Meta Splits AI Division to Chase Superintelligence Dreams

Meta is restructuring its artificial intelligence (A.I.) division in a strategic move aimed at accelerating the development of advanced AI technologies and maintaining its competitive edge in a rapidly evolving industry. The company, formerly known as Facebook, has announced plans to split its A.I. division—Meta Superintelligence Labs—into four distinct groups, as revealed by two individuals familiar with the matter. These groups will focus on A.I. research, the development of “superintelligence,” product innovation, and infrastructure, such as data centers and A.I. hardware [1].

The reorganization is intended to streamline operations and align the division with Meta’s long-term goals of advancing toward superintelligence and rapidly bringing A.I. products to market. The move follows months of internal restructurings and strategic shifts. Some A.I. executives are expected to leave the company, and there are also discussions about downsizing the A.I. division overall, which has grown significantly in recent years to include thousands of employees. However, no final decisions have yet been made on the scale of the reductions [1].

Meta is also exploring the use of third-party artificial intelligence models, marking a potential shift from its previous strategy of relying solely on its own proprietary technologies. The company could build on open-source A.I. models or license closed-source models from other firms, depending on strategic and operational considerations [1]. This shift aligns with Meta’s broader goal of accelerating development cycles and improving product capabilities, particularly as the A.I. landscape becomes increasingly competitive.

The restructuring follows a significant investment in A.I. talent and infrastructure. In June 2025, Meta announced a dedicated superintelligence lab and invested $14.3 billion in Scale AI, a start-up focused on A.I. development. The company also recruited Alexandr Wang, Scale AI’s CEO, as its new chief A.I. officer. To bolster its A.I. capabilities, Meta offered multi-million-dollar packages to attract researchers from rival firms such as OpenAI and Google, intensifying competition for top talent [1].

Under the leadership of chief A.I. officer Alexandr Wang, Meta has also faced internal tensions as his team, focused on developing the company’s most powerful A.I. model—a so-called “frontier model”—has diverged from previous strategies. Wang’s team has chosen to abandon the previous model, known as Behemoth, and instead start fresh with a new model, despite delays in Behemoth’s initial release due to performance issues [1].

Meanwhile, other key figures in Meta’s A.I. division continue to hold influential roles. Rob Fergus, who co-founded Meta’s A.I. research division in 2014, will remain as the head of the Fundamental AI Research (FAIR) lab, responsible for advancing open-source A.I. technology. However, several high-profile A.I. researchers and executives have left the company in recent months, including Joelle Pineau, who joined Cohere, and Angela Fan, who moved to OpenAI. This turnover highlights the ongoing restructuring and realignment within the division [1].

The changes underscore Meta’s commitment to staying at the forefront of the global A.I. race, as the company invests heavily in both talent and infrastructure. With projected capital expenditures of up to $72 billion in 2025, a significant portion of which will be allocated to data center construction and A.I. research, Meta is positioning itself to compete with other major players in the field. As the A.I. landscape continues to evolve, Meta’s ability to adapt and innovate will be critical to its future success [1].

Source: [1] Mark Zuckerberg, Meta’s chief executive, has spent the past few months shaking up his company’s artificial intelligence efforts (https://www.nytimes.com/2025/08/19/technology/mark-zuckerberg-meta-ai.html)