OpenLight's Open Foundry Strategy
OpenLight
OpenLight is trying to become the Arm of laser integrated photonics, not another chip vendor. Instead of building and selling finished optical parts from its own fab, it packages a process, a component library, and design support that outside customers can use at a foundry partner. That matters because customers can build custom PICs with integrated lasers without taking on fab ownership, while OpenLight scales through licenses, services, and production royalties rather than factory utilization.
-
The practical difference versus Intel or Marvell is where the customer plugs in. With OpenLight, a module maker or AI interconnect startup starts from a PDK and tapes out its own design on Tower’s PH18DA process. With vertically integrated players, the customer usually buys a finished transceiver or light engine that stays inside that vendor’s product stack.
-
OpenLight’s model is also more open than a pure foundry offer like GF Fotonix. GF sells high volume silicon photonics manufacturing capacity, while OpenLight adds heterogeneous indium phosphide on silicon building blocks, including on chip lasers, then licenses that stack across foundry relationships. The result is closer to a reusable photonics platform than a standard wafer service.
-
The tradeoff is that openness lowers capital needs but shifts differentiation into process IP and ecosystem control. OpenLight avoids owning fabs, but it is tightly tied to Tower for PH18DA manufacturing, so its moat depends on keeping its design kit, device library, and laser integration ahead of merchant silicon photonics alternatives.
If this model works, photonics will look more like mainstream semis, with independent design platforms feeding multiple product companies instead of each large vendor building everything in house. That would give OpenLight a shot at becoming core infrastructure for AI optical interconnects, especially as 800G and 1.6T designs move from prototyping into volume production.