Optical Computing

The optical computing field in 2025-2026 operates on two timescales. The near-term story is photonic interconnects: Lightmatter's Passage L200 (1.6 Tbps/fiber, 200+ Tbps/package) and Ayar Labs' first UCIe-compliant optical chiplet (SuperNova, 8 Tbps, 16-wavelength WDM) are solving the most urgent bottleneck in AI infrastructure — the bandwidth wall between accelerators. Ayar Labs' TSMC COUPE partnership and $500M Nvidia-backed raise signal that optical I/O is moving from specialized AI infrastructure to broad semiconductor ecosystem adoption. The deployment timeline is crystallizing: early adopters 2026, broader adoption 2027, standard for AI networking by 2028.

The longer-term story is photonic compute itself: neuromorphic photonic tensor cores promise 880 TOPS/mm^2 and 5.1 TOPS/W (1-3 orders of magnitude over digital), but the gap between lab demonstrations and wafer-scale production remains substantial.

Frontier — What's Moving Now

• UCIe optical chiplet standardized — Ayar Labs first UCIe-compliant optical chiplet, TSMC COUPE partnership, $500M Nvidia-backed raise.
• Co-packaged optics going mainstream — Lightmatter's record bandwidth + Qualcomm partnership + Ayar Labs standardization.
• 2026-2028 deployment timeline — Early CPO adopters 2026, 800G/1.6T in 2027, standard by 2028.
• Photonic tensor cores project massive density gains — 880 TOPS/mm^2 projected, but lab-to-production gap is real.
• Manufacturing is the bottleneck — Yield, InP supply, workforce, not physics.

Concept Map

Concepts

Entities

Timeline

See timeline.md for chronological developments (2019 through 2026).

Research Frontier

See frontier.md for active research directions, breakthroughs, and knowledge gaps.

Sources