Zuckerberg Confirms Meta Will Spend Hundreds of Billions On AI Data Centres

Meta CEO Mark Zuckerberg has confirmed the company will invest “hundreds of billions of dollars” to construct multiple next-gen data centres aimed at powering superintelligent AI systems.

The first site, dubbed Prometheus, is expected to come online in 2026.

Zuckerberg posted on Threads that Meta is developing several massive “titan clusters” across the US, including a flagship facility in Louisiana “nearly the size of Manhattan.”

The facilities will house supercomputing infrastructure at the gigawatt scale, which is far beyond the capacity of traditional data centres.

The move is part of Meta’s all-in push on generative AI. In April, the company lifted its 2025 capex forecast to as much as US$72 billion, citing infrastructure and AI as top priorities.

According to analyst group SemiAnalysis, Meta is on track to be the first company to deploy a supercluster exceeding one gigawatt in capacity, which is enough to power nearly 900,000 homes.

This infrastructure arms race comes as Meta reorganises its AI efforts under a new division called Superintelligence Labs.

In recent months, Zuckerberg has led a talent raid by poaching leading researchers from OpenAI, Google DeepMind and Apple. He also appointed Scale AI co-founder Alexandr Wang as Meta’s Chief AI Officer, following a US$14.3 billion investment in the startup.

Despite investor concerns over the major expenditure, Zuckerberg says Meta’s robust ad business will continue to fund its AI ambitions.

Meta shares have climbed over 20% in 2025 to date, with the company banking on its advanced AI tools to drive future revenue, from smart glasses to image-to-video ad platforms.

Meta’s data centre ambitions put it in direct competition with OpenAI, Nvidia, and Oracle.