Analysis: China has not yet seen the emergence of large independent AI data companies

AI Proem author Grace Shao stated on October 9 that China has long accumulated a large amount of data, but has yet to see large independent AI data companies similar to Surge AI, Scale AI, and Mercor in the United States. She wrote that the relevant background includes the formation of cutting-edge models and upstream data markets in the United States, while Chinese tech giants have long kept data within closed ecosystems, and the data of enterprises and traditional industries is scattered, with state-owned enterprises unwilling to easily hand over data to external suppliers.

The article cites data from the National Bureau of Statistics, stating that by 2025, China will generate 52.26 ZB of data, with only 22.13% of the stored data classified as structured data. An interview with Lin Zhenya, CTO of Benyuan Zhishu, by Data Monkey mentions that there are about 2,000 to 3,000 labeling companies in China, with the top three accounting for about 20% to 30% of the market, while he estimates that the top three in the United States account for about 70% to 80%. The author notes that this is an industry estimate, not an independently calculated market share.

The Information reported in September 2025 that Beijing-based company UniPat completed financing with a pre-investment valuation of $1 billion, while competitor Humanlaya's post-financing valuation is close to $1 billion. Both companies received investment from HSG and are reportedly serving Alibaba, ByteDance, The Dark Side of the Moon, and DeepSeek. A Hong Kong Stock Exchange announcement from Qian Baidu shows that the company will acquire approximately 90.91% equity in Benyuan Zhishu for 300 million yuan in June 2026. Haitan Ruisheng's 2025 annual report shows a year-on-year revenue increase of 59% to 377 million yuan.