Featured on
Data & model training · San Francisco, United States
Nearly every frontier model released since 2020 was trained in part on data Scale AI's workforce labeled or ranked, and Meta paying $14.3 billion for a minority stake - and hiring its founder outright - was one of the clearest signals yet of how much Big Tech values control over AI training data.
Scale AI was founded in San Francisco in 2016 by Alexandr Wang, then a 19-year-old MIT dropout, to provide the human-labeled and RLHF training data that large AI models are trained and fine-tuned on. Rather than building models itself, Scale sells the data infrastructure and human feedback pipelines that other labs' models depend on - a business that made it one of the most consequential companies in AI most consumers have never heard of.
In June 2025, Meta invested approximately $14.3 billion for a 49% stake in Scale, and Alexandr Wang left to become Meta's Chief AI Officer, leading its new Superintelligence Labs. Scale's board appointed Chief Strategy Officer Jason Droege as CEO, and the company has continued operating independently of Meta despite the size of Meta's stake and its founder's departure.
Scale Data Engine
Data labeling, curation and RLHF pipeline used to train and fine-tune large AI models.
Scale Evaluation
Model evaluation and red-teaming services for AI labs and enterprises.
Meta's consolidated AI division, formed in mid-2025 to unify its foundation-model, research and product AI work under one organization.
The open-source AI platform and model-hosting hub that much of the industry builds on - agreed to be acquired by Nvidia in September 2026.
The data and AI platform built around Apache Spark, now selling the infrastructure enterprises use to build and govern their own AI applications.
