3 6 Ke·2025-10-21 02:15Core Insights - The core focus of global AI competition is on chip manufacturing, with NVIDIA and TSMC unveiling the first Blackwell chip wafer made in the U.S., marking a significant milestone in the industry and symbolizing the return of advanced manufacturing to the U.S. [1][3][10] Group 1: Chip Manufacturing - The Blackwell chip wafer represents the first AI chip produced domestically in the U.S., indicating a shift in the global tech landscape towards local manufacturing [1][5] - TSMC's Arizona factory will produce advanced chips, including 2nm, 3nm, and 4nm technologies, which are crucial for AI, telecommunications, and high-performance computing applications [8][10] Group 2: Investment and Economic Impact - NVIDIA's CEO, Jensen Huang, previously committed $500 billion to U.S. AI chip manufacturing, emphasizing the importance of domestic production [5] - The total investment in the Arizona facility has reached an impressive $165 billion, highlighting the scale of this manufacturing initiative [5][10] Group 3: Technological Advancements - The Blackwell architecture features approximately 208 billion transistors and utilizes a custom 4NP process developed in collaboration with TSMC [14] - Blackwell GPUs consist of two interconnected chips, enabling a high bandwidth connection of up to 10TB/s, allowing for enhanced performance [14] Group 4: Future Developments - Blackwell is currently in full production, with plans for subsequent models, including Blackwell Ultra and the Vera Rubin superchip, aimed at addressing larger model inference and training needs [16] - The roadmap indicates that the next architecture, Feynman, is expected to be launched in 2028, with an interim release of Rubin Ultra in 2027 [16]