Arista Networks (ANET) Q3 FY2026 Earnings Call: Q&A Summary

Q: Michael Ng from Goldman Sachs asked whether the projection of greater than 10 gigawatts of deployments across 10-plus compute accelerators for next year is an industry or Arista-specific number, and how compute diversification creates networking opportunities.

A: Jayshree Ullal stated that the gigawatt number reflects new distributed deployments and mentioned that Arista supports three types of AI fabrics across multiple XPUs.

Q: Meta Marshall from Morgan Stanley asked for an update on the mix of scale across, scale out, and scale up for the business and into 2027.

A: Jayshree Ullal explained that scale-up will be in early trials in 2027 without material numbers, leaving most use cases as scale out and scale across, with scale across expected to represent at least 30% or more.

Q: Benjamin Reitzes from Melius Research asked about the 10% customers in the fourth quarter and their use cases.

A: Jayshree Ullal stated that top customers M&M are expected to continue, and Arista hopes to add one or two more depending on shipping volumes.

Q: George Notter from Wolfe Research asked about the sequential jump in R&D expenses and whether it represents a new run rate.

A: Kenneth Duda and Chantelle Breithaupt explained that R&D increased due to hyperscaler AI system opportunities and liquid cooling technology introductions, with Chantelle adding that the framework between 9% and 11% remains applicable.

Q: Amit Daryanani from Evercore ISI asked if Arista gets involved earlier in system design as customers deploy different accelerators and whether network intensity is changing.

A: Jayshree Ullal responded that Arista's close work with accelerators like Cerebrus and Microsoft allows them to tune buffers and maximize performance early in the cycle.

Q: Amit Daryanani from Evercore ISI asked about confidence in supply chain execution to meet the 38% to 40% growth guide for 2027.

A: Chantelle Breithaupt expressed confidence based on internal execution and key supplier agreements reflected in increased purchase commitments.

Q: David Vogt from UBS asked how purchase commitments and deferred revenue relate to the 38% to 40% growth guidance for 2027.

A: Chantelle Breithaupt stated that purchase commitments increased to serve demand, including a new multiyear memory agreement, while deferred revenue remains healthy and supportive of growth.

Q: Simon Leopold from Raymond James asked for the 2026 baselines for campus and AI and about the evolution of switching with CPO or NPL.

A: Chantelle Breithaupt stated that the 2026 baseline includes AI at $3.5 billion and campus at $1.25 billion. Jayshree Ullal added that Arista remains focused on pluggable optics while pioneering open CPO for the 2028 to 2030 time frame.

Q: Joseph Cardoso from JPMorgan asked if moderation in non-AI and non-campus business is assumed in the 2027 guide.

A: Jayshree Ullal responded that all sectors are expected to grow double digits and that the guidance reflects healthy demand across all segments.

Q: Tim Long from Barclays asked what enables campus and branch growth to outpace the industry and whether it is a multiyear dynamic.

A: Kenneth Duda and Todd Nightingale stated that enterprise networks are increasingly viewed as mission-critical, and Arista's quality, reliability, and EOS patching capabilities drive market share gains in a multiyear refresh cycle.

Q: James Fish from Piper Sandler asked about visibility and confidence in the scale-up transition for 2028.

A: Chantelle Breithaupt and Jayshree Ullal noted that guidance is supported by two quarters of visibility and two years of technical conversations and co-development with partners.

Q: Tal Liani from Bank of America asked about the drivers of growth and whether deferred revenue trends are decoupling from revenue growth.

A: Jayshree Ullal and Chantelle Breithaupt explained that demand is strong across AI, enterprise, and campus, and deferred revenue is a balance sheet account that reflects new products and use cases rather than predicting quarterly revenue directly.

Q: Ryan Koontz from Needham and Company asked how liquid cooling and optics changes will impact the fulfillment model and scale-up product plans.

A: Jayshree Ullal responded that Arista works with different partners depending on the design, with orthogonal, cable backplane, and cross-rack designs each differing.

Q: Karl Ackerman from BNP Paribas asked if the 2027 revenue outlook requires new power infrastructure permitting or an aggressive drawdown of deferred revenue.

A: Chantelle Breithaupt stated that Arista does not guide deferred revenue and that the outlook is based on known demand signals and customer conversations.

Q: An analyst from Deutsche Bank asked how gross margin economics for SU-144 in 2028 compare to the existing portfolio.

A: Jayshree Ullal stated that Arista's long-term gross margin target remains in the 62% to 64% range.

Q: Daniel Zhu from Bernstein asked if scale-across customer concentration is driven by cloud titans or if smaller customers and telcos are also deploying.

A: Jayshree Ullal stated that both small use cases and large customers deploying 7800 platforms exist, with revenue concentration favoring the 7800.

Q: Sean O'Loughlin from TD Cowen asked about trends in the traditional front-end compute cycle.

A: Jayshree Ullal explained that front-end and back-end investments act like a sine wave, with current focus heavily on back-end AI fabrics while front-end will catch up later.

Q: Antoine Chkaiban from New Street Research asked how Arista plans to monetize AVA and size the security opportunity.

A: Kenneth Duda and Jayshree Ullal stated that AVA is currently in alpha trial within CloudVision and that pricing models are being developed while emphasizing Arista's strong foundational security posture.

Q: Benjamin Bollin from Cleveland Research asked about lead times across enterprise and hyperscale segments and go-to-market execution.

A: Todd Nightingale and Jayshree Ullal noted that enterprise lead times are shorter with buffered inventory, while hyperscale lead times are longer, and highlighted strong go-to-market investments and momentum.

Q: Mehdi Hosseini from Susquehanna asked if the 2027 revenue target accounts for secured silicon and memory supply.

A: Todd Nightingale stated that Arista has high confidence in supply delivery due to long-term strategic agreements covering silicon, PCB capacity, and memory.