Reading depth
- Why Optical Becomes the Next BottleneckThat gap creates pricing power, mix improvement and customer priority.Jump to section
- 200G EML — Lumentum and Coherent as the Two Key NamesA 1.6T transceiver needs eight 200G EMLs, and the bottleneck hits the laser-modulator core before module assembly. Lumentum and Coherent own it.Jump to section
- 1.6T Timeline — 800G Does Not Disappear800G does not end but overlaps the 1.6T ramp, intensifying the bottleneck. For Lumentum the question is revenue mix, for Coherent absorbing residual demand.Jump to section
- CPO Commercialization — Fewer Boxes, More PrecisionCPO does not kill optical makers but makes them more precise. Module boxes shrink, but in-package precision and reliability demands rise, lifting component value.Jump to section
- Next Week's Earnings — Backlog and Mix Matter More Than the HeadlineNext week's earnings read by backlog and mix, not the headline. A bottleneck firm's good print shows up as how to allocate supply, not strong demand.Jump to section
- Conclusion — Exposure Matrix by Bottleneck LocationThe exposure matrix is read by bottleneck location, not weight. Lumentum is the EML core, Coherent the transceiver platform, with earnings validating the thesis.Jump to section
Why Optical Becomes the Next Bottleneck
For tens of thousands of GPUs to operate as a single cluster, nodes and racks must exchange data continuously. Copper still works inside short server paths, but at 800G and speeds it runs into signal loss and power limits. Light becomes the practical transport layer. Optical transceivers convert electrical signals into light and back again.
The issue is that a transceiver is not a simple assembly. Lasers, modulators, photodetectors, DSPs, packaging and test all have to work together. Among them, the 200G is the narrowest process bottleneck. It needs compound-semiconductor capacity, dedicated fabs and customer-qualified yield, so supply cannot expand instantly.
This is why the optical bottleneck resembles memory. Demand steps up with AI cluster builds, while supply moves through materials, process, yield and qualification constraints. That gap creates pricing power, mix improvement and customer priority.
