Optalysis targets AI inference with optical computing

spectrum.ieee.org13 min read

Surging energy demands from AI inference have revived interest in optical computing, with several startups and researchers developing photonic systems to perform matrix-vector multiplication and other mathematical operations using light instead of electricity. While optics promises dramatic improvements in energy efficiency by scaling with vector length rather than quadratically, companies still face significant engineering hurdles regarding data encoding costs, system scale, and integration with traditional electronic architectures. Open article →