Meta is preparing its fourth overhaul of artificial intelligence operations in six months, with a plan to split its recently created Superintelligence Labs into four units, according to Reuters.
The restructuring would create a new “TBD Lab” (to be determined), a products team that includes the Meta AI assistant, an infrastructure group, and the long-term research arm FAIR. Meta has not commented publicly on the report.
The move follows a period of rapid organisational change as the company concentrates AI work within Superintelligence Labs, a unit formed to accelerate development toward artificial general intelligence (AGI). The reorganisation aims to clarify responsibilities between research, productisation and the compute infrastructure required to train and serve larger models. FAIR would continue to focus on foundational research, while the products group advances user-facing systems such as Meta AI across Facebook, Instagram, WhatsApp and the company’s hardware lines.
Chief executive Mark Zuckerberg has set out an expansive investment programme to support this agenda. In July, he said Meta would spend “hundreds of billions of dollars” to build several multi-gigawatt AI data centres as part of a push toward “superintelligence”. He also described plans for large “superclusters”, including projects called Prometheus (targeted for 2026) and a second facility capable of scaling to 5 GW.
Financing arrangements have begun to take shape. Earlier this month Reuters reported that Meta had tapped US bond manager PIMCO and alternative asset manager Blue Owl Capital to spearhead a $29 billion package to fund a data centre expansion in rural Louisiana. According to that report, PIMCO is expected to lead roughly $26 billion in debt financing, with Blue Owl contributing about $3 billion in equity.
Meta has also adjusted its near-term spending guidance. At the end of July the company raised the bottom end of its 2025 capital-expenditure forecast by $2 billion to a range of $66 billion to $72 billion. Management has said that the cost of data-centre build-out and higher employee compensation — amid an industry-wide battle to recruit and retain senior AI researchers — will result in a faster expense growth rate in 2026 than in 2025.
The latest restructuring would come after internal changes earlier this summer that concentrated Meta’s generative-AI programmes under Superintelligence Labs and sought to streamline decision-making. The company has experienced senior staff departures in parts of its AI organisation this year, while the most recent iteration of its open-source large language model, Llama 4, drew a muted reception compared with expectations, according to Reuters’ summary of The Information’s reporting.
To manage the capital intensity of its infrastructure plans, Meta has also explored shifting some assets off balance sheet. In early August, the company said it intended to sell about $2 billion of data-centre assets and partner with external financiers to co-develop facilities, a move designed to provide flexibility while maintaining the scale of its AI build-out.
The proposed four-part structure reflects the breadth of work Meta is pursuing. The “TBD Lab” would act as a small, elite research group; the products team would translate research advances into consumer and enterprise-facing features; the infrastructure group would focus on compute, networking and efficiency; and FAIR would continue its remit on long-term, fundamental AI research. The company has not confirmed the plan or a timeline for implementation.
Meta’s direction mirrors a wider pattern across large technology firms that are consolidating AI programmes while committing substantial capital to compute. The company argues that AI is already improving its advertising business and will underpin future products, including assistants and multimodal tools. The immediate operational impact of another reorganisation is unclear; however, the company’s recent guidance indicates that higher infrastructure and compensation costs are embedded into its multi-year planning.
If enacted, this would be Meta’s fourth AI reorganisation since February, underscoring the pace at which the group is attempting to align research, product and infrastructure for larger-scale model development.
OpenAI Loses Key Researchers to Meta, Reassesses Strategy to Retain Talent
