China Mandates Domestic AI Chips in State-Funded Data Centers, Dealing Blow to Nvidia

The Chinese government has directed that all new data center projects receiving any public funding must exclusively use AI chips manufactured within the country, according to two individuals familiar with the policy.

In recent weeks, regulators have instructed data centers under 30 percent complete to strip out any foreign chips already installed or abandon plans to acquire them. Projects further along will face individual reviews, the sources added.

This directive marks one of Beijing’s boldest efforts to purge foreign technology from vital infrastructure, even as trade tensions with the United States ease temporarily. It accelerates China’s drive toward self-reliance in AI semiconductors.

Restrictions on advanced AI chips, particularly those from Nvidia, have fueled longstanding disputes between Washington and Beijing over control of cutting-edge computing resources.

In a recent interview, U.S. President Donald Trump indicated that following discussions with Chinese President Xi Jinping, the United States would permit China to acquire Nvidia products but not the most sophisticated models.

Beijing’s latest action, however, undermines Nvidia’s prospects of reclaiming market share in China and creates fresh opportunities for domestic competitors, led by Huawei.

The policy’s scope—whether national or limited to specific regions—remains unclear. The issuing authorities were not specified, and the sources spoke anonymously due to the matter’s sensitivity.

Other foreign suppliers of data center chips to China include AMD and Intel. Requests for comment from China’s Cyberspace Administration, National Development and Reform Commission, Nvidia, and AMD went unanswered; Intel declined to respond.

Nvidia Emerges as Primary Loser

Since 2021, Chinese AI data center initiatives have attracted more than $100 billion in government support, based on an analysis of public procurement records. The majority of data centers in the country benefit from some level of state assistance, though the exact number affected by the new rules is unknown.

At least one planned facility in northwestern China, developed by a private firm with public backing and intended to house Nvidia hardware, has been halted before construction began, one source confirmed.

Beijing has grown increasingly frustrated with U.S. export curbs designed to slow its technological advancement, prompting a range of countermeasures to reduce dependence on American components.

Washington defends the controls by claiming the chips could enhance Chinese military strength.

Earlier this year, China advised its major technology companies against buying high-end Nvidia processors over security risks and showcased a data center running entirely on homegrown AI chips.

In 2023, authorities barred Micron products from critical infrastructure, contributing to the U.S. firm’s decision this year to withdraw from China’s server memory market.

Nvidia’s chief executive, Jensen Huang, has actively urged President Trump and his administration to ease sales restrictions to China, contending that keeping a rival dependent on U.S. technology serves American interests.

The company’s current share of China’s AI chip market stands at zero, down from 95 percent in 2022.

Excluding foreign suppliers from major government-backed projects would erase a substantial revenue stream for Nvidia, even if an agreement allows renewed sales of advanced chips.

The guidelines encompass Nvidia’s H20 processor—the most capable model permitted for export to China under U.S. rules—as well as higher-performance units like the B200 and H200, which remain accessible in China via unofficial channels despite export bans.

Advantage and Challenges for Chinese Chipmakers

The policy reserves additional market territory for local semiconductor firms, ranging from industry leader Huawei to Cambricon, a Shanghai-listed company, and emerging players such as MetaX, Moore Threads, and Enflame.

Domestic alternatives now compete with certain Nvidia products in performance, yet adoption has lagged because developers favor Nvidia’s established software environment.

While the mandate will drive sales of Chinese chips, it may exacerbate the computing power disparity between the two nations.

American technology leaders including Microsoft, Meta, and OpenAI have committed hundreds of billions of dollars to data centers equipped with Nvidia’s top-tier processors.

In contrast, China’s premier foundry, SMIC, contends with production bottlenecks stemming from U.S. restrictions on advanced semiconductor equipment.