When Chinese start-up DeepSeek released its latest artificial intelligence model last month, it moved Beijing closer to the future it has been trying to build for years.
In a small but meaningful break from American technology, DeepSeek said for the first time that its new model has been optimized to run on chips made by Chinese tech giant Huawei. This was a milestone in China’s long-term effort to develop advanced technologies at home and reduce its dependence on Western innovation.
While most of the world’s leading AI systems still rely on semiconductors from US chip giant Nvidia, Chinese AI firms are increasingly turning to domestic alternatives.
The timing of DeepSeek’s announcement — ahead of a planned summit between President Trump and Chinese leader Xi Jinping this week — gives Beijing new confidence in trade talks that U.S. controls on exports of Nvidia chips have not derailed China’s artificial intelligence development.
Any meaningful shift by China away from US AI technology could limit the impact of US export controls and deprive Washington of a critical source of leverage with Beijing. The prospect has gained urgency as DeepSeek’s AI technology has shaken up the US tech industry and turned the company into a powerful symbol of China’s quest for technological self-sufficiency.
Before the two leaders met last year, Mr Trump said he planned to discuss Nvidia’s most powerful AI chips with Mr Xi, fueling speculation that the United States could ease restrictions on the technology.
But after years of Washington barring Chinese companies from buying certain advanced technology products, firms like DeepSeek and Moonshot AI are starting to design their AI systems around the restrictions rather than waiting for them to disappear. This includes exploring how their models can run on a wider range of processors outside of Nvidia.
“US export controls are not freezing China’s AI development,” said Wei Sun, chief artificial intelligence analyst at Counterpoint Research in Beijing. “He’s forcing China to build an alternative stack.”
DeepSeek said its latest model can use Huawei chips for inference, a process that allows the AI system to respond faster and more accurately to users. Inference generally requires less computing power than training, the demanding process of teaching a model how to operate. DeepSeek still relied on Nvidia chips to train its system, according to two semiconductor industry people who were not authorized to comment publicly on the matter.
It was not immediately clear how DeepSeek gained access to the chips, although Chinese companies can still remotely use Nvidia chips located in data centers outside of China. DeepSeek did not respond to a request for comment.
Huawei said it plans to release a training chip this year. However, he also said it will take another year for his products to match the performance of Nvidia’s current offerings.
The growing rift between China’s and America’s AI infrastructure is the result of what Jensen Huang, Nvidia’s chief executive, has long warned would result from strict export controls.
He said the restrictions only prompted Chinese companies to accelerate efforts to create domestic alternatives, which could lead to a bifurcation of the market: Chinese AI systems running on Chinese chips while the West sticks to American hardware.
As the world’s dominant AI chip maker, Nvidia has the ability to gain unfettered access to China. But Mr. Huang argued that strict restrictions would ultimately hurt the United States by reducing its influence over China’s artificial intelligence industry.
Two months after his last meeting with Mr. Xi, Mr. Trump granted Nvidia permission to sell the H200, one of its most powerful chips, to China.
But since then, the chips have been squeezed between lawmakers in Washington, who seek closer oversight of their use in China, and Beijing, which has ordered Chinese tech companies to buy domestic chips.
Commerce Secretary Howard Lutnick told the Senate Appropriations Committee last month that no H200s have actually gone to China, and Nvidia said in regulatory filings this year that it has yet to generate any revenue from H200 sales there. Ahead of this week’s summit in Beijing, the fate of Nvidia’s chips in China is no clearer than it was at the last meeting between Mr Trump and Mr Xi.
Analysts expect China’s frustration with US export controls to be part of the discussion when the two leaders meet.
“Chip export control is a persistent issue that China has opposed,” said Jiang Tianjiao, an associate professor at Fudan University in Shanghai. But as China’s chip-making capabilities improve, officials may not want to interfere with efforts to reduce its reliance on U.S. technology, he said.
While Chinese tech companies continue to release high-performance artificial intelligence systems despite export controls, China’s push for technological self-sufficiency in chipmaking continues to run into significant obstacles. Semiconductor Manufacturing International Corporation, or SMIC, the Chinese company that makes some of Huawei’s chips, is having trouble mass-producing them. The chips it makes are more prone to defects and use more power than those made by foreign rivals.
Huawei’s solution has been to combine large numbers of these weaker chips to achieve the computing power of more advanced processors—a strategy that depends on SMIC’s ability to produce in high volumes. Still, Chinese chipmakers are still expected to make only a small fraction of the advanced semiconductors made by foreign companies such as Nvidia this year.
Before Washington tightened controls, many of Huawei’s chips were made by Taiwan Semiconductor Manufacturing Company, which makes most of the world’s advanced chips, including Nvidia chips.
Export controls have limited China’s ability to produce large volumes of the advanced chips needed for AI, said Dan Kim, director of strategy at Canadian research firm TechInsights and a Commerce Department official during the Biden administration. But he added that the same restrictions have also pushed Chinese tech companies to innovate in new ways.
Chinese companies are trying to redefine what determines success in the race to build cutting-edge artificial intelligence For years, the industry’s most advanced systems have come from companies that can afford to spend billions of dollars assembling huge numbers of powerful chips.
Companies like Huawei are now betting that success could one day depend less on amassing the most computing power and more on building an integrated ecosystem of chips, AI models and applications that is good enough for most real-world uses. By working closely with AI model developers like DeepSeek, Huawei can adapt its hardware to better support the software running on it.
When DeepSeek announced its latest model, Huawei said there was a “close collaboration of both sides’ chip and model technologies.”
In technical papers describing its models, DeepSeek outlined specific ways chipmakers could modify their products to improve the performance of their systems.
“DeepSeek is calling out to Huawei and other companies, ‘Please make these changes so we can get better performance out of your chips,'” said Jacob Feldgoise, an analyst at Georgetown University’s Center for Security and Emerging Technologies.
Xinyun Wu contributed reporting from Taipei.
