Meta’s AI research division, FAIR, is experiencing significant leadership departures, with VP Jitendra Malik joining Amazon’s robotics team. This follows Chief AI Scientist Yann LeCun’s exit, signaling a shift from fundamental research to commercial AI products. The unit faces internal turmoil and resource constraints as Meta prioritizes generative AI initiatives.

Meta’s famed AI research division has lost yet another senior leader, with VP Jitendra Malik announcing his departure to join Amazon’s robotics efforts. The exit comes amid mounting turmoil at FAIR (Fundamental AI Research), which some former employees have described as “dying a slow death” as Meta pivots aggressively toward commercial AI products. Malik’s resignation follows chief AI scientist Yann LeCun’s recent departure and adds to a growing exodus of top researchers who helped build Meta’s reputation as an AI powerhouse.Malik, who split his time between Meta’s FAIR lab and UC Berkeley, announced on January 4 that he will lead Amazon’s robotics research in San Francisco starting January 5. “It is time to move on,” he wrote on X, highlighting accomplishments including video action recognition models, tactile sensors, and the recent SAM 3D breakthrough in computer vision.

FAIR’s identity crisis: From blue-sky research to product focus

The departure caps a tumultuous year for FAIR, a unit once considered Meta’s crown jewel in AI development. Since CEO Mark Zuckerberg launched his “year of intensity” push in January 2025, the lab has undergone multiple reorganizations, faced budget constraints, and seen more than half the authors of the original Llama research paper leave within months of publication.Former FAIR researchers told Fortune earlier this year that Meta has steadily deprioritized open-ended exploration in favor of product-driven initiatives under the GenAI organization. The lab now gets less computing power than teams focused on generative AI, according to ex-employees, though Meta declined to confirm resource allocation details.In August, Meta consolidated its AI efforts by dividing the Superintelligence Labs into four teams, creating confusion over project ownership and team assignments. The reshuffling was followed by 600 job cuts in October, and Chief AI Officer Alexandr Wang framed the layoffs necessary to accelerate decision-making.

Tensions simmer as LeCun criticizes LLM-focused strategy

That instability was compounded when LeCun left to pursue his own project on “world models”—a path different from that of the large language models that now define Meta’s approach to AI. In a December Financial Times interview, LeCun admitted Meta’s Llama 4 results were “fudged a little bit” and criticized the company’s “LLM-pilled” direction as incompatible with achieving superintelligence.Former research scientist Tian Yuandong, who left Meta in November, told Chinese media that internal conflicts emerged as computing resources became scarce amid the industrywide LLM race. Another ex-employee, Joena Zhang, wrote that “nobody really knew what anyone was doing” during the first half of 2025 at what’s now called Meta Superintelligence Labs.Meta says it remains committed to FAIR, and employee sentiment surveys of recent times indicate improvement: Optimism was at 80% by late 2025.