Prepared Highlights
- Q4营收2.706亿美元，环比增长17%，同比增长92% (CEO)(CFO)
- 2025全年营收8.525亿美元，同比增长115% (CEO)
- Scorpio P系列全年营收占比超10%，为市场唯一量产PCIe 6 fabric (CEO)
- Ares产品线2025年同比增长近70% (CEO)
- Taurus 2025年营收同比增长超4倍 (CEO)
- Q4 non-GAAP毛利率75.7%，环比下降70个基点 (CFO)
- Q4 non-GAAP运营费用9600万美元 (CFO)
- Q4 non-GAAP运营利润率40.2%，环比下降150个基点 (CFO)
- Q4 non-GAAP每股摊薄收益0.58美元 (CFO)
- Q1 2026营收指引2.86亿至2.97亿美元，环比增长6%至10% (CFO)
- Q1 2026 non-GAAP毛利率约74% (CFO)
- Q1 2026 non-GAAP运营费用1.12亿至1.18亿美元 (CFO)
官方指引：Q1 2026营收2.86亿-2.97亿美元；non-GAAP每股收益0.53-0.54美元 (CFO)
管理层引用：
- “Scorpio P系列全年营收占比超过10%，仍是市场唯一量产的PCIe 6 fabric。” (CEO)
- “我们预计2026年Scorpio P系列将继续增长，并开始向至少两家新 hyperscaler 出货。” (CEO)
- “Taurus 2025年营收同比增长超过四倍。” (CEO)
- “Q1营收预计环比增长6%至10%。” (CFO)
- “我们正战略性投资团队扩张以抓住AI连接机遇。” (CEO)
Prepared Metrics
| 指标 | 数值 | 发言人/背景 |
|---|
| Q4营收 | 2.706亿美元 | CEO/CFO |
| 2025全年营收 | 8.525亿美元 | CEO |
| Q4 non-GAAP毛利率 | 75.7% | CFO |
| Q4 non-GAAP运营费用 | 9600万美元 | CFO |
| Q4 non-GAAP每股收益 | 0.58美元 | CFO |
| Q1营收指引 | 2.86亿-2.97亿美元 | CFO |
Q&A Batch (1-5 of 5)
Q1 — Blayne Curtis
- Topic: UA Link adoption, NVLink fusion opportunity, and optical timing
- Key points:
- AWS announced Phranyon four (ramp in 2027) will support UA Link; AMD MI 500 series also supports UA Link in 2027.
- NVLink fusion opportunity: revenue per XPU expected to be in line with UA Link switch revenue.
- Warrant agreement with Amazon: 3.3 million warrant shares, performance conditions tied to up to $6.5 billion of smart fabric switches, signal conditioning, and optical engine solutions.
- Optical for scale-up timing: 2028; initial CPO deployment may occur with scale-out before scale-up.
- Mgmt stance: Bullish — strong customer endorsements (AWS, AMD, NVIDIA) and growing ecosystem; optical TAM doubling.
Q2 — Vivek Arya
- Topic: Scorpio sales performance and NVIDIA Vera Rubin content opportunity
- Key points:
- Scorpio (P series, scale-out) broke above 15% of sales for 2025; initial Scorpio X (scale-up) volumes shipped, with material ramp in back half of 2026.
- Astera’s opportunity with NVIDIA Vera Rubin arises only in custom deployments of the reference design; minimal opportunity with reference design itself.
- Lead hyperscaler customer plans to deploy Vera Rubin as a custom deployment.
- Mgmt stance: Neutral — Scorpio growing fast, but Vera Rubin content depends on custom deployment decisions by customers.
Q3 — Tore Svanberg
- Topic: SAM growth, scale-up customization, and Scorpio X ramp timeline
- Key points:
- SAM by 2030: $20 billion; half addressable today via PCIe, UA Link, and platform-specific scale-up topologies.
- Scale-up fabrics require customization; Astera uses software-defined silicon to manage investment.
- Scorpio X: pre-production in first half 2026, ramp with lead customer in second half; additional customer pre-production in second half with ramps in 2027.
- Mgmt stance: Bullish — increasing R&D investment now to capture growing scale-up opportunities; strong customer traction.
Q4 — Ross Seymore
- Topic: Optical development timeline and Scorpio growth trajectory
- Key points:
- Optical technology: start-to-revenue timeline of 18–24 months on the earlier side; company has been building internal capabilities and made an acquisition last year.
