Q3 2024 Summary
Published Feb 7, 2025, 7:58 PM UTC- Arista expects its AI center networking revenue to double its back-end target of $750 million, aiming for approximately $1.5 billion in 2025. This growth is due to increased demand in both back-end AI networking and front-end networks connecting AI clusters. ,
- Arista is progressing well with 4 out of 5 major AI customers, with 3 of them moving from trials to pilots this year and expected to become 50,000 to 100,000 GPU clusters in 2025. This indicates strong adoption of Arista's AI networking solutions among leading tech companies.
- Arista's enterprise opportunity has never been stronger, with increased traction in multiple verticals such as financials, healthcare, media, retail, and federal markets. The company is expanding its presence in campus and data center networking, leveraging its universal spine architecture and strength in wired networking.
- Arista Networks acknowledges weakness in its WiFi segment, indicating potential challenges in effectively competing in this market.
- Increasing competition from NVIDIA in Ethernet switching could impact Arista's market share, as NVIDIA is both a partner and a fierce competitor.
- Anticipated margin pressures due to customer mix may lead to future declines in gross margins, with it getting harder to maintain 10% customers.
| Metric | YoY Change | Reason |
|---|---|---|
Total Revenue | +20% from $1,509.50M to $1,810.94M | Driven by strong product demand for switching and routing platforms and continued expansion of service offerings as the installed base increased. Improvements in supply chain and early AI opportunities also contributed to higher revenue. |
Product | +19% | Reflects robust demand across Cloud, Enterprise, and Providers, along with new product introductions and higher shipments. AI-related deployments, though still early, provided incremental upside. |
Service | +28% | Stems from expansion of the installed base leading to increased support contracts and a growing contribution from subscription software. Ongoing customer renewals further boosted service revenue. |
Net Income | +37% | Lifted by revenue expansion outpacing operating expense growth, improved gross margins, and higher other income. The company also benefited from a lower effective tax rate, adding to net income gains. |
EPS (Basic) | +34% from $1.77 to $2.38 | Fueled by strong net income growth and share repurchases, which reduced outstanding shares. Improved operational efficiencies and margins also drove higher earnings per share. |
Americas | +25% | Reflects healthy enterprise demand and incremental service revenue in the region, with Cloud, Enterprise, and Provider customers contributing robustly. Improved supply chain conditions supported higher shipments. |
EMEA | +10% | Benefits from steady enterprise adoption and overall international expansion. Nonetheless, the region’s share of total revenue changed slightly as global sales mix shifted and other regions showed comparatively higher growth. |
Asia-Pacific | -8% | The decline reflects relatively weaker performance in APJ markets and a shift in sales mix toward the Americas. External macro factors and uneven enterprise spending in the region contributed to the slowdown. |
| Metric | Period | Previous Guidance | Current Guidance | Change |
|---|---|---|---|---|
Revenue | FY 2025 | no prior guidance | $8B with 15%–17% annual growth | no prior guidance |
AI Networking Revenue | FY 2025 | no prior guidance | $750M back-end; $1.5B total AI center networking | no prior guidance |
Campus Networking | FY 2025 | no prior guidance | $750M | no prior guidance |
Gross Margin | FY 2025 | no prior guidance | Expected declines due to customer mix | no prior guidance |
| Metric | Period | Guidance | Actual | Performance |
|---|---|---|---|---|
Revenue | Q3 2024 | $1.72B to $1.75B | $1.81B (1,810,936 thousands) | Beat |
Operating Margin | Q3 2024 | Approximately 44% | ~43.4% (785,250 / 1,810,936 thousands) | Missed |
| Topic | Previous Mentions | Current Period | Trend |
|---|---|---|---|
Strong enterprise networking demand and underpenetrated market opportunities | Consistently emphasized in Q2 , Q1 , and Q4. | Continued strength highlighted in verticals such as financials, healthcare, media, retail, and federal. Underpenetration remains a key focus. | Recurring across all periods with a consistently bullish viewpoint. |
AI networking expansions, revenue targets, and shift from InfiniBand to Ethernet | Q2 expansion of 800-gig products and pilot deployments. Similar targets in Q1 and Q4. | Focus on AI centers and Ethernet solutions. Targeting $8B in total revenue for 2025 with $750M from back-end AI and $1.5B from front-end AI. Gradual InfiniBand-to-Ethernet shift. | Consistent expansion with growing confidence in Ethernet-based AI networking. |
Progress with major AI customers (trials to pilots, large GPU clusters) | Q2: Four major AI trials scaling to pilots. Q1: Similar transition to larger GPU clusters. Q4: Trials to pilots with future production. | Working with five major AI customers. Three moving to pilots in 2024 with large-scale (>50k GPUs) growth in 2025; two slower. | Ongoing progress as customers ramp up GPU deployments. |
Margin pressures from large cloud customer mix | Q2 margins guided conservatively on expected cloud mix ; minimal direct mention in Q1 ; Q4 noted higher cloud mix lowers margins. | Non-GAAP gross margin of 64.6% influenced by pricing pressure from cloud titans, partially offset by enterprise margins. | Recurring challenge linked to customer mix. |
Slower growth in core (non-AI) business segments | Q2 noted a pivot to AI spending but core refresh still relevant. No explicit mention in Q1 or Q4. | Expected single-digit growth in non-AI markets; faster AI and campus growth. | Emerging focus on a shift toward AI, impacting core growth. |
Campus networking expansions and universal spine architecture | Q2 highlighted leaf/spine for campus modernization ; Q1 underscored universal spine’s broad applicability ; no direct Q4 mention. | Strong traction in enterprise verticals, targeting $750M campus revenue by 2025; emphasis on universal spine. | Steady expansion as a critical growth vector. |
