For the past eighteen months, the single most persistent institutional concern about Arista Networks has not been growth. It has been dependence. Microsoft and Meta together account for more than 20% of Arista’s revenue, and any portfolio manager running a risk-attribution model has flagged that number. Tuesday evening, CEO Jayshree Ullal gave that concern a direct answer.
The Big Question
Arista’s Q2 2026 numbers were not close. Arista posted EPS of $1.02 versus the $0.89 consensus estimate, a beat of 15.2%, while revenue came in at $3.04 billion against the $2.83 billion estimate. The results extended the company’s consensus EPS beat streak to five consecutive quarters. That is not the debate. The debate is what comes after the beat, and who is buying.
During the call, Ullal suggested that beyond its two existing customers that generate 10% or more of revenue, Microsoft and Meta Platforms, Arista expects to add “at least one, maybe two” new 10%-plus customers before the end of 2026. One new 10%-plus customer at Arista’s current revenue run rate represents roughly $1.2 billion in annualized spend. Two would restructure the company’s customer map entirely.
Why Wall Street Cares
Customer concentration is not just a talking point in the investment committee room. It directly affects how institutions model downside scenarios. A single delayed hyperscaler buildout cycle, a sourcing change from one of the two anchor customers, or a competitive win by Cisco or Nvidia’s Spectrum-X platform can knock 10% to 15% off Arista’s forward revenue overnight.
New entrants at the 10% threshold are increasingly valuing the EOS software stack for its reliability and observability over traditional white-box or ODM solutions. That matters. It is not just volume migrating to Arista. It is customers making a deliberate architectural choice that is expensive to reverse.
A fourth major customer officially transitioned from InfiniBand to Ethernet-based AI fabric in Q1 2026. The Q2 call confirmed that migration is accelerating. The customer diversification thesis is becoming real rather than aspirational.
The Bull Case
Revenue of $3.036 billion represented an increase of 12.1% compared to the first quarter of 2026, and an increase of 37.7% from the second quarter of 2025. GAAP and non-GAAP operating margins came in at 45.4% and 49.9%, respectively, compared to 44.7% and 48.8% in the second quarter of 2025. Margins are expanding even as the company scales rapidly, a combination that rarely lasts long enough for most technology companies to exploit.
Forward guidance was equally decisive. For Q3 2026, Arista forecast revenue of approximately $3.3 billion, well above the consensus estimate of $2.95 billion. Adjusted EPS guidance of $1.06 to $1.08 also topped the analyst consensus of $0.92, with the midpoint representing a 16% premium to expectations. Arista also raised its full-year 2026 revenue view to $12.6 billion, against a prior consensus of $11.62 billion.
The product cycle is timed well. Arista introduced the 7060XE7 Series, a next-generation 1.6-terabit portfolio for AI fabrics. The new Etherlink platforms deliver up to 100 Tbps of system bandwidth and support Linear Pluggable Optics, cutting interconnect power consumption by approximately 60% compared to traditional pluggable optics. The 7060XE7 has already been adopted by Meta, Microsoft, and Oracle in commercial deployments.
The Bear Case
Three months ago, this same earnings season looked very different. Despite reporting better-than-expected Q1 2026 revenue of $2.71 billion and EPS of $0.87, shares tumbled 13.6% the following day as investors focused on worsening supply constraints and delayed revenue recognition. The company cited wafer fab shortages, semiconductor supply de-commits, and extended lead times beyond 52 weeks as factors limiting its ability to meet strong AI-driven demand, while deferred product revenue jumped by approximately $643 million to $3.63 billion.
That deferred revenue pile is a two-edged reality. On one hand it represents contracted future demand. On the other it means Arista has been collecting orders it cannot yet ship. If supply chain resolution proves slower than management has telegraphed, the mismatch between bookings velocity and actual delivery could compress recognized revenue just as the Street has lifted its estimates.
Greater focus on AI networking ties Arista more closely to a small set of hyperscale customers, so any delay or change in their buildout plans could affect future revenue. Competition from large rivals such as Cisco, Broadcom, and Nvidia, including Ethernet and alternative fabrics for AI clusters, may pressure pricing and share in high-end deployments.
The stock currently sells for 65 times earnings and 43 times next year’s expected earnings. At those multiples, the market is pricing continued execution with little tolerance for a single quarter of guidance disappointment.
