The CITIC Securities research report highlights that as publicly accessible internet data is fully leveraged and models progressively expand into areas such as coding, agent technology, world modeling, and specialized domains, the competitive landscape among large-scale models is transitioning from a focus on 'computing power + algorithms' to a more comprehensive 'computing power + algorithms + data' paradigm. Consequently, the demand for training data is evolving from general-purpose public domain data to premium, specialized, multimodal, and process-oriented data. Given the growing scarcity of high-quality, compliant, and clearly copyrighted data supplies, it is anticipated that the value of high-quality copyrighted data assets will undergo a re-evaluation.