What is Prometheus? Meta Announces First AI Supercluster, Expected to Go Live by 2026

To attain its goal of becoming the global leader in artificial intelligence, Meta is introducing the AI Supercluster named Prometheus. Mark Zuckerberg unveiled the supercluster, which will go live by 2026. This is one of Facebook’s most ambitious infrastructure projects yet.

Zuckerberg described the company’s approach to scaling up in a lengthy Facebook post, highlighting massive investments in computing power, rapid infrastructure expansion, and the recruitment of the world’s top AI researchers. Prometheus will soon be joined by Hyperion, a future 5-gigawatt AI system that is set to become one of the largest and most powerful AI clusters in the world.

What makes Meta’s approach unusual is its urgency. Rather than waiting for the permanent infrastructure to be built, Meta has started constructing temporary data centers using tents. Quicker to establish, these tented facilities are used despite some loss of environmental control in extreme hot weather.

Zuckerberg also unveiled Meta Superintelligence Labs, a new internal division intended to provide more computing power per researcher than anyone else. Its laboratory seeks to assemble the largest aggregation of AI talent anywhere in the world. As part of this effort, Meta has begun to offer compensation packages exceeding $100 million to woo top-tier AI talent from competitors like OpenAI and Apple.

After an underwhelming reception of its Llama 4 AI model earlier this year, Meta has launched a complete strategic reset. It recently invested $14 billion in Scale AI. This investment is key to advancing the next generation of AI models, which are expected to be more accurate, efficient, and globally competitive.