Meta Platforms Chief Executive Officer Mark Zuckerberg said the company is building several multi-gigawatt artificial-intelligence data-centre clusters in the United States, marking one of the largest infrastructure bets yet in the race for so-called super-intelligence. The first site, dubbed Prometheus, is slated to come online in 2026 with capacity above one gigawatt. A second project, Hyperion, is designed to scale to as much as 5 GW over several years—roughly the power draw of a midsize U.S. city. The hardware build-out is being paired with an aggressive talent campaign. Bloomberg and Reuters report that Meta has hired former Apple AI researchers Mark Lee and Tom Gunter for its newly created Superintelligence Labs, joining Ruoming Pang, the head of Apple’s Foundation Models group whom Meta lured earlier this month with a multiyear package exceeding $200 million. About 30 of the 3,400 employees in the division now report directly to Chief AI Officer Alexandr Wang, most of them recruited from rival labs. Zuckerberg has pledged to invest "hundreds of billions of dollars" in compute capacity as Meta seeks to keep pace with OpenAI, Google and other leaders in generative AI. The move underscores intensifying competition for both electricity and engineering talent as companies race to develop systems that can outperform humans on a broad range of tasks.