Meta CEO Mark Zuckerberg announced an aggressive acceleration of the company’s AI infrastructure roadmap, revealing plans to invest “hundreds of billions of dollars into compute” for developing artificial superintelligence. In a blog post this morning, Zuckerberg said Meta will deploy some of the most powerful AI superclusters ever built, leveraging the company’s financial resources to lead in this domain.
According to Zuckerberg, Meta’s first major AI supercluster, dubbed Prometheus, is slated to come online in 2026 and will surpass 1 gigawatt (GW) in power draw, making it potentially the first AI compute cluster at that scale. A second cluster, Hyperion, will scale up to 5GW over several years, with other “titan clusters” already in development. Each is expected to span a footprint comparable to a large portion of Manhattan.
Zuckerberg stated that Meta’s Superintelligence Labs will achieve “the greatest compute per researcher,” underscoring a strategic focus on attracting elite AI talent. “We’re also going to invest hundreds of billions of dollars into compute to build superintelligence. We have the capital from our business to do this,” he wrote. The move positions Meta as a front-runner in the race toward superintelligence infrastructure, with scale and compute density as central pillars.
- Prometheus supercluster expected to go live in 2026 with >1GW of power
- Hyperion to scale up to 5GW over several years
- Several additional “titan” AI clusters are in development
- Each cluster spans a footprint comparable to parts of Manhattan
- Meta aims to lead in compute per AI researcher
“We’re actually building several multi-GW clusters,” Zuckerberg wrote. “I’m looking forward to working with the top researchers to advance the frontier!”
