China Bans Foreign AI Chips from State-Funded Data Centers to Reduce Tech Dependence

China has issued one of its most aggressive moves yet to eliminate foreign technology from its critical digital infrastructure, ordering all state-funded data center projects to use only domestically manufactured artificial intelligence chips. The directive, issued in early November 2025, instructs regulatory authorities to mandate that any new data center receiving government backing must deploy exclusively Chinese-made AI processors.​

The implications are sweeping and immediate. Data centers currently under construction and less than 30 percent complete have been ordered to remove any already-installed foreign chips or cancel pending orders, while projects in more advanced stages will be reviewed on a case-by-case basis to determine whether they can continue using imported technology. This marks a decisive escalation in Beijing’s long-running quest for technological self-sufficiency and represents the formalization of policies that were previously only informally enforced through customs checks and unofficial pressure campaigns.​

The End of Nvidia’s Dominance in China

For Nvidia, the world’s most valuable chip company and longtime leader in AI processors, the directive represents a devastating blow. The American company once commanded approximately 95 percent of China’s AI chip market in 2022, but that share has now plummeted to effectively zero, according to Nvidia CEO Jensen Huang. The new ban applies not only to Nvidia’s most advanced chips restricted under US export controls but also to the H20—a scaled-down processor specifically designed to comply with American regulations and maintain Nvidia’s foothold in the Chinese market.​

The ban also extends to Nvidia’s more powerful B200 and H200 processors, which, although officially restricted from export to China by US regulations, have remained widely available through unofficial gray-market channels. By targeting even the compliance-oriented H20 and blocking gray-market imports of higher-end chips, Beijing has effectively closed all pathways for Nvidia chips to enter Chinese government-backed data centers.

Other foreign chipmakers affected by the directive include AMD and Intel, both of which have supplied data center chips to China in recent years. The ban represents a comprehensive rejection of American semiconductor technology in China’s most critical AI infrastructure.​​

Over $100 Billion in State Funding at Stake

The scale of this policy shift cannot be overstated. According to a Reuters review of government tenders, AI data center projects in China have attracted more than $100 billion in state funding since 2021. The vast majority of data centers in China have received some form of government financial assistance for their construction, meaning the new directive will impact a substantial portion of the country’s digital infrastructure.​

Some projects have already been suspended before breaking ground as a result of the directive. One facility in a northwestern province that had planned to deploy Nvidia chips has been put on hold after the private technology company developing it received state funding. The project remains frozen as developers determine how to comply with the new requirements.​

China’s Domestic Champions: Huawei, Cambricon, and Beyond

The ban is designed to redirect massive demand toward domestic chipmakers, particularly Huawei Technologies, which has emerged as China’s most prominent AI chip manufacturer. Huawei’s Ascend series of AI processors has been positioned as the primary alternative to Nvidia’s offerings, though technical comparisons suggest performance gaps remain.​

Other Chinese AI chip companies expected to benefit from the directive include Cambricon Technologies, MetaX Integrated Circuits, Moore Threads Technology, and Enflame Technology. Cambricon, in particular, has seen extraordinary growth, with its stock price doubling in recent months and its market capitalization reaching approximately 580 billion yuan ($81 billion). The company reported a 14-fold surge in quarterly revenue and swung to profitability for the first time, reflecting surging demand as Chinese tech firms shift away from Nvidia chips.​

Founded in 2016 by brothers Chen Tianshi and Chen Yunji, who previously developed processors at the Chinese Academy of Sciences, Cambricon has positioned itself as a pure-play AI chip designer supplying chips to Baidu, Alibaba, ByteDance, Tencent, and other major Chinese technology firms. Goldman Sachs estimates Cambricon’s revenues will more than double from 6.5 billion yuan ($910 million) in 2025 to 13.8 billion yuan ($1.93 billion) in 2026, with its market share of China’s AI chip market expected to grow from 3 percent to 11 percent by 2028.​

Huawei, meanwhile, is forecast to ship between 600,000 and 650,000 AI chips in 2025, while Cambricon is on track to supply 125,000 to 150,000 units. These production volumes reflect rapid scaling of domestic semiconductor manufacturing capacity, though both companies continue to face challenges related to software ecosystems, developer tools, and high-bandwidth memory integration.​

The Performance Gap and China’s Strategy

While Chinese chipmakers have made significant strides, industry analysts acknowledge that domestically produced AI chips still lag behind Nvidia’s offerings in raw performance and software maturity. Huawei’s Ascend line, often cited as the most mature domestic option, still lacks full parity with Nvidia’s CUDA-based software stack, which has become the industry standard for AI development.​

Benchmarks suggest that Cambricon’s Siyuan 590 processor reaches approximately 80 percent of the performance of Nvidia’s A100 chip, while the upcoming Siyuan 690 aims to rival the H100. However, many developers who have spent years working within Nvidia’s software ecosystem remain reluctant to adopt domestic alternatives, creating friction in the transition.​

