Communication

Sivers Semiconductors and SemiNex Announce $3.4M Program for AI Data Center Light Sources

Listen to this story

AI NARRATED
0:00 / 0:00
Sivers Semiconductors and SemiNex Corporation have announced a program, initially valued at roughly $3.4 million, to develop next-generation indium phosphide (InP) light sources for AI data center interconnects, the optical links used to move data between servers and racks at high speed.
 
The program targets several specific technologies: high-power external laser sources for co-packaged optics (CPO), a technique that integrates optical components directly alongside chip packaging to speed up data transfer; high-channel-count DFB (distributed feedback) laser arrays, used to support wavelength-multiplexed links, meaning connections that carry multiple data streams simultaneously over different light wavelengths; and semiconductor optical amplifiers (SOAs), which boost optical signals to extend how far they can travel as the number of channels in a system increases.
 
The push behind this work reflects a broader shift in optical networking: as optics move from sitting at the outer edge of a chip package (the "faceplate") into being integrated directly within the package itself, factors like power output per wavelength, wall-plug efficiency (how much of the electrical power going in actually gets converted into useful light output), and thermal stability increasingly determine how far a given system architecture can scale. That makes designing the light source itself a system-level engineering challenge, rather than simply a matter of picking an off-the-shelf component.
 
Vickram Vathulya, CEO of Sivers Semiconductors, said the advanced light sources needed for next-generation optical networking inside AI data centers are just as critical as the compute, memory, and silicon photonics components they connect. He said the company values its partnership with SemiNex for combining innovative technologies to deliver strong light source solutions at scale.
 
Ronald Moore, CEO of SemiNex, said the optical layer is increasingly becoming the limiting factor in how far AI infrastructure can scale, with the industry needing more power, more wavelengths, and higher efficiency, gains that can't be achieved simply by pushing existing devices harder. He said those improvements instead need to come from the design and manufacturing of the InP material itself, which is exactly where the two companies' teams are focused in this collaboration.
E

EEHerald News Desk

Editor, Electronics Engineering Herald


More from Communication