Meta invests hundreds of billions of dollars in manhattan size data centers

Chatbots have become a daily tool for many people. What many forget, or which they have only a vague conscience, is that artificial intelligence consumes a lot of energy, and requires a large amount of servers. If it is often difficult to put advanced figures on AI data centers into perspective, a new META announcement makes it possible to realize the extent of the phenomenon.

Mark Zuckerberg has announced that Meta plans to invest several hundred billions of dollars to build several data centers in the United States in his quest for “superintendent” which would go beyond human intelligence. This pharaonic project will include multiple clusters of several power gigawatts. It will start with a first data center called Prometheus, and which should go into operation next year.

Data centers almost as large as Manhattan

Another center will be called Hyperion and will rise over several years to reach five gigawatts. If the name is a reference to the Titan, it is difficult not to think of Cantos of Hyperion From Dan Simmons, where AI uses human brains as scope … hope that if Meta manages to create a superintendent, she will not turn against humanity.

In addition to the cost and the titanic power of these data centers, they are also extremely large. Only one of these Titan type clusters would occupy most of the Manhattan district of New York, just under 60 km². And Meta will not stop with Prometheus and Hyperion, and plans to “multiple”Additional clusters.

Mark Zuckerberg also cited a semiianalysis report which indicates that Meta should be the first company to create a “superclusive”More than a Gigawatt. Meta has reorganized its research teams around artificial intelligence in a new structure called Meta Superintelligence Labs, directed by Alexandr Wang. The money for this investment would come directly from Meta chests, Zuckerberg having declared“Our company has the necessary capital to achieve this”.

Difficult to know if this investment will be sufficient to reach the superintendent. Some, like Sam Altman, the boss of Openai, do not hesitate to announce an imminent arrival of general artificial intelligence (IAG) in the coming years, at most. Others, like Yann Lecun, think that IAG arrives in the near future, but that large language models currently used for generative artificial intelligence will not allow it to be achieved and that it will take another approach.