2026-02-27 02:15Summary of Key Points from the Conference Call Industry and Company Involved - The discussion revolves around the supply chain for NVIDIA's LPU (Large Processing Unit) and its associated components, highlighting various suppliers and manufacturers involved in the production process [1]. Core Insights and Arguments - Foundry: TSMC is identified as the primary foundry for the LPU, utilizing N3 and N5 process technologies [1]. - SRAM Suppliers: The SRAM suppliers include Powerchip, Winbond, Etron, and Aip, all of which are Taiwanese companies. The LPU leverages on-chip SRAM to bypass the bottlenecks associated with HBM (High Bandwidth Memory) and CoWoS (Chip-on-Wafer-on-Substrate), thus reducing reliance on the three major HBM suppliers: Samsung, SK Hynix, and Micron [1]. - PCB Suppliers: The PCB (Printed Circuit Board) suppliers mentioned are Shenghong Technology and Unimicron [1]. - Resin Supplier: Dongcai Technology is noted as the resin supplier for the LPU [1]. - Q Fabric: The Q fabric supplier is identified as Fihua [1]. - CCL (Copper Clad Laminate) Suppliers: The main suppliers for CCL are Taiguang (primary supplier) and Shengyi Technology [1]. - Testing and Packaging: ASE (Advanced Semiconductor Engineering) is the main supplier for testing and packaging, along with Tongfu Microelectronics and Changdian Technology. The LPU does not require complex CoWoS advanced packaging and primarily uses traditional packaging methods [1]. - System Integration: The system integration for servers and cabinets is handled by companies such as Foxconn Industrial Internet, Quanta, and Wistron [1]. Other Important but Potentially Overlooked Content - The strategic choice to utilize on-chip SRAM is significant as it allows NVIDIA to mitigate risks associated with supply chain dependencies on HBM suppliers, which could be crucial in maintaining production efficiency and cost-effectiveness [1]. - The mention of traditional packaging methods indicates a potential shift in manufacturing strategies that could impact the overall cost structure and production timelines for NVIDIA's LPU [1].