HyperLight’s $80 million Series C arrives with a clean narrative: a foundry-aligned thin-film lithium niobate pure-play, freshly funded by MediaTek, UMC, Jabil, and Foxconn, attacking the modulator layer of the optical stack at the precise moment AI networks make that layer matter. For Lumentum and Coherent, the two listed names with the most exposure to datacom optics, the reflex read is a new and well-capitalized threat. The reflex is half right. The threat is real, but it is narrower than the headline, and the part of it that should worry the incumbents is not the part the announcement advertises.
TFLN takes the modulator, not the engine
A transceiver is a light source, a modulator, a detector, and the packaging that holds them together. HyperLight’s chiplet contests one of those four functions. It modulates light it does not generate. The laser still comes from indium phosphide, and that is the function Coherent and Lumentum spent years and capital securing. Coherent is running what it calls the world’s first six-inch indium phosphide production platform, across Sherman, Texas and Järfälla, Sweden, where a six-inch wafer yields more than four times the devices at less than half the per-device cost of three-inch. Industry-wide, InP capacity is the gating constraint on transceiver shipments, not modulator design. A TFLN socket win still pulls an InP light source behind it. The new entrant therefore competes for a slice of the engine while leaving the scarce, capital-intensive heart of it untouched.
That bounds the damage. It does not erase it. The modulator is where a meaningful share of the performance differentiation lives, and ceding it to a merchant supplier is not a neutral outcome for a company whose entire pitch is owning the stack.
The real vector is disintermediation, not displacement
Coherent’s strategy is vertical integration: it designs and builds the EMLs, the CW lasers, the VCSELs, the silicon photonics, the detectors, and the finished modules, outsourcing little beyond the DSP. Lumentum sits adjacent, both a customer and a supplier to Coherent, with its own 800G and 1.6T module franchise. The danger a merchant TFLN chiplet introduces is not that it beats either of them in a head-to-head modulator bake-off. It is that it hands the independent module houses—the InnoLights and Eoptolinks of the world—a best-of-breed modulation layer they can buy off the shelf and bond onto someone else’s laser. That arms the merchant ecosystem against the integrated players’ core argument, which is that only a vertical stack can deliver the performance. If a module maker can assemble a competitive 1.6T or 3.2T part from a UMC-fabricated TFLN chiplet and a third-party laser, the integration premium compresses. The threat is to the differentiation, not to the volume, and the differentiation is what supports the multiple.
Neither incumbent is caught flat
This is not a surprise attack. Coherent’s OFC posture was an explicit multi-technology hedge—silicon photonics, indium phosphide, and VCSEL all on the table—precisely so that no single material transition strands it. It already builds lithium niobate modulators and can fold thin-film versions into that line. Lumentum has signaled it intends to reach TFLN through partnerships and acquisitions rather than a ground-up build, which is the rational posture for a company that would rather buy a qualified platform than fund the years HyperLight just spent. The strategic question for both is build, buy, or partner, not whether to respond. HyperLight’s round actually clarifies that question by putting a price and a foundry relationship on a credible acquisition or licensing target.
The valuation backdrop cuts in the incumbents’ favor more than the narrative suggests. Coherent trades at a lower forward multiple than Lumentum despite owning the scarcer asset, the six-inch InP platform, and despite datacenter revenue growing 37 percent year over year into a record quarter. A new modulator entrant pressures sentiment faster than it pressures the income statement, and sentiment is already discounting Coherent relative to the franchise it holds.
The constraint that decides this
As long as indium phosphide is the bottleneck, the company that scales the laser wins the transceiver, and the modulator entrant competes for position inside an engine someone else still gates. HyperLight changes the modulator conversation. It does not yet change the binding constraint. The day to take the threat seriously is not this funding round. It is the day a hyperscaler writes a TFLN chiplet into a specification by name and sources the laser separately. Until then, Lumentum and Coherent are defending the scarce thing, and the scarce thing is still indium phosphide.
