Data Centres Will Use Twice as Much Energy By 2030 (nature.com)
- Reference: 0176996487
- News link: https://hardware.slashdot.org/story/25/04/10/2019233/data-centres-will-use-twice-as-much-energy-by-2030
- Source link: https://www.nature.com/articles/d41586-025-01113-z
> The report covers the current energy footprint for data centres and forecasts their future needs, which could help governments, companies, and local communities to plan infrastructure and AI deployment. IEA's models project that data centres will use 945 terawatt-hours (TWh) in 2030, roughly equivalent to the current annual electricity consumption of Japan. By comparison, data centres consumed 415 TWh in 2024, roughly 1.5% of the world's total electricity consumption.
>
> The projections largely focus on data centres, which also run computing tasks other than AI. Although the agency estimated the proportion of servers in data centres devoted to AI. They found that servers for AI accounted for 24% of server electricity demand and 15% of total data centre energy demand in 2024.
[1] https://www.nature.com/articles/d41586-025-01113-z
Obligatory XKCD (Score:3)
Obligatory XKCD: [1]Extrapolating [xkcd.com]
This kind of sounds like a report designed to push companies into building more power plants to drive supply up and prices down.
Data centers are likely to do everything they can to keep power consumption down, because power costs money. Doubling the power needs, even if supply keeps up, means doubling the cost, and I'm not sure that's really sustainable in terms of revenue.
That said, scarcity breeds innovation. Limitations to power availability are what drive companies to find ways to be more energy-efficient. For on-device models, obviously the power availability isn't going to improve significantly, and CPUs aren't going to get massively more efficient per watt, so you're going to have those constraints no matter what, but for data-center-side AI work, you need the threat of running out of power availability, because that's what forces the folks working on the models to find ways to prune them to be faster and more efficient, whereas in a hypothetical world with infinite free power and CPU, future power consumption would be unbounded.
Mind you, hardware costs also help bound things and serve as a forcing function to make things more efficient, but those are one-time capital costs, whereas power costs are ongoing expenses, so when viewed from a long-term financial planning perspective, hardware costs are noise, and power costs are a much stronger forcing function.
[1] https://xkcd.com/605/
Data centers will pollute twice as much by 2030 (Score:2)
It's pretty clear that there is no intention to be limited in any way by what kind of energy they are going to use and that means the energy use is going to be relying on polluting sources. Corporations don't care about who they harm and people don't seem to give a damn either... until it's them. This world is fucked.
Re: (Score:3)
> It's pretty clear that there is no intention to be limited in any way by what kind of energy they are going to use and that means the energy use is going to be relying on polluting sources. Corporations don't care about who they harm and people don't seem to give a damn either... until it's them. This world is fucked.
Polluting twice as much is probably best-case. Realistically, building new power plants takes longer than starting up plants that already exist but are offline, and most of the plants that are offline because of cost reasons are likely to be based on fossil fuels. So it could very well be that using 2x the power results in 200,000x as much pollution if most of their existing power comes from solar and wind and most of the extra power comes from bringing coal and natural gas plants back online.
And with 145
To replace you. (Score:2)
That's the part we're not supposed to talk about. The reason that all that electricity is going to be used is to replace workers. That's the drive coming.
I don't know if you believe it or not but the CEOs absolutely believe that they're going to be replacing a large percentage of their employees. And you know what they're right.
All this talk about AI whether it's real or not has got CEOs going top to bottom through their entire organization looking for things they can automate. Looking for places th
Better keep building more power plants (Score:1)
Or we will all be doomscrolling in the dark.
Re: (Score:3)
How about just using a lot less AI? I think whatever improvements, if any, that AI has made to mankind's lot are vastly overshadowed by their environmental damage, greenhouse gas emissions, and job market disruption.
When you factor in the degree to which AI has already become a devastatingly effective propaganda tool, AI just seems like a really, really bad idea. "We did it because it looked cool and because we could" seems is a really shitty justification.
Re: (Score:2)
> How about just using a lot less AI? I think whatever improvements, if any, that AI has made to mankind's lot are vastly overshadowed by their environmental damage, greenhouse gas emissions, and job market disruption.
> When you factor in the degree to which AI has already become a devastatingly effective propaganda tool, AI just seems like a really, really bad idea. "We did it because it looked cool and because we could" seems is a really shitty justification.
AI at a reasonable pace could be a net positive. Well reasoned research into the possibilities leading to incremental improvements. But somebody figured out a way to hype it into a profit center, and once that happened, it became an arms race. Now, any attempt at getting introspection from the AI prophets is met with shaky-voiced outrage, "But, if we don't, they will," and completely unreasonable expectations that the only "solution" to today's AI's issues is to throw more power, more hardware, and more ove
Re: (Score:2)
Although I personally think the current crop of "AI" programs is of limited value, others don't. Good luck putting the genie back in the bottle.
I can imagine someone making remarks similar to yours when the printing press was invented. Your falling sky is someone else's brand new world of tomorrow. Humanity will survive LLMs and image generators. If this is what finishes off our ancestors wasted their time surviving plagues, starvation, and other non first-world problems.