Meta has shifted to a talent- and infrastructure-centric model by reorganizing into Meta Superintelligence Labs, concentrating on attracting top researchers with generous compensation and leveraging substantial compute resources.

This pivot follows the underwhelming Llama 4 performance and concerns over benchmark handling, which prompted leadership changes, including Yann LeCun’s departure and the creation of the TBD Lab led by Alexandr Wang.

The company identifies compute per researcher and talent density as bottlenecks and plans multiple large clusters to sustain ongoing AI work, signaling a long-term, capital-intensive strategy.

Meta highlights ambitious compute projects like the Prometheus and Louisiana Hyperion campuses as evidence of its large-scale compute ambitions, aiming to fund them from advertising-driven cash flow.

CEO Mark Zuckerberg says his primary role is to recruit the best people and provide massive compute and infrastructure, not to dictate research ideas.

He reiterates that he cannot—and does not—select research directions, instead trusting researchers and infrastructure to drive AI progress.

Meta laid off roughly 600 people from the Superintelligence Labs in October as part of organizational and strategic realignment in AI development.

LeCun’s exit underscores a contrast between his open academic approach and Wang’s more controlled, infrastructure-driven model at Meta.