Behind artificial intelligence (AI) are countless supercomputers and data centers, essential to its operation. “Making France a country at the forefront of AI” will therefore necessarily be accompanied by the construction of new data centers in France, the Artificial Intelligence Commission underlines in its report submitted to Emmanuel Macron on Wednesday, March 13.
“Europe’s dependence on the United States, which is already strong in data centers, is even more striking in terms of computing power”, the document underlines. The United States has 2,109 data centers, or a global share of 37.8%, compared to 1,244 for all European Union countries (22.3%), and about 250 in France.
To encourage one “to startle”the committee proposes various measures. Firstly, simplifying the procedures for installing a data center, such as those benefiting projects of great national interest, as defined in the Green Industry Law of 2023. In particular, this status allows under the jurisdiction to fall from the state, and no longer from the mayor, planning permits for everyone “industrial project which, given its purpose and size, especially in terms of investment and employment, is of particular importance for the ecological transition or national sovereignty”.
Other measure: “Make sure that the computer and data center sector qualifies for electro-intensive status. » The latter, created by the 2015 Energy Transition Act, allows companies that use a lot of electricity “benefit from special conditions for electricity supply for all or part of their locations. In return, they commit to applying best practices in energy performance”.
“A significant expense”
The committee touches on a sensitive point here. Data centers are already very demanding in terms of electricity, but will become even more so with AI, according to a report published on January 24 by the International Energy Agency: their consumption could increase by 460 terawatt hours (TWh) in 2022, or 2% of global demand. (including 25% for cryptocurrencies), at 1,050 TWh, a jump equivalent to the consumption of an additional country, on the order of that of Germany.
“Electricity consumption represents significant costs for market players: it is therefore a key factor in the location decision”, the AI committee acknowledges. For example, she explains that the German DeepL, specialized in translation tools, has chosen to install its brand new supercomputer in Sweden in 2023, due to the most low-carbon energy mix in Europe – fossil fuels, which represent 1% of electricity production – and a low electricity price.