Meta invests $14.3B in Scale AI

2025-06-12 · Publisher : Meta · Domain : Artificial Intelligence

Meta takes a 49% stake in Scale AI, the data-labeling specialist, for about $14.3 billion and hires its founder Alexandr Wang: the kickoff of an unprecedented talent war in AI.

What changes

  • About $14.3 billion for 49% of Scale AI, with no voting rights, valuing the company at $29 billion.
  • Alexandr Wang steps down as Scale's CEO to join Meta; Jason Droege becomes interim CEO.
  • At the end of June, Meta creates Meta Superintelligence Labs, led by Alexandr Wang.
  • Sam Altman says Meta offered signing bonuses of up to $100 million to OpenAI researchers.
  • The "minority stake plus hiring" structure avoids an acquisition subject to merger control.

On June 12, 2025, Alexandr Wang confirms he is stepping down as CEO of Scale AI to join Meta, as part of an investment of about $14.3 billion. Scale AI, founded in 2016, supplies the labeled data used to train and evaluate large language models. ### What happened - **The deal**: Meta acquires 49% of Scale AI, with no voting rights, valuing the company at about $29 billion. Scale remains an independent entity; the capital goes to paying shareholders and funding growth. - **The people**: Alexandr Wang joins Meta to work on "superintelligence" while remaining a Scale board member. Jason Droege, chief strategy officer, becomes interim CEO. - **What followed**: on June 30, 2025, Mark Zuckerberg announces Meta Superintelligence Labs, led by Alexandr Wang, and steps up hiring of researchers from OpenAI, Google DeepMind and Apple. ### Key figures | Item | Value | |---|---| | Amount invested | about $14.3 billion | | Stake | 49%, no voting rights | | Scale AI valuation | about $29 billion | | Bonuses cited by Sam Altman | up to $100 million | ### Why it matters After the lukewarm reception of Llama 4 in the spring, Meta changes scale: the challenge is no longer just compute, but data and a handful of researchers able to build frontier models. This kind of structure (a minority stake combined with hiring the leadership) brings in teams and know-how without a formal acquisition, a practice already seen with Microsoft and Inflection, and Google and Character.AI. ### Caveats Scale's independence was put to the test: several clients competing with Meta, including Google and OpenAI, reportedly scaled back their work with the company after the announcement. The size of the pay offers rests on public statements, not published contracts. ### Key takeaways Summer 2025 makes AI researchers the scarcest and most expensive resource in the sector, on a par with GPUs.

Links

Tags : IA, Financement, Talents, Meta, Données