Meta Restructures Its AI Division Once Again in Pursuit of Superintelligence

Meta has reshuffled its artificial intelligence division yet again, underscoring the company’s relentless ambition to achieve one of technology’s most elusive goals: superintelligence. The company’s AI arm, known as Meta Superintelligence Labs (MSL), has now been reorganized into four specialized groups, each focused on a different pillar of AI development. It marks the fourth major restructuring in less than a year, a pace of change that reveals both the urgency and the pressure behind Meta’s high-stakes AI strategy.

At the center of this new structure is Alexandr Wang, recently elevated to the role of Chief AI Officer. Wang’s team—unofficially referred to as “TBD Lab”—is charged with developing and scaling the foundation models that Meta hopes will define the next generation of artificial intelligence. This includes successors to its Llama family of large language models and the ambitious “omni” model, which promises to extend well beyond text into multimodal intelligence.

Supporting this effort is FAIR (Fundamental AI Research), Meta’s longstanding research hub. FAIR continues to explore the theoretical underpinnings of artificial intelligence, led by Rob Fergus with Chief Scientist Yann LeCun still playing a prominent role. Both Fergus and LeCun now report directly to Wang, signaling Meta’s desire to tightly link groundbreaking research with practical model development.

Another piece of the puzzle is Products and Applied Research, a group now overseen by former GitHub CEO Nat Friedman. This team’s mission is clear: take Meta’s research and translate it into tools, features, and applications that can be integrated across its massive user ecosystem—Facebook, Instagram, WhatsApp, and beyond. In other words, they’re the bridge between advanced research and everyday consumer experiences.

The final piece is MSL Infra, run by engineering VP Aparna Ramani. Her team is tasked with building the enormous infrastructure required to power Meta’s AI ambitions—everything from advanced data centers to the compute clusters that train these ever-larger models. With Meta’s AI capital expenditures projected to reach an eye-popping $66–72 billion in 2025, infrastructure has become as critical as the models themselves.

So why another reorganization? The answer lies in speed and competitiveness. Meta is racing against the clock, chasing rivals like OpenAI, Google DeepMind, and Anthropic, all of which are advancing steadily toward more powerful systems. By splitting MSL into four tightly focused divisions, Meta hopes to accelerate progress while eliminating bottlenecks that can arise when too many priorities are funneled through a single organizational layer.

Yet this strategy is not without risk. Constant reshuffling can fracture morale and create confusion for teams that need stability to deliver results. There have also been reports of growing tension inside Meta, as the company lures external AI talent with compensation packages worth over $100 million while longtime employees feel sidelined. The result is a delicate balancing act: moving fast enough to stay competitive without undermining the very teams tasked with making the breakthroughs.

Ultimately, the reorganization reflects the scale of Meta’s ambition. By centralizing leadership under Wang and creating distinct lanes for research, productization, and infrastructure, Meta is betting that structural clarity will lead to faster innovation. The stakes couldn’t be higher. If successful, Meta could position itself as a leader in the race toward superintelligence. If not, it risks becoming a case study in how too much organizational churn can derail even the best-funded projects.

For now, the world will be watching whether this latest structural shake-up produces the momentum Meta hopes for—or whether the company finds itself restructuring yet again in the months to come.