TMTPOST -- meta Platforms, Inc. is going to strike a deal for one of its most aggressive bets on artificial intelligence (AI) , according to reports on Tuesday.

AI Generated Image

meta is close to a deal to invest $14 billion in data-labeling startup Scale AI, CNBC cited people familiar with the matter. It was reported that meta CEO Mark Zuckerberg is so frustrated with the Facebook parent’s standing in AI that he’s willing to spending billions of dollars to convince Scale AI founder and CEO Alexander Wang to join his company. Wang is well-known for its reputation as a leader who both understands AI’s technical complexities and how to build a business that’s not merely focused on research, according to former meta employees.

The Wall Street Journal echoed the investment of $14 billion, touting it as one of meta’s largest-ever outside investments. Its sources said meta was in advanced talks to make the deal and hire Scale AI CEO Wang to help lead its AI efforts to develop advanced AI called superintelligence. Under the deal, which is still being finalized, other the startup’s employees would also join meta.

Mata has agreed to invest nearly $14.8 billion to take a 49% stake in Scale AI, and the deal will be structured so meta will send the cash to Sales’s existing shareholders and place its CEO Wang in a top position inside meta, The Information learned from people with knowledge of the matter. Scale AI shareholders will maintain their holdings and the AI startup will be valued at $28 billion folling the upcoming deal, up from $13.8 billion last year, according to the report.

Both CNBC and the Journal suggested Zuckberg’s recruitment decision was due to meta’s difficulties in developing its latest Llama AI models. The Journal reported Zuckberg personally got involved in the recruitment, including by reaching out to AI researchers directly on WhatsApp. meta last month was reported to push back rollout of its large language model (LLM) dubbed “Behemoth” under its Llama 4 family due to concerns about its capabilities. meta engineers are struggling to substantially improve capabilities of the model, stirring questions among the staff about if improvements over its predecessor versions are significant enough to justify public statements about its capabilities.

If the report is accurate, meta has so far postponed its planned release of the Behemoth model twice. Early in the development of the model, the social media titan was said to internally arrange the release in April to coincide with its inaugural AI conference for developers, but later delayed the internal target of rollout to June, and the current target became fall or even later.

Earlier this week, Bloomberg reported meta is in talks to make an investment that could top $10 billion in Scale AI. Scale AI provides data labeling services for companies like Microsoft (MSFT) and OpenAI, helping to enable to training and production of large language models. meta has previously worked with Scale AI to help train its Llama large language models, as well as to develop a "Defense Llama" AI model focused on U.S. national security.

Zuckerberg is creating a new AI research lab dedicated to pursuing “superintelligence”, a hypothetical A.I. system that exceeds the powers of the human brain, and has tapped Scale AI CEO Wang to join the new lab, the New York Times reported on Tuesday. It was reported that meta has offered compensation packages between seven and nine figures to AI researchers from top competitors, some of whom have accepted. Bloomberg the same day reported Zuckberg is hiring a team of top AI talents for meta’s “superintelligence” lab. Zuckerberg has been personally leading recruiting for the team of about 50 people — and has even rearranged meta's offices so the new hires sit near him, per the report.