北京时间 2026-10-10 21:14:12 · 个股论点

分析SIVE在产能、资质及技术上的三大护城河，看好其长期竞争力。

涉及标的： $AAOI $AVGO $JBL $LITE $MTSI $SIVE

中文翻译

@Rian_2026 好问题！只是些非正式想法……看 $SIVE 有三个“护城河”： - 产能 - 资质认证（尤其是定制型分布反馈激光器(Custom DFB lasers)） - 技术 1. 产能方面，他们将于 2027 年 Q4 投产 1 亿+ 连续波(Continuous Wave) DFB 激光器（Trendforce 数据显示目前全球 EML/CW 激光器供应约 6.08 亿）。显然到 2027 年全球产能将大幅增长，$AAOI、$AVGO（3 倍扩张）等将投产。但如果按他们 2:1 的说法推算，意味着长期来自自有供应+外部代工厂的激光器达 3 亿/年。这在被引述为 3-5 年的短缺期间极具价值，且 $LITE 称其需求可见度已至 2031 年…… 且当供应商因瓶颈开始认证更多厂商（如从早期的 2-3 家增至 5-7 家）时。我确信他们实现这一惊人产量的关键在于制造工艺/良率（这是护城河）。 2. 资质方面，若看与中国厂商合作的 70mW/100mW 可插拔产品线，我认为这更通用，更多回归到产能和制造工艺以实现这些数字。他们与 Ayar 合作的定制 DFB 阵列（2023 年曾暗示还有 Celestial 等 2-3 家）在光 I/O 相关领域需数年认证，若完成将是护城河。所以 $MTSI 或其他公司很难轻易替换或设计绕过。当然共封装光学(CPO) 大规模量产在 2028 年，而我们还在 2026 年下半年。 3. 人们总可比较功率输出等原始规格，但这可能更微妙，所以最好指向客户声明。关于 $JBL 1.6T 长距可插拔(LRO)：Jabil 执行副总裁称“我认为我们的 LRO 部件有形成相对显著护城河的潜力……”尤其涉及功耗。我猜 $SIVE 是该线唯一来源。 Jabil 可能是其激光器在明年 Q2 左右大规模生产场景中表现的最重要证明。 Ayar 也类似表示“没有 Sivers 我们做不到”针对定制方案。

英文原文

@Rian_2026 Good question! Just some informal thoughts... there's three "moats" to look at for $SIVE: - Capacity - Qualifications (especially for custom DFB lasers) - Technological 1. For capacity, they're coming online with 100M+ CW DFB lasers Q4 2027 (global supply is roughly ~608M EML/CW lasers currently per Trendforce). Obviously by 2027 global capacity is going to grow a lot, with $AAOI, $AVGO (3x expansion), and others coming online. But if you extrapolate their 2:1 statement, that implies 300m lasers/year long term coming from their own supply + external foundries. This is incredibly valuable during a shortage that was quoted to be 3-5 years, and $LITE stating they already have demand visibility into 2031... And when suppliers have started to qualify many more (eg. 5-7 vendors vs. 2-3 earlier) due to bottlenecks. I'm sure a lot of how they achieve that staggering amount is due to their manufacturing process/yields (which is a moat). 2. For qualifications, if you look at 70mW/100mW pluggable lines with Chinese players, I'd argue this is more generic and moreso reverting back to a capacity argument and manufacturing process to achieve those numbers. Custom DFB arrays that they've been working on with Ayar (they hinted at 2-3 others back in 2023 like Celestial) take years for optical i/o related things and would be a moat if they finish quals. So it's not quite something you can hot swap or design out really easily for $MTSI or another company. But of course cpo scale up and such is mass volume in 2028, and we're still H2 2026. 3. People can always compare raw specs like power output, but it's probably way way more nuanced so easier to just point toward customer statements. With $JBL 1.6T LRO: "I think that there's the potential for a relatively dramatic moat with our LRO part…" from Jabil's EVP, especially related to power consumption. And $SIVE is my guess sole source in this line. Jabil is probably the most important testament to how their lasers perform in mass production scenarios maybe around Q2 next year. Ayar said something similar along the lines of "we couldn't do it without Sivers" for custom implementations.

在 X 上查看原推 ↗