Meta Makes Massive $15B Move to Grab Stake in Scale AI

Several media outlets have reported that Meta is finalising an agreement to spend $14 billion on Scale AI. Earlier this week, one media site said that an investment might exceed $10 billion, while on June 10, another media publication said that Meta will pay around $15 billion.

Various other media houses also reported that Wang, the founder of one of the most well-known AI startups, has established a reputation as an ambitious leader who not only understands the technical complexities of AI but also knows how to build a business that isn't just focused on research.

After the lacklustre debut of Meta's most recent Llama AI models, Zuckerberg will be looking to Wang to better carry out the company's AI goals.

By not immediately purchasing Scale AI, Meta seems to be following in the footsteps of Google and Microsoft, which have acquired well-known AI executives from the startups Character AI and Inflection AI by investing heavily in those businesses rather than outright purchasing them.

According to several persons with knowledge of the situation, Meta does not want to aggravate authorities by purchasing Scale AI while it is on trial before the Federal Trade Commission for antitrust charges.

Scale AI Revolutionising Generative AI Segment

Since its founding in 2016, Scale AI has made a significant impact in the generative AI age by assisting well-known tech firms like Microsoft, Google, and OpenAI in preparing the data they need to train state-of-the-art AI models.

A recently published media report, Meta is one of Scale AI's largest clients. The firm is ranked number 28 on CNBC's Disruptor 50 list and was valued at $14 billion in an investment round almost a year ago.

The company acquired over 180,000 square feet of space in a downtown building that had previously been occupied by Airbnb in mid-2024, securing one of the largest recent commercial contracts in San Francisco.

Scale AI announced a multimillion-dollar agreement with the Department of Defence in March, further solidifying its growing presence in the defence sector.

In a blog post published in November, the firm announced that it had partnered with Meta on Defence Llama, a customised version of Meta's open-source Llama foundation architecture created especially to "support American national security missions".

Meta Facing Tough Competition in the AI Race

AI was one of Meta's main concerns going into 2025. However, according to current and former Meta workers, Zuckerberg has become frustrated that competitors like OpenAI seem to be ahead in both underlying AI models and consumer-facing apps.

According to an earlier media report, Zuckerberg has been giving its more product-focused GenAI team greater priority than its Fundamental Artificial Intelligence Research unit, or FAIR, in order to help Meta advance AI and enhance its Llama family of AI models. Zuckerberg was even more irritated when developers did not like Meta's April release of its Llama 4 AI models.

Meta stated that it would soon develop a larger and more potent "Behemoth" variant of Llama 4, but at the moment, it only offered two smaller versions. As reported, Zuckerberg is worried about the model's capabilities in comparison to rival models, which is why it hasn't been released yet.

What worries people the most is how Behemoth compares to the newest models from firms like OpenAI and China's DeepSeek, whose algorithms are favoured by the larger developer community.