6 Comments
User's avatar
Atlas.ML's avatar

I think people are saying the the pluggable optical transceiver is still needed for CPO. But more chips moving from the transceiver to CPO package on board.

But you need to fit the EML in the transceiver and other chips and components to transmit that data to other PCB chipsets.

Nobody seems to know how the value chain will shift tectonically in the CPO high penetration era.

But innolight and eoptolink share price, based on forward P/E valuation do not seem to suggest value compression in CPO era in the 1.6T

I don’t know the answer yet. Maybe it’s just A-Share AI concentration premium that A-share investors don’t care about 1.6T massive transition. That’s 2H2027 growth story.

叶岚's avatar

they are developing socketed CPO.

Atlas.ML's avatar

Ya. I’m trying to figure out the % value shift from transceiver module to the CPO on board

The idea of CPO is the speed

Not quite understand their socketed approach

EML Must be socketed in case of light source burnt out

UnsolicitedNotFinancialAdvice's avatar

I'm surprised that you haven't mentioned $FN in the context of module assembly. Do I understand correctly that that's the only non-Chinese assembler? Where do they fit in this structure?

Atlas.ML's avatar

i think Zhongji Innolight and Eoptolink commands 60% of globlal transceiver module assembly.

Innolight 40% m/s and Eoptolink 20% m/s.

Innolight serves google and microsoft then Eoptolink serves AWS.

these two companies dominates all high-end leading transceiver module.

I don’t know about FN but module assembly needs workers on the production line on top of the transceiver design. it’s natural to think that only China module assembler can achieve lowest cost in this business.

UnsolicitedNotFinancialAdvice's avatar

I think FN has factories in Southeast Asia.