Meta’s founder, Mark Zuckerberg, has said the social media giant will spend hundreds of billions of dollars on building huge AI data centres in the US.

The first multi-gigawatt data centre, called Prometheus, is expected to come online in 2026, Zuckerberg said.

He said one of the sites would cover an area nearly the size of Manhattan (59.1 sq km/22.8 sq miles).

Meta has invested heavily in efforts to develop what it called “superintelligence”—technology “that it said could outthink the smartest humans.

The company, which has made most of its money from online advertising, generated more than $160bn in revenue in 2024.

In a post on his social media platform, Threads, Zuckerberg said Meta was building several multi-gigawatt clusters, and that one cluster, 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 added.

Prometheus will be built in New Albany, Ohio, while Hyperion will be built in Louisiana and is expected to be fully online by 2030, Zuckerberg said.

He said Meta would “invest hundreds of billions of dollars… to build superintelligence” and that the centres had been given “names befitting their scale and impact”.

Karl Freund, principal analyst at Cambrian AI Research, told the BBC, “clearly, Zuckerberg intends to spend his way to the top of the AI heap”.

“The talent he is hiring will have access to some of the best AI Hardware in the world,” Freund added.

Meta shares were trading 1% higher following the announcement, Reuters news agency reported. The stock has risen more than 20% so far this year.

There are at least 10,000 data centres around the world hosting the cloud, remote servers that store digital information, with most of them located in the US, followed by the UK and Germany.

AI-driven data centres are extremely energy and water intensive. One study estimates that these centres could consume 1.7 trillion gallons of water globally by 2027.

A single AI query for example, a request to ChatGPT—can use about as much water as a small bottle you’d buy from the corner shop.