Meta Platforms announced this week that it is splitting its newly formed artificial intelligence division into four distinct teams, marking the fourth major restructuring of its AI efforts in just six months as Chief Executive Mark Zuckerberg intensifies his pursuit of so-called superintelligence.
The reorganization divides Meta Superintelligence Labs, the unit established in June, into four specialized groups according to an internal memo sent Tuesday by Alexandr Wang, the former Scale AI chief executive who joined Meta as chief AI officer. The restructuring represents another attempt by the social media giant to better capitalize on billions of dollars worth of recently acquired AI talent and accelerate its artificial intelligence capabilities.
Under the new structure, Wang will lead a research lab internally known as “TBD Lab,” short for “to be determined,” which will oversee Meta’s large language models including the Llama series. The group serves as the centerpiece of Meta’s reorganized AI efforts, focusing on foundation models that underpin the company’s artificial intelligence ambitions.
The three other divisions include a products team responsible for developing user-facing tools like the Meta AI assistant, an infrastructure team handling the backbone systems that power Meta’s large-scale AI operations, and the long-standing Fundamental AI Research lab, which will continue focusing on longer-term research initiatives.
Nat Friedman, former GitHub chief executive, and Daniel Gross, who previously led Safe Superintelligence, will oversee development of new AI features under the products and applied research division. Aparna Ramani, a vice president in charge of AI, data and developer infrastructure, will run the new infrastructure team.
The changes come amid a turbulent period for Meta’s artificial intelligence operations, with the company recently investing heavily to recruit top talent from competitors. In June, Meta invested approximately $14.3 billion in Scale AI as part of the deal to bring Wang to the company. The social media giant has also been aggressively recruiting researchers from rivals including OpenAI, Google DeepMind and Anthropic, offering signing bonuses as high as $100 million according to comments made by OpenAI Chief Executive Sam Altman.
Meta’s latest AI model family, Llama 4, has underperformed expectations since its April release, receiving lukewarm response from developers and facing criticism that the company published misleading benchmark figures designed to make the models appear more competitive than they actually are. The company’s AI chatbot has also drawn negative attention after reports surfaced that it engaged in sexually explicit conversations with users, including minors.
The restructuring follows several high-profile departures from Meta’s AI teams in recent months, raising questions about stability within the division. Connor Hayes, who previously led the AI products group, was reassigned to run Threads, Meta’s competitor to Elon Musk’s X platform. Loredana Crisan, who previously led Messenger and later moved to the generative AI group, recently left Meta to join design software company Figma.
As part of the reorganization, Meta is dissolving its artificial general intelligence foundations team, with leaders Ahmad Al-Dahle and Amir Frenkel now focusing on strategic initiatives and reporting to Wang. The company confirmed that no layoffs were part of the reorganization, though sources familiar with the matter indicate Meta is considering downsizing some roles or shifting staff as its AI division has grown to thousands of employees in recent years.
The restructuring underscores the mounting pressure on Zuckerberg to demonstrate progress in artificial intelligence as Meta faces intense competition from rivals. The company has significantly increased its capital expenditure forecast to between $66 billion and $72 billion annually, up from $28 billion in 2023, as it builds massive data centers to support AI workloads.
Zuckerberg announced plans in July to invest “hundreds of billions of dollars” into AI infrastructure, including a 5-gigawatt data center called Hyperion in Louisiana with a footprint large enough to cover most of Manhattan. The company also plans to bring a 1-gigawatt supercluster called Prometheus online in 2026.
The Meta chief has been personally involved in recruitment efforts for the new AI organization, seeking what he calls the most elite and talent-dense team in the industry. His goal is to develop “personal superintelligence for everyone,” referring to AI systems that can perform beyond human level at most cognitive tasks.
However, the rapid pace of organizational changes has raised questions about the sustainability of Meta’s approach. The company paused hiring for its AI division last week, ending a spending spree that saw it acquire waves of costly researchers and engineers. A Meta spokesperson described the pause as “basic organizational planning” following its latest superintelligence efforts.
The restructuring comes as competition intensifies across Silicon Valley, with companies racing to develop artificial general intelligence systems that can outthink humans and contribute to new revenue streams. Meta’s reorganization aims to streamline its approach toward achieving superintelligence while giving the company a competitive edge in AI products, where it currently lags behind rivals.
