Compact non-volatile photonic chip design performs matrix multiplication

Researchers designed a compact photonic chip that performs matrix-vector multiplication using light, enabling non-volatile, energy-efficient neuromorphic computing with broadband tolerance across a 40 nm window. The chip's design relies on phase-change GSST cells for fast, reversible switching without continuous power, potentially overcoming electronic processor limitations for large neural networks.

This Wire brief sits within Fusion42's coverage of AI Infrastructure, Photonics & Optics and Semiconductors.