Meta spent billions poaching top AI researchers, then went completely silent. Something is cooking.

In mid-2025, Meta made an unprecedented move into the AI arms race, personally recruiting the core architects behind OpenAI's GPT-4o and o1-series models, as well as key Gemini researchers, with compensation packages reportedly reaching $100 million. The company announced a massive $14 billion investment, forming the Meta Superintelligence Labs (MSL) and committing to building a state-of-the-art, 1-gigawatt AI training cluster in Ohio. This signaled a clear intent to compete directly at the frontier of artificial general intelligence (AGI).

Since that explosive announcement, the public-facing progress has stalled into near-total radio silence. The anticipated flagship model, a 2-trillion parameter system codenamed 'Behemoth,' has been delayed multiple times with no public timeline. The MSL organization itself has undergone four major restructurings in just six months, accompanied by the high-profile departure of AI pioneer Yann LeCun and some of the newly hired talent. The release of Llama 4 was considered underwhelming against this backdrop of grand promises.

This prolonged quiet from a team of that caliber, sitting on what is likely one of the world's largest dedicated AI compute clusters (codenamed 'Avocado'), presents a stark dichotomy. Industry observers are left to wonder if this represents the most expensive operational chaos in tech history or a deliberate, secretive incubation period for a breakthrough product. The fact that architects of systems like ChatGPT and the o-series remain suggests they are working on something they believe will be transformative, waiting for the right moment to launch an undeniable contender in the AGI race.