The Module and the Laser Inside It

This index already has a collection on co-packaged optics — moving the optical engine inside the switch package. This one is about what actually ships in volume today: the pluggable transceiver at the faceplate, and specifically where its supply binds.

It does not bind at the module. Module assembly is competitive and substantially Chinese. It binds one layer down, at the laser chip — the electro-absorption modulated lasers and continuous-wave lasers fabricated on indium phosphide wafers. A research house naming three bottlenecks on transceiver capacity expansion puts tight optoelectronic chip supply first, ahead of high-precision optical alignment and thermal management.

The supplier side corroborates it plainly. One laser manufacturer reported company records in EML shipments at both 100G and 200G lane rates, with 200G devices passing a quarter of total EML revenue, while describing its pump lasers as effectively sold out after growing 80% year on year for a second consecutive quarter. Capacity is being added on semiconductor timescales, not assembly ones: a fifth indium phosphide fab is not expected to generate revenue until early 2028, and the company has locked in wafer substrate under an agreement to the end of 2031, paying a $43.5 million deposit with another of the same size due in 2028. Paying deposits years ahead for substrate is what a real constraint looks like from the buyer's side.

What the scarcity earns is visible in the results. Quarterly revenue of $1,006.3 million, more than double a year earlier, on a full year of $3,014 million — up 83.2%. Components, where the laser chips sit, were 64.5% of the quarter. Guidance implied a 39.5 to 40.5% operating margin. That is the same signature this index found at the retimer layer and in automated test equipment: when a component is hard to substitute and the thing it enables is scarce, the margin migrates to the component.

So the units and the margin sit at different layers. An AI optical transceiver market put at $16.5 billion in 2025 and $26 billion in 2026 — growth above 57% — is constrained by a component rather than by the ability to assemble modules, which is exactly why a chip supplier can guide to forty per cent while module building remains a volume business.

Ahead: 800G is the volume product, 1.6T is described as gradually entering mass production, and 2026 to 2027 is called crucial for a foothold in that supply chain. Two directions are named alongside it and they pull against each other — linear pluggable optics strips the digital signal processor out of the module to cut power while keeping it pluggable; silicon photonics changes what the engine is made of; and co-packaged optics, covered elsewhere here, removes the module altogether. The industry is pursuing a cheaper pluggable and a replacement for pluggables at once, which is what happens when nobody knows whether power, cost or serviceability binds first.

A warning about numbers in this sector. Almost every striking figure in circulation — unit shipments, market shares, the proportion of one buyer's volume held by two suppliers, production running a stated percentage below demand — originates in subscription research whose method is not published. None of it is asserted here. Chinese module makers reportedly hold seven of the top ten positions, and that is recorded on this page as a gap rather than a fact, because their own disclosures were not retrieved and the share estimates cannot be checked.