Mei Ri Jing Ji Xin Wen·2026-02-27 08:45Core Viewpoint - The dramatic market reactions highlight a shift in perception regarding AI chip demand, with Chinese AI models gaining traction and impacting the valuation of companies like Nvidia, which saw a significant drop in market capitalization following its earnings report [1][3][22]. Group 1: Nvidia's Earnings and Market Reaction - Nvidia reported a record Q4 revenue of $68.1 billion, a 73% increase year-over-year, with an adjusted gross margin of 75.2%, exceeding market expectations [3][22]. - Despite strong earnings, Nvidia's stock fell by 5.5%, resulting in a market cap loss of nearly $260 billion (approximately ¥1.77 trillion), marking its largest single-day drop since April 2025 [1][3]. - The decline in Nvidia's stock also affected other chip manufacturers, including Broadcom, AMD, and TSMC, which experienced varying degrees of stock price decreases [3][22]. Group 2: Rise of Chinese AI Models - Data from OpenRouter indicates that, as of February 2026, the weekly API call volume for Chinese AI models surpassed that of the U.S. for the first time, reaching 4.12 trillion tokens compared to the U.S.'s 2.94 trillion tokens [16][19]. - The call volume for Chinese models further increased to 5.16 trillion tokens in the following week, representing a 127% growth over three weeks [16][19]. - Four out of the top five global AI models by call volume are now from China, showcasing a "cluster rise" in the Chinese AI sector [19][22]. Group 3: Market Dynamics and Future Outlook - The market is reassessing the distribution of computational power value, moving away from a linear dependence on high-end GPUs, as Chinese models utilize more efficient architectures [3][22]. - The "Mixture-of-Experts" (MoE) architecture used by many Chinese models significantly reduces the computational cost per token, with a reported 60% reduction in memory usage and up to 19 times improvement in throughput [25][27]. - The cost of processing tokens with Chinese models is substantially lower, with prices around $0.3 per million tokens compared to $5 for comparable overseas products, indicating a cost advantage for Chinese AI [27][28]. Group 4: Implications for Domestic Infrastructure - The success of Chinese AI models is driving a massive demand for domestic computational infrastructure, as global developers increasingly favor these models [22][27]. - China's lower electricity costs, which can be as low as ¥0.2-0.3 per kilowatt-hour compared to $1-1.5 in the U.S. and Europe, further enhance the competitiveness of Chinese AI models [27][28]. - Morgan Stanley predicts a compound annual growth rate of 330% for China's token consumption from 2025 to 2030, suggesting a potential 370-fold increase in just five years [27][28].