超预期,AI算力需求爆发!寒武纪涨超2%,科创芯片50ETF(588750)连续3日“吸金”超7800万!AI芯片需求空间多少?数据说话

Group 1 - The A-share market experienced fluctuations on July 31, with the Sci-Tech Chip sector showing a slight increase, as the Sci-Tech Chip 50 ETF (588750) rose by 0.09% and saw a net subscription of over 78 million yuan over three consecutive days [1] - Most component stocks of the Sci-Tech Chip 50 ETF experienced a pullback, with notable performances including Dongxin Co., which hit the daily limit with a 20% increase, and Yuanjie Technology, which rose over 9% [3] - The top ten component stocks of the Sci-Tech Chip 50 ETF include companies like Cambrian (2.47% increase) and Zhongben International (-0.30% decrease), with significant trading volumes reported [3] Group 2 - Global tech giants reported better-than-expected earnings, with a leading software provider announcing Q4 FY2025 revenue of $76.44 billion, up from $64.7 billion year-on-year, and a net profit of $27.233 billion, up from $22.036 billion [4] - The cloud computing segment of the leading software provider saw a 39% increase in sales, with 16% attributed to AI growth, indicating a strong demand for AI infrastructure [4] - The AI computing demand is expected to drive the semiconductor equipment supply chain, with the Sci-Tech Chip sector likely to benefit from this upward trend [4] Group 3 - Dongwu Securities expressed optimism regarding the development of AI computing demand, highlighting the significant reduction in costs for large model calls and the continuous improvement of inference and training chips [5] - Predictions from various sources indicate that the demand for AI chips will continue to rise exponentially in the coming years, with estimates suggesting that China's intelligent computing scale could reach 1,271.4 EFLOPS by 2026 [5] - The AI chip market in China is projected to reach 120.6 billion yuan, with more optimistic forecasts suggesting it could exceed 300 billion yuan [5] Group 4 - The demand for AI computing is being driven by both C-end and B-end applications, with C-end applications like AI programming and video/audio generation showing robust growth [6] - B-end applications are focusing on customized agents, which are becoming increasingly important for various industries, with cloud vendors and SaaS companies prioritizing agent development [6] - Domestic IC design companies are beginning to integrate inference cards into the local supply chain, which is expected to benefit from advancements in packaging and technology iterations [6]