22 Comments
User's avatar
Bayside Growth Investment's avatar

This article is super interesting and I assume you are right that Lumentum is in a good position on CPO, at least for now. However, I am concerned that "CW lasers combined with silicon photonics has become the main alternative route for CSPs facing EML shortages.," per Trend Force.

https://www.trendforce.com/presscenter/news/20251208-12823.html

Is it possible that while Lumentum may hold a dominant position for the moment on CPOs, it may face a growing threat from silicon photonics more generally (given that as you note CW is more commoditized)?

And even in CPOs, are Broadcom and Marvell likely to be competitors? I think I read somewhere that AWS is working with them on CPOs, although I don't know how comparable they are as potential competitors in the photonics space.

Jason's Chips's avatar

Yes, it is true that SiPho is a large headwind to EML growth and would dilute Lumentum's content to the cheaper CW laser. See my recent earnings coverage. However, this is important for 1.6T generation transceivers, not for scale-up CPO, which is orders of magnitude larger. In the CPO market there is no such thing an EML, there is only CW, and the CW laser isn't the commodity low power version, it is the high power narrrow linewidth kind only Lumentum is able to make. Broadcom and Marvell are not competitors here. They make other components for transceivers like the DSP but not Indium Phosphide based components which are what these lasers are. Coherent is the only other player.

Bayside Growth Investment's avatar

Thanks so much for the reply!

Apparently, according to Broadcom's website (and AI) "despite being fabless for CMOS logic, Broadcom operates leading InP fabrication facilities, handling epitaxial growth to wafer and chip fabrication in-house to produce Fabry Perot, DFB, and EML laser devices. Broadcom's Optical Systems Division in Breinigsville, PA, specifically manufactures InP lasers and detectors, focusing on high-performance EML and DFB lasers."

https://www.broadcom.com/products/fiber-optic-modules-components/components-broadband/long-reach-inp-die

https://www.broadcom.com/products/fiber-optic-modules-components/components-broadband/long-reach-inp-die/cw-lasers

Various AI reviews claim that they make the same CW lasers you referenced. "300mW to 400mW per diode Ultra-High Power (UHP) laser." AI reviews can be mistaken, however.

I wouldn't point any of this out if it didn't matter. It *seems* as though Lumentum's main competitor on this isn't Coherent, but maybe Broadcom. Then again, those resources I mentioned above may be wrong. (If you think the AI is wrong, I will believe you, but I would love your thoughts after you look at the above).

Jason's Chips's avatar

I do not believe they offer them? Cannot find it on the product page and none of the people I talk to have mentioned Broadcom as a player in the space. High power narrow linewidth lasers are extremely hard to make, announcing a product and making it at scale are very different things. Coherent has qualified actually and I am also long Coherent but being objective no one can put a dent in Lumentum's CPO laser market share. They perfected this technology for subsea applications and became very lucky that it is now useful for scale-up networking.

Roger Breuer, CFA, ACCA's avatar

Jason,

I enjoyed this article. Many thanks for the discussion on the laser position relative to the GPU.

I wrote an article on Halma plc, which discusses their Avo Photonics unit. You might find of some interest.

https://rogerbreuer.substack.com/p/halma-plc-from-tea-plantations-to?r=1qxcim

Kannon Do's avatar

I appreciate your deep research and analysis. I learned a lot and very interesting to read.

Jason's Chips's avatar

Thank you. Appreciate your support

cuhaven's avatar
cuhaven's avatar

also wonder your thoughts on the potential cannibalization effects of CPO on its other legacy product lines. Thx!

Jason's Chips's avatar

Revenues from CPO would be so large you'd want as much cannibalization as possible. With that being said, the market is growing so fast and CPO is mostly an expansion of the optical market (scale up) rather than cannibalization, I am not worried about CPO cannibalization for most optical players, maybe besides Marvell.

Jason's Chips's avatar

Appreciate it! This is old news, Coherent is still 2 years behind. High power narrow linewidth lasers are very difficult to make and Lumentum is already ready commercially.

cuhaven's avatar

Thanks for the reply! That's what i thought - still just samples for Coherent.

John's avatar

Well written article on LITE and CPO implications.

I am asking myself isn’t the stock price already reflecting the forward years expectations with FCF with a discount rate of 12%?

Jason's Chips's avatar

Thank you. Regarding intrinsic value, I have built out my DCF model already, and according to my assumptions, LITE is undervalued. However, it obviously depends on the inputs to your model, and using conservative estimates by the street will obviously yield a DCF value far below the current price. I will cover this in part 4.

Andy Liao's avatar

Your modeling is outstanding, Jason. I really appreciate the clarity of your numbers and the transparency around your sources and underlying assumptions—this is top-tier equity research. I’m curious to hear your current thoughts on LITE’s valuation. As you highlighted, LITE is a pure-play in the optics space, which arguably justifies a premium valuation. However, given Coherent’s technological edge and its more complete supply chain—especially its independence from external manufacturers like Fabrinet, which LITE relies on—do you believe the current valuation gap between the two is fully justified?

Jason's Chips's avatar

Thanks Andy, I'm so grateful for the compliment, means a lot! Appreciate the thoughtful question.

Regarding valuation, the majority of LITE's intrinsic value is derived from CPO and OCS, not optical transcievers, in my view. Transcievers are really just a side business they entered with the acquisition of cloud light. Coherent has a big advantage here -- many, many multiples of LITE's capacity and full vertical integration. In OCS, LITE is ahead of COHR as they use MEMS technology making an engineering tradeoff of performance for reliability. COHR argues their solution has no moving parts which enhances reliability and makes it last longer, but they instead face 3db of insertion loss which means the signal gets wrecked in the transport process. LITE is already capturing more of the market as a second source to Google. COHR has to fight for the rest of the hyperscalers, where I believe LITE still has an advantage. I believe this can be $3b+ of annual revenue for LITE in 2030. For CPO, as I mentioned in the post, LITE is so far ahead in ELS they are a straight up monopoly. As CPO gets adopted in scale out I can imagine it actually cannibalizing some transciever volume which is directly bad for COHR good for LITE. In reality I think they are very different companies, even though the market perception is often the opposite. Different companies, different multiples.

Noah Truwit's avatar

Have you looked at early stage XScape as a potential laser component manufacturer? It seems like their technology increases the highway for data

Jason's Chips's avatar

Hi, thanks for your comment. I am not involved in the early stage/venture space unfortunately, so I have not been tracking them. They do have an interesting partnership with Tower Semi a few months back.

Noah Truwit's avatar

Thanks! Really great analysis in your pieces.

Flyonthewall's avatar

wonderful analysis, i am just jaw breaking to see the 5200 ASP number, is that right? how a low power CW laser selling at $6 while a high power selling at $250, and why a module needs 8 UHP laser, thanks for your thoughts

Jason's Chips's avatar

Thanks for reading! Your intuition is most likely right, I have been in contact with several HF analysts who have informed me that the UHP laser BOM assumptions I have made are too aggressive. I will be revising my model soon. However, I can confirm there are 8 UHP lasers in one ELS module. Such a setup is needed for redundancy and power.

Investor's Ledger's avatar

Great piece! The “CPO ≠ co-packaged lasers” nuance is huge. If copper really hits the physics wall, those ELS lasers might end up being the quiet choke point.