Innolight starts US NPO pilot production with Tower for AI optical interconnects

Innolight's decision to build on Tower's PH18M silicon photonics platform rather than develop a proprietary optical engine is a calculated time-buying move. NPO is the pragmatic bridge architecture before CPO reaches volume maturity—shorter deployment cycles, broader ecosystem compatibility, but ultimately a transitional state. The real prize is IP ownership at the optical-engine-to-switch-ASIC interface, which dictates BOM pricing power in next-gen AI clusters. The California production line carries strategic weight beyond capacity. By routing through an Israeli foundry's 'non-sensitive' entity status, a leading Chinese optical module maker secures access to advanced silicon photonics manufacturing before US export controls tighten further. If Washington extends restrictions to silicon photonics components in 2025-2026, this compliance corridor could close within 18 months. On the competitive front, Broadcom's Bailly and NVIDIA's Quantum-X have already locked in hyperscaler commitments. Innolight's NPO play is less a technology bet than a speed-to-market strategy—trading long-term architectural dominance for near-term ecosystem positioning. The 12-24 month inflection point won't be NPO versus CPO. It will be which player establishes the standardized IP framework for optical interconnect, turning a component category into a platform dependency.