4 comments

  • monarchwadia 15 minutes ago ago

    It is a guesstimate, and measures 400+ data centers. A lot of this is just an educated guess. The estimate has an incredibly wide range.

    Minor note, the title also makes it easy to confuse the issue and think that a single hyperscaler has this kind of energy requirement.

    From the article:

    ----

    Using satellite imagery to verify the locations of 403 hyperscale data centers and power-plant records to estimate their emissions, the Harvard-led team found that the facilities’ electricity use was between 68 and 99 terawatt-hours in the year ending April 2025.

    Based on that estimate, just those data centers would have accounted for 1.8% of the US’s total electricity consumption that year.

    At the high end, it's close to the energy that states like Michigan or Tennessee use in a year. More concerningly, that power generation would have added 25.7 million metric tons of carbon dioxide to the atmosphere – more than Denmark’s entire emissions for the same period.

  • dhx 2 hours ago ago

    The paper behind this article estimates 403 hyperscale data centres in the US they studied had an electricity demand of 82TWh between May 2024 and April 2025, with Virginia's data centre market alone taking a 21.4TWh share.[1]

    To contrast with China, their NEA published official figures for January-September 2026 (9 months) indicating (with some uncertainty due to translation) that their country-wide data centre demand was 79.3TWh over these first 9 months of year, and 9.8TWh for September alone. Hence the total Chinese data centre market perhaps could be estimated to be approximately 120TWh/y current consumption.

    Are there any other statistics more recent than April 2025 for total US data centre market electricity consumption? I couldn't find anything useful from a quick search.

    [1] Guidi, G., Dominici, F., Squartini, T. et al. Assessing the carbon emissions and energy consumption of US hyperscale data centres. Nat Sustain (2026). https://doi.org/10.1038/s41893-026-01985-7 https://www.nature.com/articles/s41893-026-01985-7

    [2] https://www.nea.gov.cn/20261023/b1ac9a8faea04deebf4c43a7dcfe...

  • kiruxa211289 4 hours ago ago

    [flagged]

  • topham 3 hours ago ago

    And?