Lightmatter OEM integration for scale
Lightmatter
The real upside is not selling a few more exotic AI systems, it is becoming a component inside the boxes that incumbent chip and networking vendors already ship at scale. Today Lightmatter sells high value photonic processors and interconnect systems directly into large AI deployments, but its newer Passage products are being shaped into standard optical engines that can sit next to ASICs, on boards, or inside co packaged systems, which is the form factor OEMs need before adoption can spread into mainstream switches, servers, and accelerator platforms.
-
The clearest sign of this shift is product packaging. Passage L20 is built for near package optics and on board optics, uses industry standard electrical interfaces, and is meant to fit XPU and switch roadmaps. That turns Lightmatter from a custom system vendor into a part supplier for mainstream hardware makers.
-
There is already a playbook for this market. Ayar Labs focused on optical I O rather than compute, then won backing and ecosystem ties with Intel, AMD, and NVIDIA as it moved toward high volume manufacturing. Lightmatter is now pursuing a similar OEM path, while keeping the option to bundle a fuller stack around Passage.
-
OEM integration expands the market because servers, switches, and accelerator boards ship in volumes far above bespoke AI clusters. Lightmatter has already partnered with GUC on commercial co packaged optics for hyperscalers, joined UALink to align with open interconnect standards, and worked with GlobalFoundries on mass production, all of which are steps toward design wins inside standard platforms.
From here the market moves from proof of performance to proof of manufacturability. If Lightmatter keeps turning Passage into standard, volume ready modules that fit existing ASIC and switch designs, photonics can spread from a narrow AI interconnect niche into the default way mainstream compute and networking gear moves data.