Core Viewpoint - Nvidia is expected to re-enter the Chinese market after the U.S. government announced it would allow the sale of its H200 AI chips to China, subject to a 25% fee per chip [3][4]. Group 1: Nvidia's H200 Chip - The H200 chip, released in November 2023 and set to ship in 2024, is designed for training and inference of generative AI models, offering performance several times that of its predecessor, the H100 [4]. - The H200 chip features a memory speed of 4.8 TB per second and provides 141 GB of memory, nearly double the capacity of the A100, with a bandwidth increase of 2.4 times [4]. Group 2: Financial Impact and Market Dynamics - In Q2 of fiscal year 2026, Nvidia reported revenues of $46.743 billion, with $2.769 billion coming from the Chinese market, a decline of nearly $900 million compared to $3.667 billion in the same quarter of the previous fiscal year [5]. - Nvidia's CEO Jensen Huang noted that U.S. policies have led to a complete exit from the Chinese market, where the company's market share dropped from 95% to 0% [5]. - Huang emphasized the unique potential of the Chinese AI market, predicting it could reach approximately $50 billion in the next two to three years, highlighting the significant loss for U.S. companies unable to participate [5]. Group 3: Competitive Landscape - During Nvidia's absence, local AI chip manufacturers in China have accelerated their growth, with companies like Tencent and Baidu indicating sufficient GPU resources and reliance on domestic chips [6]. - Nvidia faces increasing competition from self-developed AI chip manufacturers, with alternatives like Google's TPU chips offering advantages in energy efficiency and performance [6]. - Competitors such as Anthropic plan to deploy up to 1 million Google TPUs for training their AI models, indicating a shift in the competitive landscape [6].
英伟达H200“附条件”对华出售,此前中国市场份额清零