China’s strategy to overcome these performance gaps relies on two key advantages: abundant cheap energy and massive chip clusters. Rather than competing chip-for-chip with Nvidia, Huawei has focused on connecting large numbers of Ascend processors into high-performance clusters that can collectively match or exceed the capabilities of smaller numbers of Nvidia GPUs. Huawei’s CloudMatrix384, for example, integrates 384 Ascend 910C chips to deliver performance comparable to Nvidia’s GB200 NVL system, which utilizes only 72 GPUs.​

This approach consumes significantly more power, but China’s massive investments in renewable energy sources, including solar, wind, and nuclear power, provide low-cost electricity that offsets the inefficiency. Various cities across China, including Shanghai and Shenzhen, have introduced subsidies and vouchers to help lower computing costs for companies renting processing power.​

The US-China Tech War Intensifies

The Chinese directive comes amid escalating technological competition between Beijing and Washington. US export controls have repeatedly restricted China’s access to advanced semiconductors, with successive administrations defending these measures as necessary to prevent American technology from supporting Chinese military capabilities.​

In October 2025, just weeks before China’s ban took effect, US President Donald Trump stated that Washington would allow China to “deal with Nvidia but not in terms of the most advanced” chips. Trump has also indicated that Nvidia’s most advanced Blackwell chips would be reserved exclusively for US companies and kept out of China and other countries.​

The White House has reportedly torpedoed Nvidia’s push to export AI chips to China, with Trump deciding against discussing the matter with Chinese leader Xi Jinping after top aides opposed it. The administration has kept in place limits on China’s access to high-end chips as part of its broader AI Action Plan.​

Earlier in 2025, China had already begun signaling its intention to reduce dependence on American chips. In September, the Cyberspace Administration of China instructed major tech companies including ByteDance and Alibaba to cancel orders and halt testing of Nvidia products, including the H20 and RTX Pro 6000D. Chinese customs authorities simultaneously tightened controls on chip imports at major ports, conducting stringent checks on semiconductor shipments nationwide to discourage the use of US products.​

A Calculated Gamble on Self-Reliance

Beijing’s aggressive push for AI chip self-sufficiency represents both strategic necessity and calculated risk. On one hand, the policy insulates China’s critical infrastructure from potential future US export restrictions and reduces leverage that Washington can exert through semiconductor supply chains. On the other hand, forcing a premature transition to domestic chips that are not yet fully competitive with Nvidia’s offerings risks slowing China’s AI development and widening the technological gap with the United States.​

Analysts suggest that China’s confidence in issuing this directive reflects an assessment that domestic chips have reached sufficient maturity to handle most AI workloads, even if they lag in peak performance. A semiconductor industry source stated that “China’s shift from passive response to active offense reflects its self-reliance resolve and confidence”.​

However, the transition will not be frictionless. Many domestic chips still struggle with mature software ecosystems, developer tools, and high-bandwidth memory integration, meaning immediate performance or compatibility gaps may persist and slow some deployments. The directive also risks creating a bifurcated global AI ecosystem, with Chinese and Western supply chains increasingly decoupled.​

What This Means for Global Tech Companies

For global technology firms, China’s directive represents a significant revenue loss and a decisive step toward technological decoupling. Excluding foreign chipmakers like Nvidia from major state projects eliminates a substantial portion of their China revenue, even as negotiations continue about potentially resuming limited chip sales to the country.​

Nvidia has been lobbying the Trump administration to permit sales of its forthcoming B30A chip to China, which would deliver roughly half the performance of the flagship B300 at half the price. However, even if such sales were approved, the domestic-only requirement for state-funded data centers would severely limit the addressable market.

For Chinese technology companies, the directive creates both opportunity and challenge. Huawei, Cambricon, and other domestic chipmakers stand to capture billions of dollars in additional revenue as demand shifts away from foreign suppliers. However, these companies will also face intense pressure to rapidly improve performance, expand production capacity, and develop robust software ecosystems to support developers accustomed to Nvidia’s tools.​

The Road Ahead: Decoupling and Competition

China’s ban on foreign AI chips from state-funded data centers marks a turning point in the global technology landscape. What began as US efforts to restrict China’s access to advanced semiconductors has now evolved into China’s proactive push to eliminate American technology from its critical infrastructure entirely.​

The policy underscores the accelerating technological decoupling between the world’s two largest economies and raises profound questions about the future of global AI development. Will China’s domestic chip industry mature rapidly enough to support its AI ambitions, or will the performance gap with US technology widen and slow Beijing’s progress? Can Huawei, Cambricon, and other Chinese firms successfully replicate the robust software ecosystems that have made Nvidia’s CUDA platform the industry standard?

As both nations double down on technological self-sufficiency and seek to dominate the AI era, the competition is no longer just about who makes the best chips—it is about which country can build the most complete, competitive, and independent technology stack from silicon to software to services.​

For now, the message from Beijing is clear: China’s critical AI infrastructure will run on Chinese chips, regardless of the short-term costs or technical challenges involved.​

Click Here to subscribe to our newsletters and get the latest updates directly to your inbox