Scale AI

Scale AI provides the data labeling, human feedback, and model evaluation services that frontier labs rely on to train and align their models. Founded in 2016, it became one of the most important behind-the-scenes suppliers in AI. In 2025, Meta took a large stake and brought founder Alexandr Wang in to lead its superintelligence effort, reshaping Scale's independence and industry role.

Scale AI was founded in 2016 by Alexandr Wang and Lucy Guo to supply high-quality labeled data, initially for autonomous vehicles and later for training and aligning large language models. Human data — annotation, ranking, red-teaming, and evaluation — is a critical and labor-intensive input to frontier models, and Scale built the operations to deliver it at scale.

The company expanded into model evaluation and government AI work, positioning data quality and rigorous evals as differentiators as labs sought better post-training. Its customer list spanned the major labs and the US public sector.

In mid-2025 Meta invested $14.3 billion for a 49% stake and recruited Wang as Chief AI Officer to lead Meta Superintelligence Labs — a deal that gave Meta elite talent and data capabilities while prompting some rival labs to reduce reliance on Scale over neutrality concerns. It remains a pivotal, and now contested, node in the training supply chain.