2026-05-03 · 供应链分析

Sivers成为CPO关键激光供应商，NVDA引发的产能瓶颈正驱动AMD等厂商争抢上游资源。

涉及标的： $AMD $COHR $GFS $LITE $MTSI $MVL $NVDA $POET $SIVE

中文翻译

看到$SIVE成为CPO(共封装光学)的Tier 1激光供应商，这真的很疯狂。这是我的预测/猜测与大致映射： $NVDA用$COHR、$LITE的产能（在他们新的20亿美元以上支出计划之后）。早期2025年EML用了同样的手法，导致了今天的瓶颈。现在，$AMD/超大规模云商正在争抢上游激光供应商。因此，$SIVE + Win / $GFS成为可能的主要途径。你可以从以下看出： -> $MVL CPO通过Celestial（Nvidia签了协议，但他们没有激光器） -> $AMD CPO -> Ayar -> $POET -> Lightmatter -> 其他项目（如Jabil 1.6T）因此，Sivers/Win成为Tier 1尖端关键独立激光供应商。这一点的一些暗示： 1. $GFS在其生态系统中将$SIVE / $LITE列为仅有的两家公开激光供应商。 2. Ayar将$LITE / $MTSI从其网站上移除，并将$SIVE提升为主要激光供应商。所以其他所有人最终都选择了Sivers，因为$COHR / $LITE已经完全分配。我的猜测是，随着架构标准化，许多二级供应商也会抢占溢出市场份额。明年年初，考虑到$NVDA的瓶颈，这场争夺关键环节的竞争将异常激烈。而这家瑞典小型公司$1.2B规模的$SIVE将处于这场竞争的中心。

英文原文

It's pretty insane to see $SIVE become a Tier 1 laser supplier for CPO. This is my prediction/guess with est. mapping: $NVDA ate up all the capacity with $COHR, $LITE after their new $2B+ spending spree. Same playbook with EML early 2025, causing the bottleneck seen today. Now, $AMD / hyperscalers are scambling for upstream laser suppliers. Hence why $SIVE + Win / $GFS became likely primary route to go down. You can see this with: -> $MVL CPO through Celestial (Nvidia signed a deal, but they don't have lasers) -> $AMD CPO -> Ayar -> $POET -> Lightmatter -> and other programs (eg. Jabil 1.6T) As a result, Sivers/Win emerged as the Tier 1 bleeding edge + critical independent laser supplier. And there's hints for this when: 1. $GFS listed $SIVE / $LITE as the two only public laser suppliers in their ecosystem. 2. Ayar removed $LITE / $MTSI off their website and elevated $SIVE to their primary laser supplier. So everyone else ended up going with Sivers since $COHR / $LITE are fully allocated. My guess is that a lot of the secondary suppliers also capture overflow as architectures standardize. But scramble for this chokepoint will be insane early next year given $NVDA bottleneck. And a small $1.2B Swedish company in $SIVE will be in the center of it.

在 X 上查看原推 ↗