The Future of Meta Superintelligence: A 1 Year Progress Update

Key Points:

• Over the past year, Meta's AI division (MSL) has undergone a major rebuild following the Llama 4 release failure, involving a $14.3B investment to acquire top talent and a rapid expansion of compute capacity through new datacenter designs.
• The frontier AI landscape is currently dominated by OpenAI and Anthropic, with Google losing momentum despite recent model releases; Meta's latest model, Muse Spark, lags behind competitors but the company is focused on long-term growth driven by data, talent, and compute.
• Meta's unique advantage lies in its vast internal data collection, including employee screen recordings to create realistic reinforcement learning (RL) environments, supported by a newly formed 3,000-engineer applied AI team dedicated to building RL tasks.
• Meta is aggressively scaling its AI compute infrastructure with simultaneous construction of five 1GW+ datacenter "titan" clusters and innovative networking solutions like AI-Backbone to manage large-scale distributed training workloads across multiple campuses.
• Despite assembling a high-profile AI research and engineering team with competitive compensation, Meta remains early in its AI development journey, and future success depends on maintaining focus and resources without compromising resolve or losing key talent.