Voyant targets AI data centres with SiPh
The company is developing a chip-scale optical circuit switching engine based on technology industrialised for FMCW LiDAR.
Voyant Photonics is developing a silicon photonics optical circuit switching (OCS) engine for AI data-centre networks, adapting its production-proven switching architecture originally developed for frequency-modulated continuous-wave (FMCW) LiDAR.
The company plans to supply the compact switching engine to optical networking companies for integration into their own platforms, rather than developing complete networking systems.
The architecture uses solid-state optical paths without moving switching elements and is designed to deliver switching times below five microseconds.
Voyant said its technology has been engineered for trillions of switching cycles, with a focus on reliability, low optical loss and semiconductor manufacturability.
The company is adapting the platform to meet data-centre requirements, including optical loss and crosstalk, fibre attachment, port density, packaging, testability and compatibility with network control systems.
Optical circuit switching can enable dynamic reconfiguration of network connections between GPUs and other computing resources, potentially reducing the need for repeated optical-electrical-optical conversions in AI infrastructure.
Voyant also sees potential for its silicon photonics architecture to support closer integration with co-packaged optics (CPO) as data-centre networking requirements evolve.
