EDA Tools Expand Beyond Chips for AI Optics as Photonics Moves Into Volume Production
Updated
Updated · SemiEngineering · Jul 23
EDA Tools Expand Beyond Chips for AI Optics as Photonics Moves Into Volume Production
1 articles · Updated · SemiEngineering · Jul 23
Summary
Silicon photonics is moving into mainstream AI and data-center hardware, pushing EDA teams to verify full electro-optical systems rather than just photonic integrated circuits.
Traditional chip flows fall short because optical links must model waveguides, wavelength, phase, polarization, thermal drift, mechanical stress and optical-electrical conversion across die, package and system levels.
Current workflows still stitch together specialized photonic solvers, circuit simulators, layout tools and foundry-specific PDKs, leaving key tasks such as routing, DRC/LVS and functional verification less automated than in digital IC design.
70 to 80 picometers of ring-resonator drift per degree Celsius illustrates why package-aware thermal analysis and compact models must feed back into system design before tape-out.
EDA vendors and researchers say the next step is a unified, more automated flow—potentially using agentic AI—to carry accurate models from device physics through package and system sign-off for co-packaged optics at scale.
With light behaving continuously unlike digital signals, can AI truly automate the fragmented multi-physics nightmare of electro-optical chip design?
As AI pushes silicon photonics mainstream, will the extreme thermal sensitivity of co-packaged optics ultimately throttle massive data center scaling?
How will the semiconductor industry solve the massive yield bottleneck of complex fiber alignment testing before standardizing next-generation optical links?