There were a number of bearish views in the market today on CPO and the broader

序号：405
星球链接：打开网页
附件：图片 0，音频 0，文档 0
音频文件：无音频

图片

无图片

正文

There were a number of bearish views in the market today on CPO and the broader optics supply chain. We believe concerns around the CPO timeline are somewhat overdone. As an important evolution path for next-generation AI scale-out networks, CPO will naturally need to go through customer validation, system integration, reliability testing, and supply-chain ramp-up. But that does not mean the project itself is experiencing a material delay. In fact, Jensen already stated very clearly at Computex that CPO will ramp on schedule. Based on what we are hearing from the supply chain, scale-out CPO remains broadly on track, rather than reflecting weakening demand or any change in customer attitude.

From what we understand, scale-out CPO deployment at NCP customers, including CoreWeave and Lambda, as well as Oracle and one major CSP, is progressing toward mass production as planned. Supply-chain feedback has also been quite positive. The current focus is more on capacity preparation, yield ramp, customer qualification, and system-level integration, rather than project cancellation or a meaningful pushout. Therefore, we do not think the pace of the CPO ramp should be interpreted as an inflection point for optics demand.

More importantly, the investment thesis for optics does not depend solely on the CPO path. As we discussed in our previous NPO update, the doubling of bandwidth requirements on the scale-up side is actually a very strong tailwind for optics. Whether the final architecture is CPO, NPO, or a coexistence of CPO and NPO, the key point is that AI clusters will continue to rely more heavily on optical interconnects both inside and outside the cluster. System architecture is evolving toward higher bandwidth density, lower power consumption, shorter reach, and higher levels of integration. This is structurally positive for the entire optics supply chain.

For the optics chain, especially Lumentum / LITE, we believe the key issue is not whether a specific CPO product ramp is pulled in or pushed out by one quarter. The real issue is that the industry has clearly entered a supply-constrained phase for InP and laser capacity. Across the industry, the scarcest resources today are InP epitaxy, InP chip manufacturing, laser capacity, and yield. Whether this capacity is ultimately used for CPO-related ultra-high-power lasers, CW lasers, or EMLs, customers will absorb the output quickly, and suppliers have very strong pricing power.

This is why we believe the market is misreading the risk for LITE. LITE is not simply betting on a single CPO cycle. It sits in one of the tightest and most pricing-powerful parts of the AI optics supply chain. Even if the ramp of one specific product fluctuates slightly, other laser products can quickly fill the capacity and contribute similar revenue and profitability. In other words, LITE’s real problem today is not insufficient demand, but insufficient capacity.

In addition, LITE and COHR are not just CPO laser stories. Beyond lasers, both companies also directly benefit from multiple growth areas, including OCS, optical modules, ITLA, pump lasers, and Coherent Lite. The industry feedback we are hearing points to supply shortages, customers competing for capacity, and tight lead times, rather than demand peaking. Therefore, we believe that if the market turns bearish on the optics chain, especially LITE, because of short-term noise around the CPO timeline, it may be missing the bigger structural reality: the bandwidth bottleneck in AI infrastructure is still getting worse, and optics remains one of the core paths to solving that bottleneck. 今日市场上对 CPO 以及更广泛的光学供应链存在不少悲观看法。我们认为对 CPO 实施时间表的忧虑在某种程度上有些过度。作为下一代 AI 规模扩展网络的重要演进路径，CPO 自然需要经历客户验证、系统集成、可靠性测试以及供应链逐步扩大的过程。但这并不意味着该项目本身正经历实质性延迟。事实上，詹森已在 Computex 上非常明确地指出，CPO 将按计划推进。根据我们从供应链处获取的信息，规模扩展型 CPO 总体上仍进展顺利，并未反映出需求减弱或客户态度发生任何变化。

据我们所知，NCP 客户（包括 CoreWeave 和 Lambda）以及 Oracle 和一家主要云服务提供商的横向扩展 CPO 部署工作正按计划逐步迈向大规模生产阶段。供应链反馈情况也相当积极。目前的工作重点更多在于产能准备、产量提升、客户资质审核以及系统层面的整合，而非项目取消或实质性延期。因此，我们认为 CPO 部署提速不应被解读为光学需求的一个拐点。

更重要的是，光学领域的投资逻辑并不完全依赖于 CPO 路径。正如我们在先前的 NPO 简报中讨论过的，规模扩大所带来的带宽需求翻倍实际上为光学领域提供了极为强劲的推动力。无论最终架构是 CPO、NPO 还是 CPO 与 NPO 的共存模式，关键在于 AI 集群将继续在集群内外更加依赖光学互连技术。系统架构正朝着更高的带宽密度、更低的功耗、更短的传输距离以及更高的集成水平方向发展。这对整个光学供应链而言具有结构性利好效应。

