OpenLight embedded laser and modulator IP

This pushes OpenLight up the value stack from selling parts for pluggable optics to supplying the photonic building blocks that sit inside the switch or GPU package itself. In practice, that means its laser and modulator IP can move from a transceiver module at the edge of the board to the substrate beside the compute or switching die, where power loss, footprint, and signal integrity matter most for 1.6T to 3.2T AI links.

• OpenLight already sells a drag and drop PDK with lasers, modulators, amplifiers, photodiodes, and reference PIC designs, then makes money again through engineering work and production royalties. Embedded IP extends that same model into co-packaged optics, where the customer is no longer just a module maker, but also a switch ASIC, GPU, or system company.
  1 sacra 2 sacra
• The competitive bar is set by companies already proving in package optics. Intel has shown a 4 Tbps optical I O chiplet designed to sit next to CPU, GPU, and other SoC devices, with on chip lasers and amplifiers. Marvell has demonstrated both 1.6T light engines and a 6.4T CPO light engine for AI interconnects.
  3 intel 4 marvell 6 marvell
• Ayar Labs shows the adjacent model. It sells an optical I O chiplet and external light source that fit standard chip packaging flows for hyperscaler and XPU designs. OpenLight is taking a more IP and foundry oriented path, which can let many chip companies embed photonics without building a full captive optics stack themselves.
  2 sacra 5 ayarlabs 7 ayarlabs 8 ayarlabs

The next step is a market split between closed full stack optical engines from large chip vendors and open photonics IP layers that others can design into their own packages. If co-packaged optics becomes standard in AI clusters after pluggables hit their power and density limits, OpenLight can become the picks and shovels supplier for that transition.