Snorkel AI, which had its roots in a Stanford University laboratory, initially concentrated on marketing AI labeling software. Starting from 2025, the company has pivoted its business model to directly supply specialized and highly complex training data to AI enterprises. Within a year of this strategic shift, its annualized revenue catapulted from around $20 million to surpassing $350 million. Recently, Snorkel AI successfully secured a $350 million Series E funding round, propelling its valuation to $3.5 billion.

Currently, the evolution of large-scale AI models has sparked a significant uptick in the demand for specialized, high-quality training data. This trend has attracted a variety of companies from different sectors to enter the market. By leveraging the expertise of professionals such as programmers, lawyers, doctors, and PhDs, these companies are able to tailor training data specifically for AI enterprises, leading to swift revenue expansion. The concept of providing AI with “exam questions” and “grading” services is quickly gaining traction.

In the domestic market, Haitian Ruisound also witnessed synchronized growth in both revenue and profit during the first half of 2026. It's worth noting that the standardized data products, which can be repeatedly sold, incur significantly lower costs compared to customized services. Hence, transforming expert knowledge into products that can be sold repeatedly has emerged as a crucial factor for the future growth of such data companies.