Before this week’s Sino-American summit, Beijing reached a significant milestone in its pursuit of technological self-sufficiency, writes the NYT.
When the Chinese startup DeepSeek released its latest artificial intelligence model last month, it brought Beijing closer to a future it has sought to build for years.
In a small but meaningful break from American technology, DeepSeek announced for the first time that its new model was optimized to run on chips produced by the Chinese tech giant Huawei. This is a milestone in China’s long-standing commitment to develop advanced technologies domestically and reduce its dependence on Western innovation.
THE TIMING OF DEEPSEEK’S ANNOUNCEMENT
Although most of the world’s most advanced artificial intelligence systems still rely on semiconductors from the U.S. giant Nvidia, Chinese companies in the sector are increasingly turning to domestically produced alternatives. The timing of DeepSeek’s announcement, made ahead of this week’s scheduled summit between President Trump and Chinese leader Xi Jinping, gives Beijing renewed confidence ahead of trade talks, demonstrating that U.S. export controls on chips
Nvidia has not hindered the development of artificial intelligence in China.
A significant shift away from American AI technology by China could limit the impact of U.S. export controls and deprive Washington of a key leverage point against Beijing. This prospect has become more pressing since DeepSeek’s AI technology shook the U.S. tech industry and turned the company into a powerful symbol of China’s push for technological self-sufficiency.
Before last year’s meeting between the two leaders, Trump had said he wanted to discuss Nvidia’s most powerful AI chips with Xi, fueling speculation about a possible easing of U.S. restrictions on this technology. But after years of Washington blocking Chinese companies from purchasing certain advanced tech products, companies like DeepSeek and Moonshot AI are beginning to design their AI systems with these limitations in mind, rather than waiting for them to disappear.
This includes exploring how their models can run on a broader range of processors beyond Nvidia’s. “U.S. export controls are not blocking AI development in China,” said Wei Sun, lead AI analyst at Counterpoint Research in Beijing. “They are forcing China to build an alternative platform.” DeepSeek stated that its latest model can use Huawei chips for inference, the process that allows an 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.
THE ROLE OF NVIDIA CHIPS
According to two semiconductor industry sources who were not authorized to speak publicly on the matter, DeepSeek still relied on Nvidia chips to train its system.
It was not immediately clear how DeepSeek accessed those chips, although Chinese companies can still remotely use Nvidia chips hosted in data centers outside China. DeepSeek did not respond to a request for comment.
The growing divergence between Chinese and American AI infrastructures is a consequence that Jensen Huang, Nvidia’s CEO, had long anticipated due to strict export controls. He stated that the restrictions have only pushed Chinese companies to accelerate efforts to build national alternatives, which could lead to a bifurcated market: Chinese AI systems running on Chinese chips while the West continues to use American hardware.
As the world’s leading producer of AI chips, Nvidia would benefit from unrestricted access to the Chinese market. However, Mr. Huang argued that strict restrictions would ultimately harm the United States by reducing its influence over China’s AI 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. Since then, however, these chips have come under scrutiny from Washington lawmakers calling for tighter controls on their use in China, and from Beijing, which has mandated Chinese tech companies to buy domestically produced chips. Commerce Secretary Howard Lutnick told the Senate Appropriations Committee last month that no H200 chips had actually arrived in China, and Nvidia stated in regulatory filings this year that it had yet to generate any revenue from H200 sales in that country. Ahead of this week’s summit in Beijing, the fate of Nvidia chips in China remains as uncertain as it was at the last Trump-Xi meeting.
HUAWEI’S BET
Chinese companies are seeking to redefine the criteria for success in the race to build cutting-edge AI systems. For years, the most advanced systems in the industry were built by companies that could afford to spend billions assembling a large number of powerful chips.
Today, companies like Huawei bet that success might one day depend less on accumulating maximum computing power and more on creating 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 customize its hardware to best support the software running on it.
When DeepSeek announced its latest model, Huawei stated there had been “close collaboration between the two parties regarding chip and model technologies.” In technical documents describing its models, DeepSeek outlined specific ways chip manufacturers could modify their products to improve performance with its models.
(Excerpt from eprcomunicazione press review)
