Meta Chief Executive Mark Zuckerberg announced plans to invest hundreds of billions of dollars over the next several years to build a series of massive artificial intelligence (AI) data centers — each with power demands measured in gigawatts — as part of the company's accelerating push toward developing superintelligence.
In a post on Threads, Zuckerberg said Meta has broken ground on its first gigawatt-scale facility in Ohio, code-named Prometheus, which is slated to come online in 2026. A second center, Hyperion, is projected to eventually expand to 5 gigawatts in capacity. Some of Meta's future AI clusters will span sites as large as Manhattan, he noted.
Zuckerberg said Meta's advertising business provides the financial backbone for this vast infrastructure expansion. He cited a report from chip industry research firm SemiAnalysis, which predicts Meta could become the world's first company to operate an AI compute cluster exceeding 1 gigawatt.
The infrastructure spree is part of Meta's broader ambition to build what Zuckerberg calls superintelligent. In April, the company raised its 2025 capital expenditure forecast to between US$64 billion and US$72 billion — an unprecedented investment level. That same month, Meta launched a new "Superintelligence Lab" and recruited elite AI researchers from OpenAI and Google DeepMind.
In a parallel move, Meta acquired a 49% stake in data labeling startup Scale AI for US$14.3 billion and appointed its co-founder, Alexandr Wang, as Meta's first-ever Chief AI Officer. Wang will co-lead the Superintelligence Lab alongside former GitHub CEO Nat Friedman.
Meanwhile, the company's next-generation open-source model, Llama 4 Behemoth, originally expected to launch in April, has been delayed. According to sources cited by The New York Times, internal dissatisfaction with the model's limited improvements has prompted Meta to consider replacing it with a closed-source alternative — a notable departure from its prior open-source approach.
Meta's bold investments position it as a major contender in the global race for AI dominance — but they also raise questions about the sustainability, governance, and ultimate direction of the superintelligence arms race.
Article edited by Jack Wu
