Meta CEO Mark Zuckerberg recently announced plans to build massive AI superclusters in Louisiana and Ohio, with a goal of powering its new Superintelligence Labs initiatives.
The first of these, called “Prometheus,” will be a 1GW supercluster scheduled for launch in 2026. In parallel, the “Hyperion” supercluster is set to expand from 2GW to 5GW over the coming years. The Hyperion site in Louisiana is particularly notable, as it will span an area comparable to the size of Manhattan — making it one of the largest AI infrastructure projects globally.
Zuckerberg said that Meta is investing “hundreds of billions” into computing power to achieve the highest compute-per-researcher ratio in the industry. Additionally, Meta is reportedly discussing a pivot from open-source models to more closed, proprietary AI systems.
This shift could signal a dramatic redirection from Meta’s previous LLaMA open-source approach. With large-scale investments in both infrastructure and talent, Meta appears determined to establish a formidable presence in the AI domain.