive, needs one more print.
Valuation at all-time highs: Stretched but earned. At $190, the stock is near its 52-week high of $194.35. The 45.4% YTD run has priced in continued execution. The upside from here requires another beat, not just meeting expectations. Neutral - the price reflects competence, which is exactly what Arista has been delivering.
Customer concentration: Managed risk, real tailwind. Two customers at 16-20% each are a structural risk that becomes a structural threat when hyperscaler capex decelerates. That's not the current environment. Directionally positive for the next 6–12 months, negative beyond that without diversification.
The Judgment
Arista's Q2 results are legitimate. The EPS beat is large enough to rule out accounting noise, and the stock's move toward all-time highs reflects a genuine shift in the supply chain narrative that sank the stock in Q1. The deferred revenue backlog from Q1 has likely converted, and the earnings print suggests supply is no longer the binding constraint it was three months ago.
But the market is now pricing Arista as if supply will never be the binding constraint again. That's the gap between the Q2 beat and the Q2 narrative.
The investor implication is straightforward: Arista remains the highest-quality networking name in AI infrastructure, with margins that hardware companies of its growth rate shouldn't be able to sustain. The stock is not cheap at all-time highs. It is not unjustified either - the 54.4% FCF margin and 28% ROIC earn the premium. But the next catalyst has to be another supply chain confirmation, not just another earnings beat on a high base. The bar for the stock to move higher from here is execution, not surprise.
The real question for the November earnings print - when Arista reports Q3 - is whether the $3.63 billion deferred revenue number has come down meaningfully. If it has, supply is solved and the $3.5 billion AI target is on rails. If it hasn't, the Q2 beat was a backlog release, not a supply fix, and the stock has run ahead of the operational reality.
You decide which was marketing fluff and which one was analysis.
