Meta, the parent company of Facebook, is gearing up to invest hundreds of billions of dollars in the construction of extensive AI data centres in the United States, as announced by CEO Mark Zuckerberg. The first of these high-capacity centres, named Prometheus, is projected to be operational by 2026. Zuckerberg revealed that these facilities would be monumental in size, with one site almost equivalent to the dimensions of Manhattan, covering 59.1 square kilometers (22.8 square miles).
This push for "superintelligence," an initiative aiming to create AI that surpasses human intelligence, marks a bold direction for the company historically buoyed by advertising revenues, which exceeded $160 billion in 2024 alone. In a social media update on Threads, Zuckerberg detailed plans for multiple clusters, including one named Hyperion, projected to expand to a total of five gigawatts over the years.
Prometheus will be situated in New Albany, Ohio, while Hyperion's location is set for Louisiana, with an aim to be fully operational by 2030. Zuckerberg emphasized that these facilities will not only represent a substantial investment in AI technology but will also serve as landmarks reflective of their significance and scale within the industry.
Analyst Karl Freund from Cambrian AI Research remarked on Zuckerberg's strategy, noting that the plethora of talent being recruited will have access to some of the world's most advanced AI infrastructure. Following the announcement, Meta's stock saw a slight increase of 1%, reflecting a year-long rise of over 20%.
Currently, the global landscape of data centres includes around 10,000 facilities, predominantly in the U.S., followed by the UK and Germany. However, the intense energy and water demands of AI-driven data centres have raised concerns, with projections suggesting a potential consumption of 1.7 trillion gallons of water worldwide by 2027. Notably, processing a single AI query might equate to the water usage of a small bottled drink.
