The administration led by \u00a0Donald Trump\u00a0 has realized that her predecessor and she have made a very serious mistake. The packages of \u00a0sanctions\u00a0 deployed over the last three years to prevent China from acquiring the most advanced \u00a0semiconductor manufacturing equipment\u00a0 and cutting-edge chips for \u00a0artificial intelligence (AI)\u00a0 have forced the Xi Jinping government to become independent from foreign technologies.<\/p>\n
<\/p>\n
The \u00a0Chinese integrated circuits industry\u00a0 has made significant advancements in the last five years and will likely continue to do so. By \u00a02026\u00a0, it is expected that Chinese manufacturers will possess their own \u00a0extreme ultraviolet lithography equipment (EUV)\u00a0. Currently, companies like \u00a0Huawei\u00a0 and \u00a0Moore Threads\u00a0 are producing GPUs that compete with Nvidia and \u00a0AMD\u00a0 chips. The \u00a0independence\u00a0 of the Chinese semiconductor industry is becoming a reality.<\/p>\n
<\/p>\n
The best approach for the U.S. in this situation would be to supply China with \u00a0advanced AI chips\u00a0 in a controlled manner, albeit less powerful than those manufactured by Nvidia or AMD. Such a strategy could alleviate some of China\u2019s ambitions for rapid development and autonomy. This appears to be the direction of the Trump administration, which has allowed Nvidia to reintroduce its \u00a0H20 GPU\u00a0 to the Chinese market. As highlighted by \u00a0Chris Miller\u00a0 in his \u00a0newsletter\u00a0, “The War of Chips,” this action may mitigate tensions between the two economies.<\/p>\n
<\/p>\n
The future of \u00a0Nvidia\u00a0 in China is heavily contingent on the \u00a0Cyberspace Administration of China (CAC)\u00a0, the country’s leading internet regulatory body. The CAC is currently scrutinizing the \u00a0Nvidia H20 GPU\u00a0, expressing concerns that it might contain \u00a0backdoors\u00a0 difficult for local experts to detect. If such suspicions prove valid, China might reconsider its dealings with this GPU.<\/p>\n
<\/p>\n
In an effort to bolster the company’s reputation amid growing distrust in China, \u00a0David Reber Jr.\u00a0, Nvidia’s Security Director, published an article on the company’s blog stating that there are “no backdoors in Nvidia chips, no kill switches, and no spyware.” The stakes are high; if the CAC’s final ruling is unfavorable, Nvidia could lose a substantial portion of its revenue from this pivotal market.<\/p>\n
<\/p>\n
The Chinese government is urging local companies to utilize homegrown integrated circuits in their AI development.<\/p>\n<\/p><\/div>\n<\/div>\n
Regardless of the CAC’s decision, it\u2019s clear that the Chinese government is actively encouraging local companies engaged in developing large AI models to utilize \u00a0Chinese-manufactured integrated circuits\u00a0 in their servers. Key players in this arena include \u00a0Huawei\u00a0 and \u00a0Moore Threads\u00a0. Huawei, for instance, has been refining its \u00a0Ascend AI chips\u00a0 for more than five years, working to match or even surpass Nvidia’s \u00a0A100\u00a0 and \u00a0H100\u00a0 chips.<\/p>\n
<\/p>\n
Huawei’s most ambitious initiative is the \u00a0Ascend 910D chip\u00a0, which aims to outperform the H100 GPU. Additionally, the company recently introduced the \u00a0Ascend 920 chip\u00a0, targeted to fill the market gap left by the \u00a0H20 GPU\u00a0 in China. This new chip is slated for large-scale production in the latter half of \u00a02025\u00a0, utilizing advanced \u00a06nm\u00a0 integration technology reportedly developed in collaboration with \u00a0SMIC\u00a0 and other partners.<\/p>\n
<\/p>\n
Meanwhile, \u00a0Moore Threads\u00a0 has also made strides with several AI-centric GPUs that claim to rival advanced solutions from Nvidia, AMD, and Huawei. The \u00a0MTT S4000\u00a0 and \u00a0MTT S3000\u00a0 are among their most promising products. Interestingly, the company’s \u00a0MTT S80 card\u00a0, designed for gaming and content creation, boasts a calculation capacity of \u00a014.4 TFLOPS\u00a0 in single precision floating-point operations. While this may not be groundbreaking, it showcases the company’s growing aspirations.<\/p>\n
<\/p>\n
Image | Gage Skidmore<\/a> | Wikipedia<\/a><\/p>\n More information | Financial Times<\/a><\/p>\n In Xataka | Ten Chinese companies in Chips and AI have allied with a common goal: to put an end to the dominance of Nvidia.<\/p>\n