BEIJING — China has ordered data centers funded by the state to stop using foreign made artificial intelligence (AI) chips, according to two people familiar with the matter, marking one of Beijing’s most assertive steps to cut reliance on US technology.
The move, reported to have been quietly issued in recent weeks, underscores China’s push toward semiconductor self sufficiency amid enduring geopolitical tensions with Washington.
Regulatory authorities instructed that new data center projects receiving any government support must exclusively use domestically produced AI chips, the sources told Reuters.
Projects still under construction those less than 30 percent complete must remove any foreign chips already installed or cancel pending purchases, while more advanced projects will be reviewed on a case by case basis.
The order arrives at a sensitive moment in US China relations. Washington has tightened export restrictions on advanced semiconductors and chipmaking tools to limit Beijing’s access to the hardware essential for training large scale AI models.
The restrictions have already barred Nvidia, a global leader in AI chips, from selling its most powerful processors to Chinese firms. China’s guidance on chip procurement appears to be a direct response to these measures. It also aligns with President Xi Jinping’s long stated goal of achieving “independent controllability” in key technologies.
“The leadership views AI chips as the foundation of national competitiveness,” said Dr. Li Yuxin, a semiconductor policy researcher at Tsinghua University. “By localizing chip supply, China hopes to shield its data infrastructure from future sanctions.”
The directive could severely impact Nvidia’s business in China, a market it has dominated for years. It also threatens to curtail opportunities for other US chipmakers such as AMD and Intel, which supply processors to many of China’s largest data centers.
Analysts say the order marks a major escalation in China’s efforts to insulate its critical sectors from Western technology dependencies.
This is not just about chips, said Ben Harcourt, an Asia Pacific technology analyst at Meridian Research in Singapore. “It’s about controlling the full stack of AI infrastructure from hardware and algorithms to data and applications.”
Nvidia, which generated billions in annual revenue from Chinese customers before the export restrictions, may face a near total freeze out from future government funded projects.
While US President Donald Trump recently said Washington would “let them deal with Nvidia but not in terms of the most advanced” chips, the Chinese government’s latest move effectively removes even lower tier Nvidia hardware from eligible projects.
“The writing has been on the wall for Nvidia,” said Harcourt. “Beijing’s clear intent is to replace imported chips with those made by Huawei, Biren Technology, and other domestic players. This policy gives them a guaranteed market.”
According to a Reuters analysis of government tenders, AI related data center projects in China have attracted over $100 billion in public and private funding since 2021.
The majority of these projects received some level of state support meaning the directive could affect most of the country’s large-scale data infrastructure developments.
For comparison, US based companies such as Amazon, Microsoft, and Google spent a combined $200 billion globally on data centers in the same period, but with far more diversified supply chains.
China’s new rule could result in short term disruptions, analysts say, as domestic chipmakers scramble to meet demand for replacements.
Huawei’s Ascend and Kunpeng processors are among the few domestic alternatives currently available. However, experts note that these chips lag behind Nvidia’s H100 and A100 series in terms of computing performance and energy efficiency.
“China’s biggest challenge is bridging the performance gap,” said Zhao Qiang, an engineer at a Beijing based AI startup. “Domestic chips work fine for inference tasks, but for training large models, they’re not yet at par.”
Reactions among industry insiders in China are mixed. Some see the policy as a necessary step toward long term independence, while others worry about near term economic consequences.
A manager at a Shanghai data center construction firm, who asked not to be named for fear of repercussions, said two projects had been delayed after authorities revoked approval to use Nvidia hardware.
“We were 20 percent into installation when the order came through,” he said. “Now everything is on hold until we source approved domestic alternatives.”
Meanwhile, AI researchers are adjusting to the changing landscape. “It will slow us down, no doubt,” said Dr. Cheng Wen, a computer scientist at a Guangzhou university lab. “But if this pressure accelerates local innovation, it might be a blessing in disguise.”
The extent of the directive remains unclear. Sources told Reuters that it may not apply nationwide and that implementation could vary by province.
Neither the Cyberspace Administration of China nor the National Development and Reform Commission has commented on the guidance, and major US chipmakers have declined to issue statements.
Still, experts believe the policy reflects a broader trend that will continue regardless of short term political fluctuations. “China’s path toward technological decoupling is now irreversible,” said Harcourt.
Every new restriction or sanction only reinforces Beijing’s determination to build its own ecosystem. If the policy expands to private sector or cloud service providers, the ripple effects could extend well beyond state projects, reshaping the global AI supply chain and accelerating the bifurcation between US and Chinese technology spheres.
China’s ban on foreign AI chips in state-funded data centers signals a pivotal shift in its technology policy one aimed at safeguarding national data sovereignty and bolstering its domestic semiconductor industry.
While the move may disrupt current projects and foreign suppliers like Nvidia, it reinforces Beijing’s long term strategy to achieve AI chip self reliance, regardless of the immediate economic cost.
As global competition for computing power intensifies, China’s decision underscores how the race for AI dominance has evolved from corporate rivalry to a geopolitical contest shaping the next era of technological order.