← Back to context

Comment by T-A

2 hours ago

> Did your datacenter consume 9GW?

I think the largest data centers consume about 1 GW.

https://www.enconnex.com/data-center-energy-consumption-and-...

https://techplustrends.com/1gw-data-center-power-consumption...

> I think the largest data centers consume about 1 GW.

The largest AI datacenters. And there are multiple planned that consume beyond 1 GW. In particular, I'm referring to the box elder county data center which I believe is one of the largest planned. [1]

Prior to these AI datacenters, 100MW would have been considered an extremely large amount of power for a single datacenter.

But also, 1GW is a huge amount of power. More than a lot of states use, and not the really small ones like Wyoming. Just for reference, steel mills use 40MW and there's about 8 of them in the US.

1GW is an absurd amount of power to consume.

[1] https://www.tomshardware.com/tech-industry/kevin-o-learys-9-...

  • > But also, 1GW is a huge amount of power. More than a lot of states use

    Not even close to a true statement.

    EIA says that the lowest electricity use by state in 2024 was Vermont, at 5.5B kwh. That’s about 0.6GW, if you assume continuous uniform use. By the time you get to #48 (Hawaii) you’re over the 1GW mark [1].

    According to the EIA, US utility-scale generation capacity in 2025 was 1.3 million megawatts [2], so assuming that these new 1GW datacenters were on the public grid, they would each consume about 1/1300th of the 2025 supply. Not nothing, but not nearly the dramatic impact that you’re suggesting.

    For perspective, the Grand Coulee Dam has a generating capacity of about 7,000MW. Robert Moses dam in Niagara Falls does about 2,700MW.

    [1] https://www.eia.gov/state/seds/data.php?incfile=/state/seds/...

    [2] https://www.eia.gov/energyexplained/electricity/electricity-...

    • Fine, an exaggeration but not off by much.

      If the Box Elder datacenter were a state it'd rank #19 in terms of power consumption (78 TWh).

      It is correct to say these massive data centers will represent a significant percentage of US power consumption when built out.

      2 replies →

  • > 1GW of power datacenters, more efficient GPUs, more efficient training, Xiaomi/Deepseek disk based caching + other efficiency gains.

    I haven't done the math but if I see how many tokens I can burn per month for 100 USD it's pretty mind boggling to imagine how much capacity is being added by this build out.

    This is not the place to discuss it but even in my wildest predictions I find it hard to imagine that we're really going to need all this capacity anytime soon (although I guess anything spare can go into training).

    • It’s not ‘spare goes to training’. These data centres are training first. Inference is really rather efficient, so not why datacenters are so big.