July 16, 2025

Mark Zuckerberg reveals Meta’s massive new AI data centre

mark zuckerberg reveals meta’s massive new ai data centre
Photo source: Flickr

Meta CEO Mark Zuckerberg has announced plans for Hyperion, a vast new AI data centre designed to provide five gigawatts (GW) of computing power to support the company’s artificial intelligence research.

This centre, likely to be built in Richland Parish, Louisiana—where Meta has already invested $10 billion in data infrastructure—aims to bring two gigawatts online by 2030 and eventually scale up to five gigawatts, a capacity comparable to multiple large power stations.

Alongside Hyperion, Meta is developing Prometheus, a 1 GW AI supercluster expected to be operational in 2026 in New Albany, Ohio. These projects are part of Zuckerberg’s ambition to build the “most elite and talent-dense team in the industry” and invest heavily in superintelligence.

The massive increases in computing power aim to position Meta competitively against leaders like OpenAI, Google DeepMind, and Anthropic.

However, the immense energy and water demands of data centres at this scale raise concerns for local communities. For example, Meta’s Newton County facility in Georgia reportedly caused water shortages for residents, while other expansions, such as CoreWeave’s planned data centre near Dallas, Texas, threaten to markedly increase local electricity consumption.

U.S. policymakers have recognised AI infrastructure as a growing energy burden. Energy Secretary Chris Wright described AI as “the next major energy-intensive frontier” and urged for expanded energy production across coal, nuclear, geothermal, and natural gas to meet rising demand. Data centres are forecast to consume around 20% of U.S. electricity by 2030, up from 2.5% in 2022.

Government support, including endorsements from President Donald Trump, has facilitated AI infrastructure projects like OpenAI’s Stargate initiative with Oracle and SoftBank. Other players, such as Elon Musk’s xAI with its “Colossus” supercomputer, are also quickly scaling their computational capabilities.

Subscribe for weekly news

Subscribe For Weekly News

* indicates required