Arista Networks, Inc. Q3 FY2024 earnings call

• Arista 2.0 strategy: Networks are at the epicenter of mission-critical transactions. Arista's modern networking platforms, including the state-oriented public subscribed network data lake EOS software stack, guide the company with principles like best-in-class proactive products, zero-touch automation, and prescriptive insights for AI. • AI networking: Networking for AI is gaining traction, moving from trials in 2023 to more pilots in 2024, with expectations of more production in 2025 and 2026. Arista's Ethernet portfolio for AI includes 3 families with over 20 switching products. • Platform and supply chain: Arista 7700 R4 distributed Ecolink switch, 7800 R4 Spine, and 7060 X6 AI leaf entered production. Supply chain team increased purchase commitments for advanced semiconductors and on-hand inventory to address AI network deployment, with lead times for advanced semiconductors remaining extended but supply chain returning to predictability post-pandemic.

Revenues for the third quarter of 2024 were $1.81 billion. Services and software support renewals contributed approximately 17.6% of revenue. International revenues for the quarter were $330.9 million, representing 18.3% of total revenue, while the Americas accounted for 82%. Non-GAAP gross margin in Q3 was 64.6%, influenced by cloud titan customer pricing pressure offset by favorable enterprise margin and supply chain hygiene. Total revenues showed a 20% year-over-year increase, exceeding the guidance range of $1.72 billion to $1.75 billion.

• Q4 2024: Revenues expected $1.85 billion to $1.9 billion, gross margin 63% to 64%, operating margin ~44%. Effective tax rate expected ~21.5% with diluted shares ~321 million. Anticipate increased working capital requirements in Q4 due to inventory build for AI network deployment. • FY 2025: Projected revenue growth 15% to 17%, gross margin 60% to 62%, operating margin ~43% to 44%. Focus on investing in R&D, go-to-market, and scaling the company, aiming for ~$8 billion in revenue in 2025. • Stock split: Board approved 4-for-1 stock split to enhance stock accessibility.

• Supply chain constraints: Lead times for advanced semiconductors remain extended from pre-pandemic levels. • Component costs and manufacturing output: Uncertainties in component costs and manufacturing output affecting inventory management. • Inflationary pressures: Impact on business operations and financials. • Inventory management: Variability in purchase commitments and inventory levels based on demand for new product introductions.

Q: Update on campus opportunities and applications?

A: Arista's enterprise opportunity is strong. Strength in spine and wired, with progress on wired and plans to improve WiFi. Vertical sectors seeing strength include financials, healthcare, media, retail, Fed, and sled. Focus on federal market with new subsidiary setup.

Q: Deferred revenue acceleration and revenue deceleration in 2025?

A: Chantelle Breithaupt explained it's due to bespoke time frames of use cases, customers, and product mixes, with variables in deferred revenue and product deferred revenue, and lengthening of trials' time frames.

Q: Tier 2 opportunities and AI trials?

A: There are more than 5 trials, with the 5 main trials being largest AI titans spread across Tier 1 and Tier 2. Classic enterprise has smaller GPU counts in trials but more are ongoing.

Q: Ethernet portfolio and 800-gig ports in pilots?

A: Majority of trials and pilots in 2024 are on 400 gig due to ecosystem waiting for 800 gig, but expecting better split between 400 and 800 gig in 2025.

Q: Margin declines in 2025 outlook?

A: Chantelle Breithaupt attributed it to customer mix, with supply chain discipline continuing, and the denominator getting bigger making it tougher to maintain margins, but expecting to see how it plays out in 2025 and beyond.

Q: Lift from AI on front end and bandwidth pressure?

A: Jayshree Ullal stated for every dollar spent on back end, there can be 30% to 200% spent on front end, with the average being 100%, making it difficult to track pure AI as it includes various elements like inference, training, and classic cloud.