Meta might invest $10 billion in Scale AI, following the company's underwhelming Llama 4 launch earlier this year. The potential investment would tap into Scale AI's massive data labeling operation, potentially strengthening Meta's position in the AI race.
According to Bloomberg, Meta is in talks to invest about $10 billion in Scale AI, though the deal isn't finalized and terms may still change. Both Meta and Scale AI declined to comment on the negotiations.
If completed, this would be Meta's largest external investment in artificial intelligence so far. While Microsoft, Amazon, and Google have already poured billions into AI firms like OpenAI and Anthropic, Meta has mostly focused on building its own models in-house.
Meta previously participated in an earlier Scale AI funding round. Scale AI posted revenue of roughly $870 million in 2024, with a forecast of $2 billion for 2025 and a potential valuation reaching $25 billion. As recently as May 2024, the company was valued at $14 billion.
Founded in 2016 in San Francisco, Scale AI specializes in providing annotated data for AI model training. The company relies on thousands of contract workers to clean and label vast amounts of text, images, and other data types. Its client list includes tech giants like Microsoft and OpenAI.
Scale AI has become a key infrastructure player in the generative AI ecosystem. The data it processes is crucial for the performance of large language and multimodal models. Backers include OpenAI co-founder Greg Brockman and PayPal co-founder Peter Thiel, and both Amazon and Nvidia have invested as well.
Data quality becomes key competitive advantage
After Llama 4's lukewarm reception, Meta might be looking to secure exclusive datasets that could give it an edge over rivals like OpenAI and Microsoft. The investment could also help Meta gain tighter control over its own data pipeline.
Military applications are also on the table. Both Meta and Scale AI are deepening ties with the US government. The two companies are working on Defense Llama, a military-adapted version of Meta's Llama model. Scale AI recently landed a contract with the US Department of Defense to develop AI agents for operational use.
Until now, Meta has leaned heavily on open-source models. The shift toward a major investment in Scale AI could signal a move to a more closed and tightly regulated infrastructure—especially for projects like Meta's Aria headset, which collects multimodal user data. Labeling this kind of data is complex, but essential for making it useful in AI applications.
Recent developments suggest that advances in large AI models may depend less on architectural innovations and more on access to high-quality training data and compute. A closer relationship with Scale AI could help Meta process data more efficiently and at greater scale, potentially giving it an edge in the ongoing AI race.
