cross-posted from: https://slrpnk.net/post/32465427
Datacentres consume just 1% of the world’s electricity but may soon demand much more. Their share of US electricity is projected to more than double to 8.6% by 2035, according to BloombergNEF, while the IEA projects datacentres will account for at least 20% of the rich world’s growth in electricity demand to the end of the decade.
“This idea that the lower cost of renewables alone will drive decarbonisation – it’s not enough,” said Daly. “Because if there’s a huge source of energy demand that wants to grow, it will land on these stranded fossil fuel assets.”
Tech companies have resisted pressure to provide detailed data on their AI energy footprints,
The IEA estimates that AI could boost technically recoverable oil and gas reserves by 5% and cut the cost of a deepwater offshore project by 10%. Big oil is even more bullish. “Artificial intelligence is, ultimately, within the industry, going to be the next fracking boom,” Mike Sommers, head of the American Petroleum Institute, told Axios.
At the same time, the oil and gas industry says AI can cut its carbon intensity, for instance by analysing satellite data to spot methane leaks. But even here, critics say there is a gap between digital insights and corporate actions.
“This idea that the lower cost of renewables alone will drive decarbonisation – it’s not enough,” said Daly. “Because if there’s a huge source of energy demand that wants to grow, it will land on these stranded fossil fuel assets.”
Corporations don’t think in savings, they think in budgets. If the cost of electricity goes down, thanks to e.g. renewables being cheaper, that just means that they can afford to use more electricity.
AI is exacerbating the damage we are causing by staying on fossil fuels. I see the issue is mainly political as we have green energy and all
“This idea that the lower cost of renewables alone will drive decarbonisation – it’s not enough,” said Daly. “Because if there’s a huge source of energy demand that wants to grow, it will land on these stranded fossil fuel assets.”
There’s also a huge issue around water use which isn’t mentioned in the article.
The water use issue is largely blown out of proportion or not compared to other sources of water use.
The US Corn industry alone uses 80x the water usage of the entire global AI industry.
Hank Green on YT goes over it in more details and with citations: https://www.youtube.com/watch?v=H_c6MWk7PQc
Fair, but corn has a use. (Also not one I agree with but you get my meaning.)
Only 1% of the corn grown in the US is eaten by people. Most of it is used as animal feed (which is catastrophic on their systems as their stomachs aren’t designed to digest corn, so their corn diet slowly kills them) and a large portion is made into ethanol and added to gasoline. The corn subsidies are so dumb that they require a minimum amount of the corn being grown to be turned into ethanol, which is extremely inefficient.
The Hank Green video someone linked above goes over some of the stuff I mentioned.
Yeah, exactly.
It’s also going to get more efficient, like Amazon’s data centers have been over time. By the way, where was this zeal over water usage from people back when AWS and other data centers started popping up all over the place?
The water issue will end up being more in how local water tables are affected, not overall consumption.
Yeah but when but gets 20% more efficient they build a second one. You still end up with a net increase.
This was a subtly sponsored video for which he got a lot of shit iirc. Didn’t he retract this?
I haven’t seen anything saying it was sponsored (other than by a news aggregator) and getting a lot of shit on the Internet isn’t any indication of being incorrect. The video is still up and no mention of anything like what you’re talking about.
Then i might be remembering wrong and if so I apologise.
Edit: ah it was the financial advice related to ai that got him a lot of criticism
I disagree. AI is causing huge increases in the demand for energy in general. Renewables are the most economic energy source and pretty much all new generation will be renewable since its cheaper. AI is just accelerating the transition process.
Good news is there is increased investment in nuclear energy for data centers, which will go a long way to combat this.
The U.S. government is shelling out a whopping $2.7 billion to three companies in an effort to strengthen domestic uranium enrichment, amid surging electricity demand from AI data centers.
God no, it will not. Aside from the discussion whether nuclear is really a good way to generate electricity (and I think it’s not): The demand is so insanely huge that it’s actually stacked: green plus coal plus gas plus oil plus nuclear is currently getting “assigned” to genai.
What do you think will replace fossil fuels as our baseload source? Because (to my understanding) renewables don’t have the output and stability required to fill that void.
Watt for watt, nuclear is one of the safest methods of generation and generates tons of energy with minimal waste (which already has methods of storage and reprocessing).
random, barely related thought
It always amuses me to point out that fossil fuel plants like coal are more radioactive than nuclear power plants. Because nuclear plants have strict regulations they have to follow, but coal plants concentrate radioactive materials into the ash as part of their normal operation, which can make it to the outside.
So fossil fuels better? Nuclear works great for France.
I’d rather just not build them
Is Coal and Gas is your preferred energy source? Because that’s what nuclear would be replacing.
Nuclear investments will probably just meet the increase in energy consumption due to new datacenters, not make an energy transition
The datacenters.
I’d rather it didn’t rain today, but it did and I still needed to bring an umbrella no matter how much I didn’t like the rain.
Nuclear energy is never good news.
Solar energy can boil water too. At much lower cost, 10x faster build times, and MUCH less waste … none that has to be guarded for centuries.
Never safe, never clean, never too cheap to meter. The exact opposite of the sales pitches. Rarely built without taxpayer dollars. Name the companies willing to insure one.
Crazy people still get downvoted in Lemmy for reminding everyone that Nuclear energy is the most expensive form of generating power while solar, wind, and water are the cheapest.
People just eat the “nuclear waste isn’t a problem actually ignore that in some places we’re already seeing it wasn’t stored safely aftet all” propaganda from the nuclear lobby right up.
And forget that just because nuclear plants are pretty damn safe when everything is done properly, people are notoriously great at not doing things properly, hence why 2 of the things have melted down so far (though i should say the same applies to hydro, except I only know of 1 disaster instead of 2, and the financial damage is less because water doesnt contaminate the ground for forever. Killed a lot of people though).
I’ll take it over fossil fuels still because co2 is also a huge problem, and having nuclear waste at all is a bigger problem than adding slightly more while we transition to full renewables.
Shew, what a headline. Keep in mind that AI is not just all LLM. I’m not sure how much more juice we can squeeze out of LLMs, but we’re just scratching the surface with other prediction models.
The ability to have smart cars that improve fuel efficiency by adjusting to traffic conditions may very well compensate for the increased electricity demand created by data centers.
New chemistry models may finally help us produce batteries that can meet the demands of a renewable energy grid.
We’re standing at a precipice. What we’re doing as a society is not working today - it’s not sustainable. And I’m not shy to say that I’m one of the few here on ActivityPub that think we may be able to leverage AI to dig us out of this hole we’ve started to dig ourselves in. But, in order to do that, we need clear heads, with clear goals and the incentives to encourage others to execute on them.
What we’re doing right now, just complaining about what’s not working isn’t going to save us.
The ability to have smart cars that improve fuel efficiency by adjusting to traffic conditions may very well compensate for the increased electricity demand created by data centers.
That just reads mindblowingly stupid for me after only one semester of the “automatic guidance of trains” subject with a few simple methods of numeric optimization. And I wasn’t studying very well, to say the least.
The rest of what you write feels as if you’d missed the whole “digital computer” thing and what it already allows us to do since 1970s and that is being done since 1970s.