对于光学产业链，尤其是 Lumentum / LITE 而言，我们认为关键问题并不在于某一特定 CPO 产品是否会在某个季度被叫停或推进。真正的问题在于，行业显然已步入 InP 和激光产能供应受限的阶段。纵观整个行业，目前最稀缺的资源是 InP 外延、InP 芯片制造、激光产能以及产量。无论这些产能最终是否会被用于 CPO 相关的高功率激光器、连续波激光器或 EML，客户都将迅速吸收产出，而供应商则拥有极强的定价能力。

正因如此，我们认为市场正在错误地评估 LITE 所面临的风险。LITE 并非单纯押注于单一的 CPO 周期。它所处的位置是 AI 光学供应链中最为紧张且定价能力最强的环节之一。即便某一特定产品的增长势头稍有波动，其他激光产品也能够迅速填补产能，并带来相似的营收与盈利能力。换言之，LITE 目前真正面临的问题并非需求不足，而是产能不足。

此外，LITE 和 COHR 不仅仅是 CPO 激光领域的故事。除了激光技术以外，这两家公司还直接受益于多个增长领域，包括 OCS、光学模块、ITLA、泵浦激光器以及 Coherent Lite。我们收到的行业反馈表明，当前面临的状况是供应短缺、客户竞相争夺产能以及交货周期紧张，而非需求见顶。因此，我们认为，如果市场对光学产业链（尤其是 LITE）的看法因围绕 CPO 时间的短期噪音而转向悲观，那么这可能忽略了更为重要的结构性现实：人工智能基础设施中的带宽瓶颈问题仍在不断加剧，而光学技术依然是解决这一瓶颈问题的核心途径之一。

总体总结

主题正文

There were a number of bearish views in the market today on CPO and the broader optics supply chain. We believe concerns around the CPO timeline are somewhat overdone. As an important evolution path for next-generation AI scale-out networks, CPO will naturally need to go through customer validation, system integration, reliability testing, and supply-chain ramp-up. But that does not mean the project itself is experiencing a material delay. In fact, Jensen already stated very clearly at Computex that CPO will ramp on schedule. Based on what we are hearing from the supply chain, scale-out CPO remains broadly on track, rather than reflecting weakening demand or any change in customer attitude.
From what we understand, scale-out CPO deployment at NCP customers, including CoreWeave and Lambda, as well as Oracle and one major CSP, is progressing toward mass production as planned. Supply-chain feedback has also been quite positive. The current focus is more on capacity preparation, yield ramp, customer qualification, and system-level integration, rather than project cancellation or a meaningful pushout. Therefore, we do not think the pace of the CPO ramp should be interpreted as an inflection point for optics demand.
More importantly, the investment thesis for optics does not depend solely on the CPO path. As we discussed in our previous NPO update, the doubling of bandwidth requirements on the scale-up side is actually a very strong tailwind for optics. Whether the final architecture is CPO, NPO, or a coexistence of CPO and NPO, the key point is that AI clusters will continue to rely more heavily on optical interconnects both inside and outside the cluster. System architecture is evolving toward higher bandwidth density, lower power consumption, shorter reach, and higher levels of integration. This is structurally positive for the entire optics supply chain.
For the optics chain, especially Lumentum / LITE, we believe the key issue is not whether a specific CPO product ramp is pulled in or pushed out by one quarter. The real issue is that the industry has clearly entered a supply-constrained phase for InP and laser capacity. Across the industry, the scarcest resources today are InP epitaxy, InP chip manufacturing, laser capacity, and yield. Whether this capacity is ultimately used for CPO-related ultra-high-power lasers, CW lasers, or EMLs, customers will absorb the output quickly, and suppliers have very strong pricing power.
This is why we believe the market is misreading the risk for LITE. LITE is not simply betting on a single CPO cycle. It sits in one of the tightest and most pricing-powerful parts of the AI optics supply chain. Even if the ramp of one specific product fluctuates slightly, other laser products can quickly fill the capacity and contribute similar revenue and profitability. In other words, LITE’s real problem today is not insufficient demand, but insufficient capacity.
In addition, LITE and COHR are not just CPO laser stories. Beyond lasers, both companies also directly benefit from multiple growth areas, including OCS, optical modules, ITLA, pump lasers, and Coherent Lite. The industry feedback we are hearing points to supply shortages, customers competing for capacity, and tight lead times, rather than demand peaking. Therefore, we believe that if the market turns bearish on the optics chain, especially LITE, because of short-term noise around the CPO timeline, it may be missing the bigger structural reality: the bandwidth bottleneck in AI infrastructure is still getting worse, and optics remains one of the core paths to solving that bottleneck.
除了激光技术以外，这两家公司还直接受益于多个增长领域，包括 OCS、光学模块、ITLA、泵浦激光器以及 Coherent Lite。
因此，我们认为，如果市场对光学产业链（尤其是 LITE）的看法因围绕 CPO 时间的短期噪音而转向悲观，那么这可能忽略了更为重要的结构性现实：人工智能基础设施中的带宽瓶颈问题仍在不断加剧，而光学技术依然是解决这一瓶颈问题的核心途径之一。

源文件：2026-06-10/topics/405.md

站内副本：2026-06-10/topics/405.md