According to sources, Mark Zuckerberg is prepared to spend billions of dollars to entice Scale AI CEO Alexandr Wang to join his business, citing frustration with Meta’s position in artificial intelligence.

Meta is completing a $14 billion investment in Scale AI, according to a source familiar with the subject who asked not to be identified because the terms are secret. Bloomberg reported earlier this week that the investment might be worth more than $10 billion, while The Information claimed on Tuesday that Meta will pay around $15 billion.

As the founder of one of the most prominent AI startups, Wang has established himself as an ambitious leader who understands both the technical complexities of AI and how to build a business that is more than just research, according to two former Meta AI employees who agreed to speak anonymously. Following the underwhelming rollout of Meta’s newest Llama AI models, Zuckerberg will rely on Wang to better implement the company’s AI objectives.

Meta’s decision to not immediately acquire Scale AI aligns with businesses like Google and Microsoft, who have hired AI executives from startups like Character. AI and Inflection AI by acquiring significant shares in such firms rather than purchasing them completely. Meta is now on trial before the Federal Trade Commission for antitrust violations, and the business does not want to further anger authorities by acquiring Scale AI, according to numerous people familiar with the case.

According to The Information, Meta will acquire a 49% share in the data-labelling and annotation firm, while Wang will manage a new AI research lab within the social networking company, along with several of his colleagues. The New York Times was the first to report about the new AI lab.

Scale AI, founded in 2016, has made a splash in the era of generative AI by helping major tech companies like OpenAI, Google and Microsoft prepare data they use to train cutting-edge AI models. Meta is one of Scale AI’s biggest customers, according to two people familiar with the matter.

Meta’s AI challenges

AI was one of Meta’s primary concerns as we approached 2025. However, present and former Meta workers claim that Zuckerberg is becoming increasingly concerned that competitors such as OpenAI look to be ahead in both underlying AI models and consumer-facing applications.

To assist Meta in making progress in AI and improving its Llama family of AI models, Zuckerberg has deprioritized its Fundamental Artificial Intelligence Research unit, or FAIR, in favor of its more product-oriented GenAI team.

The introduction of Meta’s Llama 4 AI models in April was met with criticism from developers, further upsetting Zuckerberg, according to the sources. At the moment, Meta only published two minor versions of Llama 4 and promised to create a bigger and more powerful “

According to the sources, Zuckerberg is concerned about the model’s capabilities in comparison to rival models, hence it has yet to be released. There is especially worry over how Behemoth compares to the most recent offerings from businesses like as OpenAI and China’s DeepSeek, whose approaches are chosen by the larger developer community.

Following Llama 4’s disappointing launch, Meta reorganized its GenAI unit, dividing it into two. Amir Frenkel, previously a vice president of engineering and product for Meta’s Reality Labs hardware branch, and Ahmad Al-Dahle, the previous head of GenAI, were assigned responsibility for AGI Foundations, while Connor Hayes, a long-time Meta employee, was given leadership of AI Products.