Increasing software and services revenue with security and observability offerings | Q2 saw UNO for observability ; Q1 introduced CloudVision UNO ; Q4 noted subscription-based software at ~16% of revenue. | No mention this period. | Mentioned previously, not discussed in Q3. |
Deferred revenue growth and unpredictable revenue recognition | Q2 reported $2.1B deferred revenue with unpredictable recognition ; Q1 and Q4 also noted variable timelines. | Mentioned significant increase in product deferred revenue but with longer timeframes for recognition. | Consistently noted; timing remains difficult to forecast. |
Competition from NVIDIA's in-network computing solutions | Q2 sees NVIDIA Spectrum switch in limited scenarios ; Q1 sees synergy with NVIDIA GPUs but limited Ethernet overlap ; Q4 references ongoing collaboration. | Acknowledged NVIDIA as both partner and competitor. Rarely see its Ethernet solutions; confident in Arista’s Ethernet expertise. | Ongoing competitive dynamic with limited direct Ethernet overlap. |
Conservative guidance implying potential upside | Q2 and Q1 guidance historically cautious but raised with each quarter ; Q4 also reflected a measured approach. | Visibility ~6 months; conservative forecasts may underestimate faster AI pilot acceleration. | Repeated pattern of conservative outlook that could beat estimates. |
Impact of data center refresh cycles on demand | Q2 highlighted speed upgrades (100G→200G/400G) ; Q1 noted 3–5-year server refresh cadence ; Q4 mentioned correlation with GPU/CPU availability. | No direct mention of refresh cycles in Q3. Some references to cloud infrastructure updates. | Recurring theme previously, less explicit in Q3. |
-
AI Trials Progress
Q: How are the five major AI trials progressing and what scale is expected?
A: Arista now believes they are 5 out of 5 in AI trials, progressing well in 4 out of the 5 clusters. Three customers are moving from trials to pilots this year, expecting to become 50,000 to 100,000 GPU clusters in 2025. The fifth customer is in trial mode and moving slower than expected due to challenges like awaiting new GPUs and power cooling issues. Overall, they feel good about their guidance for 2025. -
Competition with NVIDIA
Q: Are you seeing increased competition from NVIDIA in data center switching?
A: Arista views NVIDIA as both a good partner and a fierce competitor. While NVIDIA has increased its market share, Arista doesn't frequently encounter their Ethernet capabilities, perhaps at 1 or 2 customers. Arista is considered the expert in Ethernet switching, especially in large scale-out networking and smaller enterprises. They welcome competition with NVIDIA on Ethernet switching while partnering on GPU connectivity. -
Competitive Landscape with Cisco
Q: How is the competitive landscape changing with AI, including competitors like Cisco and Juniper?
A: In the back end, Arista is gaining credibility in connecting to GPUs but doesn't claim market leadership; any share they get, including the $750 million, is incremental. In the front end, Arista is viewed as the gold standard, with a superior portfolio and competitive differentiation. They feel extremely bullish with their flagship 7800 product, the newly introduced 7700, and the 7060 series. -
Revenue Guidance and Non-AI Growth
Q: Why does guidance imply slower growth in non-AI, non-campus business next year?
A: Visibility only extends to about six months, so they are cautious in guidance. Large cloud customers are continuing to refresh but are pivoting aggressively to AI. They expect to grow faster in AI and campus, and slower in classic markets like data center and cloud. -
Gross Margin Expectations
Q: What's underlying the expectations of gross margin declines in the 2025 outlook?
A: The outlook reflects customer mix. They expect supply chain discipline to continue. As the denominator gets bigger, having multiple 10%+ customers gets tougher. They believe existing major customers will remain significant but don't anticipate new ones at that level in the near term. -
Deferred Revenue Increase
Q: Why is product deferred revenue increasing significantly while revenue guidance decelerates?
A: The increase in deferred revenue is due to the type of use case, customers, and product mix, which have bespoke time frames and can take longer to manifest, sometimes multiple years. It may not all happen in 2025. -
Adoption of 400G vs. 800G Products
Q: Are hyperscalers deploying 800G switches or staying with 400G?
A: Majority of trials and pilots are on 400G because the ecosystem for 800G is still developing, including NICs and packet spraying capabilities. While there are early trials on 800G, the majority of 2024 will be at 400G. They expect a better split between 400G and 800G as they go into 2025. -
Expansion into Campus and Verticals
Q: Where are you seeing traction in campus opportunities and which verticals?
A: Arista's enterprise opportunity has never been stronger, with traction in data center, campus center, WAN center, and some AI. They're seeing strength in verticals like financials, healthcare, media, retail, and federal markets. Activity with their universal spine has started, and they're focusing on wired and spine, aiming to improve in WiFi. -
Additional AI Opportunities
Q: Are there opportunities beyond the five main AI trials, including Tier 2 or sovereigns?
A: Yes, there are additional trials, but the five main ones can go to 100,000 GPUs and more, significantly impacting numbers. They have 10 to 15 trials in classic enterprises with smaller GPU counts. There will be more opportunities in both Tier 1 and Tier 2 categories. -
Ethernet Portfolio Opportunities
Q: Could you comment on opportunities across the three Ethernet families?
A: The fixed 7060 switches are very popular in terms of units. The 7800 is strategic in dollars, serving as a flagship product. The 7700 offers capabilities of the 7800 in a mini configuration up to 10,000 GPUs. Customers are optimizing a mix of these products based on deployment sizes and GPU counts.