- Scorpio crossed 15% of sales in 2025 (P series only); Scorpio X (scale-up) has larger TAM, initial volumes in first half 2026, material step-up in back half.
- Scorpio expected to become the biggest product line at some point, but Ares, Taurus, and Leo are also growing.
- Mgmt stance: Bullish — optical built “the Astera way” with customer input; Scorpio trajectory strong, though crossover timing uncertain.
Q5 — Mike Tate
- Topic: Optical development lead time and Scorpio growth bogeys
- Key points:
- Optical development lead time longer, but new opportunities can turn into silicon in 18–24 months.
- Company has been working on optical since inception; internal build plus last year’s acquisition.
- Scorpio crossed 15% of sales in 2025 (P series); X series for scale-up has much larger TAM, initial volumes in first half, material ramp in back half.
- Mgmt stance: Bullish — optical built cost-effectively internally; Scorpio on track to become largest product line.
Q&A Batch (6-10 of 10)
Q6 — Sebastien Cyrus Naji
- Topic: Custom solutions (Scorpio X) ASP and margin profile; Taurus line growth drivers
- Key points:
- Custom solutions (e.g., NVLink fusion) have higher attach rate (one per accelerator vs. one switch shared across few accelerators), but ASP may not be at same level as native switches; revenue content evens up.
- Every generation of XPU, Astera’s content has grown; trend continues.
- Taurus line growth driven by both market transition (400G to 800G) and Astera’s share gain; total revenue for Taurus up ~4x for full year.
- Astera provides modules inside cable assemblies, not whole cables; expects to come in strong as 800G volume picks up.
- Mgmt stance: Bullish — sees continued content growth per XPU generation and strong Taurus momentum from speed transition and business model.
Q7 — Sean O'Loughlin
- Topic: Amazon warrant agreement details and NVLink Fusion financial arrangement
- Key points:
- Warrant agreement filed in 8-K; warrants earned as revenue milestones achieved ($6.5 billion outlined).
- Non-cash charge for warrant value vests, hitting gross margins by ~2 points per quarter starting Q2.
- Warrants have a life of seven years (to 2033).
- NVLink Fusion business model not disclosed due to NDAs; opens new opportunities in NVLink ecosystem, additive to existing business.
- Mgmt stance: Neutral — limited detail on NVLink Fusion; warrant impact quantified as non-cash margin headwind.
Q8 — Karl Ackerman
- Topic: Optical investment opportunity vs. signal conditioning; Scorpio P design wins and 2026 outlook
- Key points:
- Optical opportunity for scale-up is “very large,” likely larger than signal conditioning opportunity; first deployments for scale-up optical expected in 2028.
- OpEx increase also funds AI fabric investments (e.g., Israel team) for near-, mid-, and long-term growth.
- Scorpio P has design wins with three hyperscalers (two new); these are for scale-out use cases, supporting both merchant GPU and custom accelerator platforms.
- New Scorpio P customers go into production late 2026, with meaningful revenue expected in 2027.
- Mgmt stance: Bullish — optical opportunity sized as larger than signal conditioning; Scorpio P traction strong, with multiple design wins and only Gen6 switch in high-volume production.
Q9 — Srinivas Pajjuri
- Topic: Retimer growth drivers; UA Link vs. ESON protocol positioning
- Key points:
- Retimer growth driven by PCIe Gen5 to Gen6 transition, chip-down applications, and cabled applications (smart cable modules with higher ASP).
- Copper preferred by customers; optical transition expected first via pluggable optics, then co-package optics for scale-up around 2028.
- Astera can design ESON-based solutions if needed, but customers currently focus on UA Link; aspiration is to address full connectivity TAM.
- With additional resources, can pivot to other protocols if market shifts.
- Mgmt stance: Bullish on retimer growth; neutral on protocol debate — follows customer demand, capable of pivoting.
Q10 — Quinn Bolton (Robert on for)
- Topic: 800G AEC qualification and Scorpio P new customer scale
- Key points:
- 800G AEC qualification progressing; Astera expects to gain revenue as volume expands, with multiple customers using 800G AEC.
- New Scorpio P hyperscaler customers (two new, US-based) have potential to be as big as the main hyperscaler; revenue impact expected in 2027.
- These are mainstream use cases.
- Mgmt stance: Bullish — 800G AEC opportunity broad-based; new Scorpio P customers have potential for similar scale as existing large customer.