China is a \u00a0crucial\u00a0 market for Nvidia. During the last fiscal year, which ended on January 26, 2025, the Asian nation contributed approximately \u00a013%\u00a0 to the company\u2019s total revenue, netting around \u00a0$17 billion\u00a0. In fact, China is Nvidia’s third-largest market, following the United States and Taiwan. However, recent actions taken by the Trump administration threaten Nvidia\u2019s foothold in this pivotal market.<\/p>\n
<\/p>\n
In mid-April, the U.S. Department of Commerce enforced new \u00a0export restrictions\u00a0 on GPUs meant for \u00a0artificial intelligence (AI)\u00a0, specifically targeting the H20 model. As a result, the distribution of these chips to Chinese customers came to a halt. After extensive negotiations, which even included several face-to-face meetings between Donald Trump and Nvidia CEO Jensen Huang, the Department of Commerce granted permission for Nvidia to resume supplying the H20 chip in China. However, this approval came at a cost: Nvidia is now obligated to pay \u00a015%\u00a0 of its revenue from these sales back to the U.S. government.<\/p>\n
<\/p>\n
AMD has faced similar regulatory hurdles, indicating that this strategy by the Trump administration sets a \u00a0new precedent\u00a0 that requires U.S. companies to remit a portion of their foreign sales back to the state. For over three years, Nvidia has navigated various \u00a0sales prohibitions\u00a0 regarding most of its AI chips in China and now faces this additional regulatory challenge. While the U.S. government claims it is safeguarding national interests and security, these actions inadvertently promote a \u00a0rapid acceleration\u00a0 of technology development within China\u2014a classic case of \u00a0self-sabotage\u00a0.<\/p>\n
<\/p>\n
The future of Nvidia in China largely rests on the decisions made by the \u00a0Cyberspace Administration of China (CAC)\u00a0, the primary regulatory body for Internet governance in the country. The CAC is currently conducting an in-depth investigation into the NVIDIA H20 GPU amid suspicions that it could contain hidden backdoor vulnerabilities. If these concerns are substantiated, access to this GPU in China could be severely restricted.<\/p>\n
<\/p>\n
In a bid to reclaim trust and mitigate growing skepticism, Nvidia\u2019s Security Director, David Reber Jr., recently published a blog post titled, \u201cThere are no backdoors in Nvidia chips. There are no kill switches. There are no spyware.\u201d This statement marks a strategic attempt by Nvidia to reassure stakeholders in China, underscoring the importance of the Chinese market to its overall operations. However, if the CAC’s decision turns unfavorable, it could lead to a significant revenue loss in one of Nvidia\u2019s key markets.<\/p>\n
<\/p>\n
If the final resolution of the CAC turns out to be unfavorable, NVIDIA will likely lose a substantial portion of its income from China.<\/p>\n<\/p><\/div>\n<\/div>\n
Furthermore, regardless of the CAC verdict, the Chinese government is encouraging domestic companies focused on \u00a0developing AI models\u00a0 to utilize locally sourced integrated circuits. This push gives \u00a0Huawei\u00a0 and \u00a0Moore Threads\u00a0 a competitive advantage. Huawei has been developing its AI GPUs, the \u00a0Ascend Ai\u00a0, for over five years, continually refining their performance to compete with Nvidia’s A100 and H100 chips.<\/p>\n
<\/p>\n
Currently, Huawei\u2019s most ambitious project is the \u00a0Ascend 910D\u00a0 chip, aiming to outperform the H100 GPU. Furthermore, they have introduced the \u00a0Ascend 920\u00a0, geared specifically to fill the market gap currently occupied by Nvidia’s H20 GPU. This chip is set for large-scale production during the second half of 2025, utilizing advanced \u00a06nm integration technology\u00a0, reportedly developed in collaboration with SMIC.<\/p>\n
<\/p>\n
On the other hand, Moore Threads has also developed several AI GPUs that could potentially rival those from Nvidia, AMD, or Huawei. Notably, their \u00a0MTT S4000\u00a0 and \u00a0MTT S3000\u00a0 cards are garnering interest. Interestingly, they also offer the \u00a0MTT S80\u00a0 card, which is marketed for gaming and content creation, boasting a calculation capacity of \u00a014.4 TFLOPS\u00a0 in single-precision floating point operations. While not astounding, it’s a competitive entry into the market.<\/p>\n
<\/p>\n
Image | Gage Skidmore<\/a> | Wikipedia<\/a><\/p>\n In the face of these challenges, Nvidia must navigate a complex landscape where regulatory compliance, market demands, and geopolitical tensions intertwine. The company’s ability to adapt will determine its success or failure in sustaining its revenue streams from one of the most significant markets for AI technology today.<\/p>\n