Facebook parent Meta Platforms is planning to spend hundreds of billions of dollars on artificial intelligence data centre infrastructure, and will seek to bring its first multi-gigawatt compute facility online next year, the firm’s chief executive Mark Zuckerberg said.
Zuckerberg has personally led a drive to poach top engineering talent from other companies in recent months, following a disappointing reception for its most recent Llama 4 open-source AI model.
He said on social media that Meta’s new Superintelligence Labs would lead competitors in its data centre capacity and would have more compute per researcher than others.
Meta’s first multi-gigawatt data centre cluster, Prometheus, is located in Ohio and is planned to come online next year, while another called Hyperion, located in Louisiana, should be able to scale up to 5 gigawatts in the coming years, Zuckerberg said.
He said Meta was planning “multiple” additional data centre clusters, just one of which covered “a significant part of the footprint of Manhattan”.
He said the costs would be covered by Meta’s core advertising business.
Meta made nearly $165 billion (£122.8bn) in revenue last year.
The company last month reorganised its AI efforts under the new Meta Superintelligence Labs as it seeks to keep up with AI rivals such as Google and OpenAI.
The unit is Meta’s latest large-scale bet after forming Reality Labs in 2020, which has spent more than $60bn in the past five years while developing VR headsets and Ray-Ban smart glasses as well as highly-touted “metaverse” applications that have never gained popularity.
Credit: SILICONUK