In the November 2024 issue of Wiley Journal Climate and Energy, Senior Managing Director Jeff D. Makholm and Director Laura T.W. Olive examine the growth of data center electricity use in the United States. “Hyperscale” data centers accompanying the rapid growth of artificial intelligence (AI) are one of the fastest growing sources of electricity demand in the US and the rest of the world. Indeed, such growth, with the retirement of dispatchable fossil fuel electricity plans, drives recent warnings about potential reliability problems for the US power system. The prospect is proving disruptive at the federal and state regulatory levels for transmission owners and local utilities.
Local electricity prices and ready connectivity to pre-existing fiber optic networks drive the geographic location decision for data centers. The US not only benefits from an extensive submarine network directly connecting Europe, Latin America, and Asia-Pacific but also holds one of the most developed terrestrial fiber optic systems globally.
The authors note increases in energy efficiency and estimates that AI use of electricity for training models may plateau as AI models become more efficient mean data center electricity demand may not be as large as expected today.
The authors emphasize data center problems reflect the looming problems associated with the administrative constructs cobbled together to deal with electricity restructuring 25 years ago. Hyperscale data centers represent potentially huge energy loads with a great ability to choose among competing geographic locations. The authors conclude that whether central planning (and ratepayers) or unregulated interests bear the risk of hyperscale data center development is the fork in the road for the FERC as it investigates this pressing issue in its November 2024 Technical Conference.