Meta’s Zuckerberg Says New AI Data Center ‘Supercluster’ Is Set to Come Online Next Year

Meta Platforms (META) CEO Mark Zuckerberg said Monday that his company expects its first in a series of AI data center “superclusters” to come online in 2026.

In a Facebook post, Zuckerberg said Meta will spend “hundreds of billions of dollars” on setting up the computing power necessary to build “superintelligence,” a type of AI that has more capabilities than a human brain. Meta has said it expects to spend $64 billion to $72 billion this year on capital expenditures as it turbocharges its AI capacity.

Meta’s first facility with more than a gigawatt of capacity has been dubbed Prometheus, and others are currently being built, Zuckerberg said. The CEO expects another such cluster, known as Hyperion, to reach up to five gigawatts over several years.

Meta’s CEO has been personally involved in recruiting new hires for the superintelligence unit, including offering lucrative pay packages to poach AI talent. Last week, Meta reportedly offered a pay package valued at more than $200 million to lure an Apple (AAPL) executive in charge of AI models, and its other recent hires have included former Github CEO Nat Friedman and ex-Scale AI CEO Alexandr Wang.

Seeking ‘Most Elite and Talent-Dense Team’

“For our superintelligence effort, I’m focused on building the most elite and talent-dense team in the industry,” Zuckerberg wrote.

The heavy investment in Meta’s superintelligence team comes as Zuckerberg has reportedly expressed frustration with the company’s progress in AI development. In May, Meta reportedly pushed back the launch of its latest Llama 4 large language model amid concerns about whether enough improvements had been made compared with previous iterations.

Shares of Meta edged 0.5% higher to close just under $721 Monday. The stock has added close to one-quarter of its value in 2025.