Hua Hong Grace eyes silicon photonics expansion as AI chip shortages lift pricing power

Hua Hong Grace's silicon photonics push isn't a catch-up play—it's a strategic lock-in on the mature-node dividend in AI interconnect. SiPho core processes sit at 28nm-90nm, squarely in Hua Hong's sweet spot, a window TSMC and Intel are ceding as capacity tilts toward 2nm. The 147% profit spike reflects pricing power from AI-server PMIC supply-demand dislocation, not mere volume growth. Technical ripple: SiPho COB packaging will pull upstream EML laser and AWG demand, forcing optical module makers to reassess foundry concentration risk. On competition, GlobalFoundries' 45nm SiPho and Intel's Foveros are the main rivals, but both carry a customer-trust discount under the export-control overhang. Hua Hong's de-risking narrative holds structural appeal for Middle East and Southeast Asian AI infrastructure buyers. 12-24 month outlook: SiPho interconnect will migrate from 800G to 1.6T/3.2T. If Hua Hong secures hybrid-bonding capacity by 2027, it redefines "mature-node foundry" as "AI interconnect infrastructure provider"—shifting the valuation anchor from P/E to capacity-scarcity premium.