The Chinese government has ordered that all new data center projects receiving any level of state funding must use only domestically produced artificial intelligence (AI) chips.
According to two sources familiar with the directive, and shared with Reuters in an exclusive, the decision marks one of Beijing’s boldest efforts yet to remove foreign technology from its critical digital infrastructure and accelerate the nation’s drive toward AI self-sufficiency.
Targeting foreign technology
In Reuters’ report, it states that regulators instructed data centers that are less than 30% complete to remove all foreign AI chips or cancel plans to purchase them. Projects already nearing completion will be reviewed on a case-by-case basis, the sources said.
The new guidance, which has not been publicly announced, is expected to affect major U.S. semiconductor firms such as Nvidia, AMD, and Intel. U.S. President Donald Trump said Washington will let them deal with Nvidia but not in terms of the most advanced chips, in an interview aired on Sunday following talks with Chinese President Xi Jinping.
The move could undermine Nvidia’s hopes of regaining market share in China, while giving local rivals like Huawei Technologies a stronger foothold in the rapidly expanding AI infrastructure market.
Growing divide
The order comes as trade relations between the two superpowers remain tense despite temporary pauses in hostilities. Washington has long restricted exports of advanced AI chips to China, citing concerns that such technology could strengthen the Chinese military’s computing capabilities.
In response, Beijing has accelerated efforts to cultivate homegrown semiconductor champions. The new data center restrictions follow earlier steps, including the 2023 ban on Micron Technology products from critical infrastructure, which paved the way for the U.S. memory giant’s exit from China’s server chip market.
China’s push to replace foreign components in its most strategic sectors — energy, telecommunications, and now AI — reflects its determination to reduce dependence on Western technology amid growing geopolitical uncertainty.
Massive state investment
AI data center projects in China have attracted more than $100 billion in state funding since 2021, according to a Reuters review of government tenders. Most large data centers in China have received at least some level of government financing, meaning the new rule could apply to the vast majority of current projects.
Some projects have already been halted. One facility in a northwestern province that had planned to use Nvidia chips has been suspended before breaking ground, one source said.
Regulatory bodies believed to be involved include the Cyberspace Administration of China and the National Development and Reform Commission, though neither responded to Reuters’ requests for comment. Nvidia and AMD did not reply, while Intel declined to comment.
Impact on Nvidia and U.S. firms
The guidance reportedly covers Nvidia’s H20 chips, the most advanced AI processors the company is permitted to sell in China, as well as the higher-end B200 and H200 models, which are already restricted under U.S. export controls but are often available through grey-market channels.
Nvidia’s market share in China’s AI chip segment has collapsed from 95% in 2022 to zero today, according to company figures. CEO Jensen Huang has lobbied Washington to ease export rules, arguing that maintaining China’s dependence on U.S. chips aligns with American strategic interests.
However, Beijing’s latest move could eliminate any remaining foothold for U.S. chipmakers in state-backed projects, further reducing revenues from one of the world’s largest AI markets.
Opportunity and risk for Chinese chipmakers
For domestic firms, the directive represents both an opportunity and a challenge. Companies such as Huawei, Cambricon Technologies, and startups like MetaX, Moore Threads, and Enflame, are now positioned to fill the supply gap.
While their products increasingly rival some of Nvidia’s mid-range offerings, they have struggled to achieve widespread adoption due to limited software compatibility and developer trust. The new order could force a shift by compelling state projects to use these local alternatives.
Yet analysts warn that China risks widening its technological gap with the U.S. in the short term. American giants like Microsoft, Meta, and OpenAI are pouring hundreds of billions into AI data centers powered by Nvidia’s most advanced chips — hardware that Chinese firms cannot yet match in performance or availability.
Over in Germany, tech giants Nvidia and Deutsche Telekom are building a €1 billion ($1.14 billion) AI facility that aims to change how European companies tap into AI.