Step 2: find “thing 2” with property x closest to (result of Step 1) && x <= (result of Step 1)
Step 3: Write clickbait.
Optional: Publish some permutations - resample list of “thing 1” to yield min(number of “thing 1”) matching “thing 2” for each “thing 2” in list
>The team then modeled these facilities’ energy use in a variety of scenarios and calculated the energy mix of the regional grids to estimate individual facilities’ carbon emissions.
So is it just made up? The facilities energy usage that is?
Downstream re-reporting (Wired, NYT, etc) mixes these interchangeably, or picks 1 and treats it the same as the other.
AI on the other hand..
The point still stands that 400 of the gazillion datacenters in the US use more energy than the entire population and industry of an entire state. That's a ridiculous amount of energy to pour into any number of datacenters.
Why? Your objection sounds arbitrary.
Its citing Hyperscaler Data Centers as 1.8% of the US's power usage, right? Sounds like Michigan just doesn't use that much power.
Is it? What are you comparing it to? Someone up a thread compared it to the top 10 steel plants and it turns out they use more energy than all the data centers.
This is like the water discussion where people discuss absolute values of consumption for huge scale industries and then put that in front of people that have no context to generate outrage.
I would assume you wouldn't select the 400 smallest data centers and 400 random data centers would be weird to write an article about, so I defaulted to thinking the 400 largest(with available data). Was this assumption wrong? Where is the deception? Is it Michigan?
Here's how it was misleading to me personally: when I heard "Just 400 Data Centers" I interpreted that initially as there are lots of data centers, say, 1000, but it only takes 400 of them to equal the power of Michigan, so wow! the full set of data centers must use an enormous amount of power!
But now this comment clarified that the initial study looked at all the hyperscalars in the US (403). So it wasn't like it was choosing a subset to equal a known amount of power, it was putting all the power into context. If the headline were instead: "All the Hyperscalar Data Centers in the US equal the power usage of Michigan" it doesn't kind of gesture at a much larger amount of power being used.
Taking the headline at face value, you might imagine that there are 4,000 data centers and they chose the biggest 400 and said that they consume as much power as Michigan. That would lead you to believe that all the data centers would consume multiple states worth of power because that's only 10% of them.
But that's not the case. This isn't 10% of all data center electrical usage. It's more like 90% of all data center electrical usage (or something). That's the distinction that's being discussed.
When did we start doing that. and what do you mean by a lot higher? Two Michigans? Is that deceptive?
>That would lead you to believe that all the data centers would consume multiple states worth of power because that's only 10% of them.
Why would someone assume that? Cause the 400 data centers are just a random sampling?
>That's the distinction that's being discussed.
Is it?
I said the percentage was a lot higher. I'm measuring percentage using integers between 0 and 100. I'm not sure how you measure percentage in "Michigans". There's a unit error there.
> Why would someone assume that?
They would assume the percentage is small because of the use of the word "just". Again, the 400 was not a random sampling.
> Is it?
Yes, you responded to a comment that was focusing on that.
"One thing to note is that the framing "Just 400 US Data Centers" makes it seem like 400 out of the gazillion data centers use this much electricity, but the original research seems to be that they found exactly 403 "Hyperscale Data Centers" and evaluated them."
That's the current topic. You seem to be working pretty hard to not get the point that's being made.
I replied to them because they said it was deceptive and no one can give me a concrete example of this deception.
>the data did not allow them to differentiate between power used for general-purpose versus AI-specialized operations at each HDC
That seems like an important distinction, given how this study will likely be used by DC opponents.
>HDC electricity demand corresponded to approximately 1.8% of total U.S. electricity consumption
I don't see a problem with this. The only real issue is in cases of added load to local grids increasing costs. It is useful to look at historical energy production in the US.[0] If anything the period of stagnation starting from 2007 is an anomaly. That stagnation was probably caused by the GFC, de-industrialization and improved energy efficiency. I don't think anyone protested the 25% increase in electricity production from 1990 to 2000, for example.
By the way, your first link appears to be cut off, full link is this:
Hyperscale data centers’ energy consumption, carbon emissions newly estimated by researchers https://hsph.harvard.edu/news/hyperscale-data-centers-energy...
[0] https://www.eia.gov/totalenergy/data/browser/?tbl=T07.01#/?f...
At this point I am not even sure why should we continue to read anything from mainstream media. Other than some sort of civil duty to continue correct them or bring it context / communities notes.
I am also wondering why Datecenter and AI is now becoming the new Anti-Christ / Nuclear Reactor.
Heavy industry gets/got a lot of pushback, and that brings both local jobs and outputs that are essential to many.
Datacenters, on the other hand, bring basically no local jobs once construction is finished, and--worse even--their output is perceived to compete with the job people already have to some degree.
With a headline like this, I'd always expect the 400 datacenters to be among the biggest, not a random sample (or the smallest, even). Why would any piece of media ever deliberately make their headline less "interesting" (sensationalist)?
Deceptive framing would've been: "5 datacenters use more electricity than Michigan" which could've still been technically correct assuming 5 datacenters the size of the currently unfinished AWS New Carlisle campus (2.3GW).
One thing to note is that the framing "Just 400 US Data Centers" makes it seem like 400 out of the gazillion data centers use this much electricity, but the original research seems to be that they found exactly 403 "Hyperscale Data Centers" and evaluated them.
The framing is technically correct (since there exist non-hyperscale data centers) but is clearly intended to be deceptive.
[1] https://hsph.harvard.edu/news/hyperscale-data-centers-energy...