[USA] Energy Department’s advisers call for agency’s own AI data center testbed

On July 30, 2024, the Department of Energy (DOE)’s advisory board proposed the establishment of its own AI data center to study how to lessen energy usage, according to a recent report. [1] This is relevant because the rapid growth of energy requirements from AI data centers, as well as the threat they present to climate goals, has landed on the department’s main agenda. The advisory board calls on the DOE to create a “data-center-scale AI testbed” in the agency for two reasons: it would allow scientists from academia, the DOE, and the industry, to help make these data centers more energy efficient and flexible in taxation, and it would be separate from the agency’s pre-existing computing facilities. The report focuses on hyperscale facilities that require between 300 to 1,000 megawatts of electricity or greater, which are being built with lead times of only 1-3 years. The board consulted with big technology firms, power suppliers, and other stakeholders, and concluded that there is an urgent need for a more flexible and firm electricity supply.

[1] https://www.energy.gov/sites/default/files/2024-08/Powering%20AI%20and%20Data%20Center%20Infrastructure%20Recommendations%20July%202024.pdf