Meta founder Mark Zuckerberg announced that hundreds of billions of dollars will be invested in the coming years to build massive data centers that will power artificial intelligence technologies. Zuckerberg announced that the first of these very large-scale data centers, to be built in the US, will be operational in 2026.

Meta is spending hundreds of billions of dollars on massive AI data centers.

The first facility, Prometheus, will reportedly be located in Ohio, while the second, Hyperion, will be built in Louisiana and will be fully operational by 2030. Zuckerberg stated that each of these data centers will reach several gigawatts of energy capacity, with some spanning an area the size of Manhattan Island. Prometheus will be the first center with a capacity of more than one gigawatt, while Hyperion could eventually reach up to 5 gigawatts.

Meta stated that these investments will enable what he calls “superintelligence,” artificial intelligence technologies capable of surpassing human intelligence. In a statement on Threads, Zuckerberg stated that these data centers are named after their scale and that Meta has allocated significant budgets to reach the top of the line in this field.

Emphasizing the critical importance of this infrastructure for artificial intelligence hardware and large data sets, Cambrian AI Research principal analyst Karl Freund stated that Zuckerberg has recruited the world’s best AI engineers and provided them with top-tier hardware.

Following Meta’s announcement, the company’s stock value increased by 1 percent. Shares have gained more than 20 percent throughout 2024, with the company generating more than $160 billion in revenue that year.

However, such data centers also have a significant environmental impact. AI-based data processing infrastructures are reportedly significant resources in terms of water consumption as well as energy. According to studies, these centers could use a total of 1.7 trillion gallons of water worldwide by 2027. It is stated that a single AI query, for example a question posed to ChatGPT, consumes as much resources as a small bottle of water.