U.S. Data Restrictions Spark China's Digital Infrastructure Push

AsianFin — A new wave of U.S. data restrictions targeting China has intensified concerns over global data decoupling and is prompting renewed urgency in Beijing's digital infrastructure ambitions.

On May 18, reports surfaced that the U.S. National Institutes of Health (NIH) has explicitly barred access from China to several critical biomedical databases, including those related to the human genome and disease research.

Major U.S. repositories such as SEER, TCGA, GTEx, and the Genomic Data Commons (GDC) have imposed new access restrictions, citing Executive Order 14117, issued by President Biden in February 2024. The order aims to "prevent access to Americans' bulk sensitive personal data and government-related data by countries of concern"—with China squarely in the crosshairs.

The NIH's move has triggered alarm across China's scientific and technological communities, highlighting vulnerabilities in the nation's reliance on foreign data ecosystems. For Chinese companies and research institutions engaged in genomics, AI, and life sciences, the action underscores the increasing importance of securing domestic data resources to sustain innovation and competitiveness.

In response, Beijing is fast-tracking its digital development agenda.

The National Data Administration (NDA) recently issued its first-ever Digital China Construction 2025 Action Plan to local governments. The blueprint targets a sweeping upgrade of national data capabilities, with goals that include raising the value-added contribution of the digital economy's core industries to over 10% of GDP, expanding computing capacity beyond 300 EFLOPS, and strengthening the data element market.

China's total data production reached 41.06 zettabytes in 2024, a 25% increase year-on-year. The domestic data sector now comprises more than 190,000 enterprises, generating over 2 trillion yuan in annual output. At its current growth trajectory—exceeding 20% annually—the industry is expected to reach 7.5 trillion yuan ($1.03 trillion) by 2030.

At the 2025 Data Security Development Conference held on May 17, Liu Liehong, Director of the NDA, emphasized that China possesses rich data resources, a complete industrial ecosystem, and vast market potential. He highlighted the dual imperative of ensuring data security while enhancing the efficient flow and utilization of data to empower the real economy.

"Data is no longer just a byproduct of digitization—it's the new 'industrial blood' of AI-driven development," said Liu.

As AI technology matures, high-quality datasets are becoming strategic assets. Experts warn that without accurate and well-labeled data, large language models (LLMs) are prone to hallucinations—false or misleading outputs that undermine credibility and utility.

Liu Bo, CTO of cybersecurity firm Anheng Information, said that privatized and vertical-specific data are crucial for building reliable AI agents. He noted that AI is already transforming business software, enterprise competitiveness, and real-world problem solving. "Every application is worth redoing in the AI era," Liu said. "Every industry will undergo a new round of transformation."

China's digital policymakers are laying out long-term plans to establish a vertically and horizontally integrated national data infrastructure. By 2029, the NDA aims to complete the main framework of this data backbone, supported by special treasury bonds and local-level fiscal investments.

According to Professor Huang Tiejun of Peking University, the next frontier in data sourcing will shift from text-based inputs—like papers and books—to images, videos, and multimodal data, all essential for training next-generation AI systems.

In 2024, the number of Chinese enterprises developing or applying AI jumped 36% year-on-year, while the number of public open data platforms and high-quality datasets grew by 7.5% and 27.4%, respectively, according to a report by the China Academy of Electronics and Information Technology.

Despite the momentum, China's data ecosystem still faces hurdles. Liu Wenqiang, Deputy Director of the CCID Research Institute, cited issues such as fragmented data aggregation, low output, inconsistent quality, and a lack of benchmark datasets. The nation is investing heavily in data annotation infrastructure, with over 17 trillion terabytes of data annotated across 335 high-quality datasets to support more than 120 homegrown LLMs.

Meanwhile, cybersecurity remains a top concern. At the conference, Academician Shen Changxiang from the Chinese Academy of Engineering introduced China's Trusted Computing 3.0 framework, designed to defend against emerging cyber threats in an increasingly open and>Industry experts say the next chapter of AI will be led by intelligent agents capable of full autonomy. "AI now understands data, business, and behavior better than humans," said Fan Yuan, chairman of DBAPPSecurity. "The future is dynamic data security and intelligent agents that actively participate in decision-making."

In Zhejiang Province, local governments are already deploying AI agents based on DeepSeek, an open-source Chinese LLM platform. Medical assistants powered by DeepSeek are being tested at Zhejiang Provincial Hospital of Traditional Chinese Medicine across various departments.

According to An Xiaopeng, Vice President of Alibaba Cloud Intelligence Group, AI is redefining not just productivity but the underlying logic of enterprise workflows. Tasks that once took months—such as training models for autonomous driving—can now be completed in under a week using automated AI-powered data pipelines.

"AI models are no longer just tools," An said. "They are becoming autonomous participants in the digital world, transforming industries from the ground up."

He added that with vertical fine-tuning, foundational models are now enabling companies to cut development costs by 70%, speed up time-to-market by 50%, and reduce data analysis cycles by 60%.

China currently produces about 23% of the world's data and is expected to increase its share to nearly one-third by 2025, reaching 48.6ZB in annual output. Meanwhile, the global data trading market is forecast to hit $144.5 billion by 2025 and more than double to $301.1 billion by 2030.

Academician Xu Tao of the Chinese Academy of Sciences said life and health data will be among the most strategically valuable. Technologies like blockchain and privacy-preserving computation are being leveraged to enable compliant, secure sharing of sensitive biomedical information, potentially slashing R&D costs by 30%.

As the U.S. and China deepen their divide over data access, both sides are racing to build self-sufficient data ecosystems. For China, the message is clear: data sovereignty is now inseparable from technological sovereignty—and the stakes have never been higher.

相关文章：

U.S. Data Restrictions Spark China's Digital Infrastructure Push05-19