Lightmatter

Lightmatter is a photonic computing company founded in 2017, building processors that use light instead of electricity to perform AI computations. The company has raised over $400 million and is developing technology that could fundamentally change how AI chips communicate and process data.

Lightmatter's core technology uses photonic (light-based) interconnects and computing elements to achieve dramatically higher bandwidth and lower power consumption compared to traditional electronic chips. The company's Passage product provides photonic interconnects that connect AI chips with much higher bandwidth than electrical connections — addressing the critical bottleneck of chip-to-chip communication in large AI training clusters. Envise is Lightmatter's photonic AI accelerator chip.

As AI models grow larger and training clusters expand to thousands of chips, the electrical interconnect between those chips becomes a major performance bottleneck and power drain. Lightmatter's photonic approach could solve this fundamental physics constraint, enabling AI systems to scale more efficiently. The company represents one of several bets on novel computing architectures (alongside Cerebras' wafer-scale and Extropic's thermodynamic computing) that could complement or eventually challenge the dominant GPU paradigm.