Meta to spend hundreds of billions on massive AI data centres, starting with Prometheus and Hyperion in the US, aiming to develop superintelligence and rival top AI players.

Meta's founder, Mark Zuckerberg, has announced that the social media giant will invest hundreds of billions of dollars in the construction of massive artificial intelligence (AI) data centres across the United States. The first of these, a multi-gigawatt facility named Prometheus, is expected to become operational in 2026.

Zuckerberg stated that one of the upcoming sites would span an area nearly the size of Manhattan, measuring approximately 59.1 square kilometers (22.8 square miles). The development is part of Meta’s aggressive pursuit of “superintelligence” — a form of advanced AI that the company claims could surpass the intellectual capabilities of the smartest humans.

Meta, which earns the majority of its revenue from online advertising, generated over $160 billion in 2024. In a post shared on his platform, Threads, Zuckerberg revealed that Meta is building several multi-gigawatt clusters. One of these, called Hyperion, could scale up to five gigawatts over several years.

“We’re building multiple more titan clusters as well. Just one of these covers a significant part of the footprint of Manhattan,” he said.

The Prometheus cluster will be located in New Albany, Ohio, while Hyperion is slated for construction in Louisiana and is expected to be fully operational by 2030. Zuckerberg emphasized that Meta will “invest hundreds of billions of dollars... to build superintelligence,” noting that the data centres have been given “names befitting their scale and impact.”

Commenting on the development, Karl Freund, principal analyst at Cambrian AI Research, told the BBC that Zuckerberg “clearly intends to spend his way to the top of the AI heap.” Freund added, “The talent he is hiring will have access to some of the best AI hardware in the world.”

Following the announcement, Meta’s shares were trading 1% higher, according to Reuters. The stock has risen more than 20% so far this year.

There are currently more than 10,000 data centres worldwide hosting the cloud — remote servers that store digital information. The majority are located in the US, followed by the UK and Germany.

AI-driven data centres are highly resource-intensive, requiring substantial amounts of energy and water. One study estimates that these facilities could consume 1.7 trillion gallons of water globally by 2027. A single AI query — such as a request to ChatGPT — can use roughly the same amount of water as a small bottle purchased from a corner shop.