In a bold move to bolster technological self-reliance, the Chinese government has mandated that state-owned data centers must source more than 50% of their AI chips from domestic manufacturers.
This directive, part of China’s broader strategy to reduce dependence on American technology amid ongoing U.S. export sanctions, aims to strengthen the country’s AI and computing infrastructure. Initially trialed in Shanghai in 2023, the policy has been enforced nationwide since early 2025, according to industry sources, with a focus on specialized AI data centers that are rapidly expanding across the country.
Driving Technological Independence
China’s push for domestic AI chips aligns with its goal of achieving technological sovereignty in the face of escalating trade restrictions.
U.S. sanctions, particularly those targeting advanced semiconductor exports, have limited China’s access to cutting-edge chips from companies like Nvidia, prompting Beijing to accelerate the development of its homegrown semiconductor industry.
The new mandate requires state-run data centers to prioritize chips from Chinese manufacturers such as Huawei, Cambricon, and Loongson, reducing reliance on foreign suppliers like Nvidia, AMD, and Intel.
This policy is part of a broader national strategy that includes significant investments in AI infrastructure.
China is constructing numerous AI-focused data centers to support applications in machine learning, big data analytics, and smart city initiatives.
By mandating that over half of the chips in these facilities come from domestic producers, the government aims to create a robust ecosystem for Chinese-made hardware and software, fostering innovation and resilience against external supply chain disruptions.
Challenges in Performance and Compatibility
While Chinese AI chips have made significant strides, they still lag behind Nvidia’s offerings, particularly in training large-scale AI models. Nvidia’s GPUs, powered by the CUDA software ecosystem, remain the industry standard for high-performance computing tasks like training foundation models.
In contrast, Chinese chips, such as those developed by Huawei under its Ascend series and Compute Architecture for Neural Networks (CANN) platform, are considered more suitable for inference tasks — where trained models are deployed for real-time predictions — than for the computationally intensive process of model training.
A key technical challenge lies in adapting AI models developed for Nvidia’s CUDA to Huawei’s CANN framework. This transition requires significant reengineering, as developers must rewrite or optimize code to ensure compatibility with Chinese chips.
While companies like Huawei are investing heavily in software tools and developer support to ease this process, the gap in performance and ecosystem maturity remains a hurdle. Industry insiders note that Chinese chipmakers are making rapid progress, with firms like Cambricon and Hygon closing the gap in specific use cases, but achieving parity with global leaders will take time.
Implementation and Industry Impact
The mandate, which began as a pilot in Shanghai in 2023, has been rolled out nationwide following successful regional trials. State-owned data centers, which play a critical role in China’s AI ambitions, are now required to meet the 50% domestic chip threshold, with compliance monitored closely by regulatory bodies.
This policy is expected to drive demand for chips from Chinese manufacturers, boosting their market share and encouraging further innovation. Companies like Huawei, which has invested billions in its Ascend processors and Kunpeng CPUs, stand to benefit significantly, as do emerging players like Biren Technology and Moore Threads.
The policy also has implications for global tech giants. Nvidia, which has dominated the AI chip market, faces reduced access to China’s state-run data centers, though private-sector demand for its chips remains strong.
Meanwhile, Chinese firms are seizing the opportunity to capture market share, with Huawei reportedly doubling its Ascend chip shipments in 2024 compared to the previous year.
The mandate could accelerate the development of a self-sufficient AI ecosystem in China, potentially reshaping global supply chains.
Also read:
- True Detective Season 5: Nicolas Cage to Lead in Jamaica Bay, Issa López Returns
- Hollywood’s Post-Strike Fracture: The Battle for SAG-AFTRA’s Soul
- First Chinese Bank Launches Cryptocurrency Trading in Hong Kong
A Strategic Leap Forward
China’s mandate reflects a strategic response to geopolitical tensions and technological competition. By prioritizing domestic chips, the government is not only reducing reliance on foreign technology but also fostering a competitive homegrown industry.
However, challenges remain, including bridging the performance gap with global leaders and overcoming software compatibility hurdles. As China continues to invest in AI infrastructure and semiconductor innovation, the success of this policy will depend on the ability of domestic chipmakers to meet the demands of increasingly sophisticated AI applications.
With over 50% of AI chips in state data centers now required to be domestically produced, China is sending a clear message: it is committed to building a future where its technological destiny is firmly in its own hands.
