us chip export controls: will china's ai innovation efforts succeed?
shanghai, Thursday, 10 July 2025.
us export controls on ai chips are pushing china to develop its own semiconductor industry. this move impacts nvidia’s market share and opens doors for new competitors. china’s ai chip market is projected to reach $453 billion by 2030, growing annually by 14% from 2025. investors should watch china’s progress in chip tech, as it could cut reliance on us tech. one model, Llama 3.1-405B, needs a massive 16,384 nvidia h100 nodes and 2,142 hours for training.
us restrictions and china’s response
Since October 2022, the U.S. government has tightened export controls, restricting China’s access to advanced computing chips and semiconductor manufacturing equipment [1]. These measures aim to curb China’s ability to develop supercomputers and advanced AI technologies [1]. The Bureau of Industry and Security (BIS) further expanded these controls in December 2024, adding more semiconductor manufacturing equipment and software tools to the restricted list [1]. These restrictions have spurred Chinese firms to seek domestic alternatives, impacting companies like Nvidia, which had $80 billion in revenue from data center GPUs in 2024 [1].
nvidia’s modified chips for china
Nvidia plans to launch a modified version of its Blackwell RTX Pro 6000 chip in September, designed to comply with U.S. export rules [6]. This new chip removes high bandwidth memory (HBM) and NVLINK, features that enhance data transfer speeds but violate export laws [6]. Nvidia is awaiting U.S. government approval to ensure the chip meets all regulations [6]. Chinese customers are currently testing sample units and initial feedback is positive, indicating potential for large orders [6]. This approach reflects Nvidia’s attempt to balance regulatory compliance with market opportunities in China [6].
china’s ai ambitions and data centers
Despite U.S. restrictions, Chinese companies plan to install over 115,000 Nvidia AI chips in about 36 data centers in western China [4]. These data centers, located in areas like Xinjiang, aim to boost China’s computing capabilities for training large language models [4]. One project by Nyocor Co. involves a data center utilizing 625 H100 servers [4]. Bloomberg estimates completing these projects would require over 14,000 data servers or 115,000 Nvidia H100 or H200 chips [4]. These efforts highlight China’s determination to advance its AI sector despite limited access to U.S. technology [4].
southeast asia as a point of contention
The U.S. government is drafting new rules to restrict AI chip exports to Malaysia and Thailand, fearing China might use these countries as a workaround to obtain advanced AI chips [5]. This move aims to prevent China from circumventing export controls through Southeast Asia [5]. There are concerns that restricted GPUs are being rerouted to China via Singapore, where Nvidia has a significant revenue share [5]. These concerns reflect the ongoing geopolitical tensions and the U.S.’s efforts to contain China’s technological advancement [5].
market impact and competition
Nvidia’s CEO, Jensen Huang, acknowledged that U.S. export restrictions have backfired, accelerating China’s development of domestic AI technology [6]. He noted Nvidia’s market share in China has decreased from 95% to 50% in four years [6]. Despite this, Nvidia sees a $50 billion AI market opportunity in China and is striving to maintain its presence [6]. Chinese tech giants like Alibaba and Tencent are testing domestic alternatives, signaling increased competition for U.S. chipmakers [6]. The evolving landscape necessitates that investors closely monitor the competitive dynamics and technological advancements within China’s semiconductor industry [1].