The Equation Series
Menu

Artificial Intelligence · Profile from AI Research Companion

Arista Networks

The Santa Clara networking company, founded in 2004 by Sun Microsystems co-founder Andy Bechtolsheim, Stanford professor David Cheriton and engineer Ken Duda, and run since 2008 by former Cisco executive Jayshree Ullal, that took the high-speed data-centre switching market from Cisco by selling fast, software-defined Ethernet switches to the largest cloud companies — and then became the main independent supplier of Ethernet networks for AI clusters, the open alternative to Nvidia's InfiniBand. Revenue grew from ~$584m (2014) to $9.006bn in 2025 (+28.6%, company), with Microsoft and Meta historically its two largest customers; the "cloud and AI titans" were 48% of 2025 revenue (company). Arista raised its 2026 outlook to ~$12.6bn (~40% growth) and its 2026 AI-networking goal to at least ~$3.5bn (August 2026, reported). It operates at ~64% gross and ~48% operating margins (non-GAAP, company), unusually high for hardware, because the value it sells is mostly software (its EOS operating system) running on Broadcom merchant silicon.

At a glance

Founded
2004 as Arastra, Menlo Park/Palo Alto, by Andy Bechtolsheim, David Cheriton and Ken Duda; renamed Arista Networks 2008
Headquarters
Santa Clara, California
Ownership / structure
Public company, NYSE: ANET (IPO June 6, 2014); S&P 500 member; 4-for-1 stock split December 2024. Bechtolsheim and Ullal are large individual holders (verify)
Leadership (2026)
Jayshree Ullal (chair and CEO, verify chair title); Andy Bechtolsheim (co-founder, chief architect and chairman, verify); Ken Duda (co-founder, CTO); Todd Nightingale (president and COO, from 2025, verify); Chantelle Breithaupt (CFO, from 2024)
AI products
Etherlink AI switches (7060X, 7700R4 distributed Etherlink switch, 7800R4 AI spine) on Broadcom Tomahawk and Jericho chips; EOS operating system; CloudVision management; AI back-end (scale-out) and front-end networks; scale-up Ethernet (2026); VeloCloud SD-WAN (from 2025)
Scale
Revenue ~$584m (2014), ~$2.4bn (2019), ~$4.4bn (2022), ~$5.9bn (2023), ~$7.0bn (2024), $9.006bn (2025); Q2 2026 $3.04bn (+37.7%); 2026 outlook ~$12.6bn (August 2026); Etherlink >100 cumulative customers (2026, company)
Capital raised
Founder-funded (Bechtolsheim and Cheriton); IPO ~$226m at $43 (June 2014, reported); no further equity needed; large buybacks since
Key partnerships
Broadcom (merchant switch silicon); Microsoft and Meta (largest customers, co-design); Ultra Ethernet Consortium (founding member, July 2023); AMD and others in UEC/UALink-era ecosystems
Signature moves (incl. failures)
Founding and EOS (2004–08); the NYSE IPO (2014); the Cisco patent war and $400m settlement (2014–18); the cloud-titan concentration bet (2017–); AI back-end Ethernet and the Ultra Ethernet Consortium (2023); the VeloCloud acquisition (2025); the AI revenue targets and the $9bn year (2025); scale-up Ethernet and the $12.6bn outlook (2026)

8 strategic moves analysed

Each one gets a full strategy file in the companion: context, the decision, the financing, what happened, and the lesson.

  1. 2004

    Founding Arastra and EOS

    Founding

  2. 2014

    Cisco Patent War and the 400m Settlement

    Lawsuit or settlement

  3. 2014

    NYSE IPO

    IPO

  4. 2017

    Cloud Titans and the Concentration Bet

    Go-to-market strategy

  5. 2023

    AI Back End Ethernet and the Ultra Ethernet Consortium

    Launch / partnership (industry standard)

  6. 2025

    AI Revenue Targets and the 9bn Year

    Growth strategy / public targets

  7. 2025

    VeloCloud Acquisition

    Acquisition

  8. 2026

    Scale Up Ethernet and the 12 6bn Outlook

    Launch / guidance raise