Big Tech's A.I. Data Centers Are Driving Up Electricity Bills for Everyone

6 months ago (nytimes.com)

Not just bills. These data centers, a major driver of new energy use, are contributing to climate change. Sadly it seems to be another way for large companies to offload externalities onto the public.

  • Are they? IIRC MS & Google were running on carbon neutral sources.

    •   Are they? IIRC MS & Google were running on carbon neutral sources.
      

      It's like saying "our crypto mining energy is carbon neutral" but ignoring the fact that if the carbon neutral energy is going into crypto mining, it's not going into something more useful. If crypto mining buys all the carbon neutral energy sources, then those sources aren't being used by other industries.

      We have limited carbon neutral energy production.

    • Carbon neutrality doesn’t refill a drained reservoir used to cool off these machines. Running your servers on wind power doesn’t make the millions of gallons they’re dumping into cooling systems any less gone.

      1 reply →

    • Google stopped claiming "operational carbon neutrality" shortly after the release of chatgpt.

  • To say nothing of the exorbitant amount of water used to cool these machines, we’re on track to face a water shortage crisis long before any other climate change impact.

    • Water availability is a regional climate change impact, which does not apply everywhere due to differences in water sourcing and weather patterns and how climate change affects these.

      It's very stupid to evaporate potable water on purpose in dry regions, but note that many numbers in this area are highly sensationalized by taking e.g. the maximum design capacity of the cooling system instead of the actual load, and that there are several other cooling solutions. Most proper facts die tragic deaths before they make it to mainstream news media. :/

      5 replies →

It's usually the opposite.

When a company requires lot more energy, power plants are expected to produce lot more.

When a power plant produce a lot, the low consumption rates tend to get cheaper. It's gets cheaper to produce energy as the demand increases.

This is, of course, considering the input material is not scarce, like hydro power plants or wind power. Everything else (coal, oil, nuclear, gas, solar) should be easy to increase supply/demand.

But the vast majority of my $500+ a month PG&E bill is for transmission, not generation.

  • 500 a month sounds steep. I'm assuming you live somewhere that requires AC every day?

    The article referred to driving prices up from 2020 due to making the infrastructure stronger by as much as 30%. Which, yeah, about 150ish of your bill.

    It is less clear on how much it will need to go up because of increased demand? The prediction is 8%. Which, again, not nothing. But it is telling that there is more increase from infrastructure than there is generation? I don't know that that will change?

  • And according TFA, those poles and wires for transmission are a large part of the increase in costs that are forecasted.

    Ideally, the folks who request the new plants and transmission lines pay for them, but it appears tech cos are attempting to pass the transmission cost burden onto residential consumers.

  • That isn't unique to PG&E. Clear across the country, my utility bill is close to 60% transmission fees.

    Also what are you doing? Running a flux capacitor?

    • PG&E peak residential rates hit over 50 cts/kwh.

      PG&E works with a capture PUC and cost plus accounting where the only way for the company to increase profits is to drive up expenses.

Oh hey, not even crypto pulled this kind of thing off, but here we now see crickets from certain parts of the tech community (on this site especially) about this particular gargantuan use of electricity. I guess it's no longer a problem when the people who employee you are doing it, despite the vast amount of slop, spam and other garbage being generated among the occassionally (only apparently so far) productive uses of LLM technology.

As far as that goes, i'd love to see a well researched breakdown of just what percentage of LLM technology is actually being used for anything productive, and what percentage is just being pissed into industrialized spam.

LLMs even have the whole crime angle neatly covered, considering all the innovative uses they're being put to by the same people who brought us ransoms paid in crypto. Would be interesting to see the numbers on how that breaks down too.

this is nonsense and the author even admits it

> In the coming years, artificial intelligence could turbocharge those increases

the cost of residential power is going up because of the shift away from natural gas towards solar

failing to admit this or worse lying about it is not going to actually help long term

[flagged]

  • > “… grandmother knows about AI energy demand.”

    Uh huh. Tell us more about how gramma should have planned for the AI future.

  • yes let’s listen to the well thought out and impartial criticisms of “meta_ai_x”

    obviously the big AI companies are using lots and lots of energy, making big claims like you are about “brainwashing” and such without anything to back it up is cancerous to real discussion. as are the bad faith arguments.

  • Tech companies are trying to lock the market and create monopolies to abuse later. And have pretty much established track record or genuinely not caring who or how they harm ... or even brag about it.

    Tech companies literally brag about disruption, about making groups of people obsolete and dont even pretend their compliance with rules is anything but bad faith.

    Like common here. We as a tech lost the goodwill because of the culture of people we celebrate.