The Evidence
The supply chain narrative that crushed the stock in May appears to be resolving faster than feared. KeyBanc viewed earlier supply and deferred revenue issues as largely temporary even before Tuesday’s results. The Q2 beat and the sharply higher Q3 guidance suggest that at least some of the deferred backlog is converting to recognized revenue on schedule.
The total addressable market argument has materially strengthened. The total addressable market has expanded from $50 billion to $105 billion, with projections of $160 billion by 2030. The AI networking cycle is expected to be multi-year or even multi-decade, with capital expenditure growth driven by both hyperscalers and emerging cloud providers.
Financially, the fortress balance sheet gives Arista room to navigate supply disruptions that would damage a more leveraged competitor. With roughly $12 billion of cash and investments and no debt, the company maintains a fortress balance sheet supporting its premium valuation and financial flexibility.
The Mavens’ View
Professional investors are not debating whether Arista is a good business. They are debating whether it is priced as though the next three years of growth are already in the stock. The answer depends almost entirely on whether the new-customer signal Ullal telegraphed Tuesday evening converts into disclosed revenue before year-end.
The shift from proprietary back-end technologies to Ethernet-based AI fabrics is redefining data center architectures and driving growth. Arista is the primary beneficiary of that architectural shift, and it is not a close race. Management reported achieving the number one market share in high-speed switching in the greater than 10 gigabit Ethernet category, overtaking several incumbent vendors based on 2025 industry data.
The institutional position is: own the business, watch the multiple. Most buy-side teams that have modeled the new-customer disclosure see it as the unlock that justifies the current valuation. If two new 10%-plus customers appear by December, the concentration risk that has capped the stock’s multiple for more than a year disappears from the risk model.
What Investors Are Missing
The conversation has centered on whether Arista can sustain hyperscaler spending tailwinds. The overlooked question is what happens to Arista’s pricing power when its customer base widens. Today, Microsoft and Meta have leverage. Each represents enough of Arista’s revenue to negotiate hard on price. Market expansion beyond hyperscalers includes tier 2 clouds and enterprises. AI installations are seeing a shift toward branded vendors due to reliability and mission-critical requirements. That shift reduces buyer leverage. A more fragmented customer base, paradoxically, supports higher average selling prices rather than threatening them.
There is a second implication almost nobody is discussing. Arista detailed new innovations for scale-up, scale-out, and scale-across AI fabrics, outlining techniques including multi-planar leaf-spine designs, the open Multipath Reliable Connection protocol, and segment routing over IPv6, aimed at maximizing AI cluster utilization and resilience. These are not incremental product updates. They are infrastructure decisions that lock a customer into Arista’s EOS software stack for five to seven years. Every new 10%-plus customer is not just an incremental revenue line. It is an annuity.
Stocks to Watch
Arista Networks (ANET). The direct read-through from Tuesday’s results. CFO Chantelle Breithaupt noted that second-quarter revenue increased 37.7% while adjusted EPS grew 39.7% compared to the prior year period. The new-customer signal is the single most important disclosure Arista has made in several quarters. Watch for any identification of the new customers on the Q3 call.
Cisco Systems (CSCO). The most direct competitive threat to Arista and the clearest loser if Arista continues to win enterprise and campus deployments. Arista was named a Leader in the 2026 Gartner Magic Quadrant for Enterprise Wired and Wireless LAN, territory Cisco has historically dominated. Each point of campus share Arista captures comes directly from Cisco’s installed base.
Broadcom (AVGO). Arista’s 7060XE7 Series runs on Broadcom Tomahawk 6 silicon, making Broadcom a quiet beneficiary of every Arista rack that ships. Arista’s capacity constraints and growing backlog translate directly into Tomahawk 6 demand. The relationship is less visible than the Nvidia GPU story and is priced accordingly.
Meta Platforms (META) and Microsoft (MSFT). Both are confirmed existing anchor customers. Both are committing to multi-year AI infrastructure buildouts that run through Arista switches. Tuesday’s guidance implies each will continue spending aggressively. Their capex trajectories are now a secondary indicator for Arista’s revenue visibility. When Microsoft or Meta raises capital expenditure guidance, Arista’s forward model improves in real time.
