Despite a significant $14.3 billion investment in Scale AI two months prior, Meta is now reportedly leveraging competitor resources to train its next-generation AI models. This shift suggests a potential re-evaluation of Meta's exclusive reliance on Scale AI for AI development. The initial investment granted Meta a 49% stake in Scale AI, with founder Alexandr Wang joining Meta to lead a new 'superintelligence' unit.
However, the partnership appears to have triggered a market shake-up, with major clients like Google and OpenAI reducing or ending their engagements with Scale AI. These companies are seeking alternative data providers, potentially due to concerns over data confidentiality and competitive advantage. This situation highlights the critical importance of diverse, high-quality data in the competitive AI landscape and the strategic implications of Meta's evolving approach to AI model training.
Scale AI, which specialises in providing labelled training data for AI models, had a revenue of approximately $870 million in 2024 and projected $2 billion for 2025. The company's shift towards more specialised data labelling reflects the growing complexity of AI models.
