是说芯语·2026-02-02 23:56Core Viewpoint - The semiconductor industry is facing unprecedented demand for advanced process wafers driven by AI model training, necessitating significant capacity expansion from TSMC to meet the needs of companies like NVIDIA [1][4]. Group 1: NVIDIA's Demand and TSMC's Response - NVIDIA's CEO Jensen Huang emphasized the urgent need for TSMC to double its advanced process wafer production to meet the explosive demand for AI training and inference chips, predicting a potential growth of over 100% in TSMC's capacity over the next decade [1][4]. - Huang estimated that NVIDIA will require approximately 1 million advanced process wafers annually by 2035, while TSMC's current monthly capacity for 12-inch wafers is about 1.5 million, with only 500,000 at advanced nodes [3][4]. Group 2: TSMC's Capital Expenditure Plans - In response to Huang's call for increased production, TSMC has raised its capital expenditure for 2026-2030 to $100 billion, a 30% increase from previous plans, focusing on expanding 3nm capacity at Fab 20, building a 2nm line at Fab 22, and establishing advanced processes at Fab 21 in Arizona [4][5]. - Despite this significant investment, TSMC's plans still fall short of NVIDIA's demand projections, indicating a growing supply-demand gap in the semiconductor market [4]. Group 3: Opportunities for Chinese Semiconductor Companies - TSMC's expansion efforts are creating unprecedented opportunities for domestic semiconductor equipment and materials manufacturers in China, with companies like AMEC and North Huachuang entering TSMC's procurement lists for new equipment [5]. - The expansion is expected to generate a procurement demand in the hundreds of billions, as Chinese firms make technological breakthroughs in specific segments, positioning them as key players in this capacity race [5][6]. Group 4: Global Semiconductor Landscape Transformation - Huang's warnings signal the beginning of a global semiconductor "computing power arms race," necessitating a complete restructuring of the supply chain to meet the surging demand for AI chips [6]. - The competition will span all aspects of the semiconductor industry, from equipment manufacturers to material suppliers, and will determine which companies can capitalize on the AI computing power era's opportunities [6].