Meta is taking a historic leap into artificial intelligence by unveiling its most ambitious infrastructure initiative to date: a multi-gigawatt data supercluster named Prometheus, set to launch in 2026. CEO Mark Zuckerberg announced the project, revealing that Meta has earmarked between $64 billion and $72 billion in capital expenditures for 2025 alone to support this AI push.
Prometheus will deliver over one gigawatt of computational power, making Meta the first company to deploy AI-ready hardware at such a massive scale.
Building Gigawatt Giants for AI Dominance
In addition to Prometheus, Meta is also building Hyperion, a second supercluster expected to scale up to five gigawatts over the coming years. Located in Richland Parish, Louisiana, Hyperion complements Prometheus’s deployment in New Albany, Ohio as part of a $10 billion phased rollout.
Meta is using unconventional construction methods to fast-track deployment and scale compute rapidly. These clusters form the backbone of Meta’s Superintelligence Labs and will support future versions of its LLaMA models.
Meta Bets Big on AI Dominance, Infrastructure Costs
Industry experts estimate that each gigawatt of AI compute capacity costs around $30 billion, placing Meta’s infrastructure among the largest investments in tech history. With both Prometheus and Hyperion, Meta is laying the groundwork for an AI arms race.
Zuckerberg Luring Top AI Tech Leads
To support its superclusters, Meta is aggressively hiring from OpenAI, Google DeepMind, and Apple. Its Superintelligence Labs, now a central pillar of Meta AI, are led by industry heavyweights like Alexandr Wang and Nat Friedman. Meta is offering compensation packages worth up to $200 million over four years to attract elite researchers.
By scaling infrastructure and talent simultaneously, Meta aims to reclaim its relevance in the global AI race after recent setbacks with LLaMA 4 and internal staff churn.