Mark Zuckerberg Overhauls Meta’s Artificial Intelligence Division Amid Internal Turmoil
By Mike Isaac and Eli Tan — Reporting from San Francisco, August 19, 2025
Meta CEO Mark Zuckerberg has once again shaken up the company’s artificial intelligence (A.I.) efforts, announcing a significant restructuring aimed at accelerating the development of cutting-edge A.I. technologies. The latest changes come amid growing internal tensions and ongoing competition in the fast-evolving A.I. landscape.
Major Reorganization of Meta’s A.I. Labs
According to sources familiar with the matter, Meta has internally announced the division of its A.I. department—known as Meta Superintelligence Labs—into four distinct groups. Each group will have a focused mandate: one dedicated to A.I. research, another on developing a powerful A.I. dubbed “superintelligence,” a third on product development, and a fourth on infrastructure including data centers and A.I.-related hardware.
This realignment is aimed at better organizing Meta’s efforts to accelerate the company’s ambitious goal of achieving superintelligence—a level of artificial intelligence surpassing human capabilities—while simultaneously speeding up the rollout of advanced A.I. products to remain competitive.
Potential Downsizing and Staff Changes
Internal sources also disclosed that some A.I. executives may leave the company as part of this shuffle. Meta is reportedly considering downsizing the overall A.I. division, which has ballooned to thousands of employees in recent years. This could involve job eliminations or employee reassignments to other parts of the company. Final decisions on these measures had yet to be made, with discussions remaining fluid as of the announcement.
Shift Toward Integrating Third-Party A.I. Models
In a notable departure from its traditional practice of relying exclusively on proprietary A.I. technologies, Meta is reportedly exploring the integration of third-party A.I. models into its product ecosystem. This includes open-source models freely available within the developer community, as well as licensed “closed-source” models from other companies. Such a strategy marks a shift in Meta’s approach to A.I. development and may allow faster innovation by leveraging external advancements.
Context: An Intense Internal Battle for A.I. Leadership
This restructuring follows months of turbulence at Meta’s A.I. division. Earlier this year, after struggles to advance its new A.I. models, Zuckerberg announced the creation of a dedicated superintelligence lab with a $14.3 billion investment in the start-up Scale AI. Alexandr Wang, Scale AI’s CEO, was appointed Meta’s chief A.I. officer, tasked with leading the company’s most ambitious A.I. development efforts.
Zuckerberg has demonstrated a clear determination to dominate the A.I. space, stating in a recent investor call that superintelligence will usher in “a new era of individual empowerment.” He highlighted how A.I. has already strengthened Meta’s advertising business and revealed plans for capital expenditures potentially reaching $72 billion this year, mainly aimed at data centers and hiring top A.I. talent.
Internal Tensions and New Leadership Dynamics
The arrival of Wang and his team has sparked tensions inside Meta. They have focused on creating the company’s most advanced A.I. “frontier model,” and controversially decided to make this next-generation model a “closed” system—keeping the technology secret rather than open-sourcing it, a long-standing Meta philosophy.
Wang’s group has abandoned the previous frontier model, code-named Behemoth, after disappointing test results led to delays earlier this year. The decision to start fresh underscores the ambitious yet risky nature of their project.
Meta has aggressively recruited talent from rival organizations, sparking a Silicon Valley “poaching war.” Notable new hires include Shengjia Zhao, a former OpenAI researcher and co-creator of ChatGPT, appointed as Meta’s chief A.I. scientist. Zhao has reportedly begun extensive interviews with Meta’s veteran A.I. teams, probing their previous work as the company reshapes its workforce.
Meanwhile, other key Meta A.I. leaders remain. Rob Fergus, who co-founded Meta’s A.I. research division in 2014, will continue to lead the Fundamental AI Research lab (FAIR), which focuses on open-source research. Ahmad Al-Dahle and Amir Frenkel are working under Wang on strategic generative A.I. products.
Personnel Departures Signal Changing Guard
The restructuring has coincided with notable departures. Joelle Pineau, a leading computer scientist at Meta, left to join the A.I. startup Cohere. Angela Fan, involved in Meta’s open-source Llama model, moved to OpenAI. Loredana Crisan, vice president of generative A.I., is joining Figma as chief design officer.
Meta’s A.I. Journey Continues Amid High Stakes
Mark Zuckerberg’s readiness to restructure and invest heavily in artificial intelligence signals Meta’s commitment to maintaining relevance in a rapidly evolving and highly competitive technology sector. However, the internal friction and strategic shifts also underscore the challenges tech giants face as they race to develop superintelligent A.I. systems and integrated products.
Meta declined to comment on the restructuring. Meanwhile, industry observers and investors are closely watching, as leadership in A.I. could define the next era of technology, creating new winners — and losers — in Silicon Valley.
Correction: An earlier version of this article misspelled the surname of Amir Frenkel, who reports to Alexandr Wang on Meta’s strategic A.I. initiatives.
About the Authors
Mike Isaac is a technology correspondent for The New York Times based in San Francisco, regularly covering Facebook and Silicon Valley.
Eli Tan covers the technology industry for The Times from San Francisco.
This article originally appeared in print with the headline: “Zuckerberg Shakes Up A.I. Efforts At Meta.”
