Meta will invest hundreds of billions of dollars to build some of the world’s largest artificial intelligence data centres in the United States, as part of its bold push to develop “superintelligence”, CEO Mark Zuckerberg has announced.

The company’s first such facility, named Prometheus, will go online in 2026 and is being built in New Albany, Ohio. Another centre, Hyperion, is set for Louisiana and will be capable of scaling up to five gigawatts over the next several years. Zuckerberg said it would eventually occupy a site nearly the size of Manhattan — around 59 square kilometres.

“We’re building multiple more titan clusters as well,” Zuckerberg wrote in a post on Threads, Meta’s social media platform. “Just one of these covers a significant part of the footprint of Manhattan.”

The scale of Meta’s ambition signals a major leap in its AI efforts. Zuckerberg described the data centres as essential infrastructure to power what he termed “superintelligence” — AI systems capable of surpassing the cognitive abilities of even the smartest humans.

The tech giant has been investing heavily in what it calls Superintelligence Labs, hiring top talent from leading AI companies such as OpenAI and Google DeepMind. The team is reportedly being led by Scale AI co-founder Alexandr Wang.

Meta, which generated more than $160 billion in revenue in 2024 — largely from online advertising — is now staking its future on being a global leader in artificial general intelligence (AGI), a form of AI that can perform any intellectual task a human can do, across multiple domains.

In his post, Zuckerberg confirmed that Meta is constructing several multi-gigawatt “clusters” with names “befitting their scale and impact.” Alongside Prometheus and Hyperion, others are expected to follow, each likely to be among the largest data centre projects ever attempted.

“Clearly, Zuckerberg intends to spend his way to the top of the AI heap,” said Karl Freund, principal analyst at Cambrian AI Research, speaking to the BBC. “The talent he is hiring will have access to some of the best AI hardware in the world.”

Recommended reading

• Users Accidentally Leak Personal Data via Meta AI
• ChatGPT Search Inches Toward Strict EU Oversight as Users Surge
• AI Fuels Consumer Demand for Stronger Privacy Laws

While the announcement sent Meta’s stock up by 1% — adding to a 20% increase so far this year — it also drew attention to the environmental footprint of AI infrastructure. Data centres, especially those powering large-scale AI models, are extremely energy and water intensive. A single AI query can use as much water as a small bottle of mineral water, and one study estimates global AI data centres could consume up to 1.7 trillion gallons of water annually by 2027.

Currently, there are more than 10,000 data centres worldwide, with the majority located in the US, followed by the UK and Germany. Most generate only a few hundred megawatts. Meta’s gigawatt-scale facilities — each potentially 10 times larger — mark a radical new phase in global computing infrastructure.