Industry Analysis
UBS’s findings reflect a systemic shift driven by escalating compute density: as GPU scaling nears physical limits, AI infrastructure demands holistic innovation. High-frequency PCBs (e.g., Shenghong’s Rubin bet), optical interconnects replacing copper, and SiC-based thermal solutions are no longer optional—they redefine material stacks from LCP films to silicon photonics. Geopolitically, looming U.S. controls on advanced packaging and liquid cooling could inflate Chinese OEMs’ BOM costs by over 15%, given reliance on Japanese and American SiC/cooling equipment. TSMC and Samsung will likely localize CoWoS capacity faster, while SMIC accelerates domestic SiC epitaxy validation. Within 18 months, AI servers will pivot from GPU-centric to thermo-electro-optical co-design, opening a narrow window for second-tier material suppliers—though technological moats remain fluid, signaling imminent consolidation.
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.