What is optical computing, and could it replace electronic chips?

The answer

Optical computing uses photons instead of electrons to process data, promising speeds up to 1,000 times faster than electronic circuits with dramatically lower energy consumption - since photons don't generate heat through resistance. In 2024, startup Lightelligence demonstrated its Hummingbird optical chip performing matrix multiplications for AI at 10 times the efficiency of equivalent electronic chips. MIT researchers created a programmable photonic processor using meshes of optical interferometers that can perform neural network inference at the speed of light. Lightmatter's Passage chip uses photonic interconnects to eliminate the data bottleneck between conventional processors. The fundamental advantage is parallelism: unlike electrons, multiple light beams can cross paths without interfering, enabling massive parallel computation. Current limitations include the difficulty of miniaturizing optical components to electronic scales and the challenge of optical memory storage, meaning hybrid opto-electronic systems are the most likely near-term architecture.