China Builds AI Data “Granary” to Fuel Scientific Research

The Chinese Academy of Sciences launched the “Ao Cang” Science Data Repository to provide high-quality data for artificial intelligence research and development in China. The repository, unveiled during the 2026 World Artificial Intelligence Conference, aims to address the critical need for strategic data resources in the AI field.

It aggregates over 320 petabytes of scientific data, 150 million scientific papers, 120 million invention patents, and 5 million scientific books spanning diverse data types. The “Ao Cang” repository isn’t simply a data collection, but a systematically engineered project involving nearly 100 domestic institutions and experts.

It employs a “1+7+N” structure, one national-level comprehensive repository, seven specialized domain repositories, and numerous application-specific sub-repositories, and a three-tiered knowledge refinement model to maximize semantic density. Currently, the repository is supplying data to foundational science models like “Panshi” and specialized models in areas like chemistry and healthcare, with plans to expand to commercial applications in aerospace and manufacturing. The Academy hopes “Ao Cang” will establish a solid data foundation for China’s AI development and promote scientific independence.