Tech Disruptors Lightmatter Taps Photonics to Solve AI at Scale
Feb 12, 2026
Nick Harris, Co-founder and CEO of Lightmatter and silicon photonics pioneer, explains using lasers and photonic interconnects to scale AI infrastructure. He discusses Guide, a VLSP light engine, and Passage, a photonic interconnect platform. Topics include near‑package vs co‑package optics, micro‑ring modulators for density, and partnerships that bring photonics into standard semiconductor workflows.
AI Snips
Chapters
Transcript
Episode notes
Bigger Chips Plus Faster IO
- Modern AI scale requires both ever-larger chips and extremely fast interconnect to make multiple chips behave like one.
- Lightmatter's Passage stacks chips and uses optics to provide the 100+ Tb/s IO needed for that illusion.
From Pluggables To Co‑Package Optics
- Silicon photonics is already shipping in pluggables but the optical engine is moving closer to the chip via near‑package then co‑package optics.
- Moving to single‑silicon transceivers reduces components and improves reliability dramatically.
Make Fiber Connections Replaceable
- Design detachable fiber connections so field technicians can replace optics without scrapping expensive chips.
- Prioritize serviceability to protect $100k GPUs from fiber installation errors.