- cross-posted to:
- climate@slrpnk.net
- technology@lemmy.zip
- technology@beehaw.org
- cross-posted to:
- climate@slrpnk.net
- technology@lemmy.zip
- technology@beehaw.org
The leap in emissions is largely due to energy-guzzling data centers and supply chain emissions necessary to power artificial intelligence (AI) systems such as Google’s Gemini and OpenAI’s ChatGPT. The report estimated that in 2023, Google’s data centers alone account for up to 10% of global data center electricity consumption. Their data center electricity and water consumption both increased 17% between 2022 and 2023.
Google released 14.3 million metric tons of carbon dioxide just last year, 13% higher than the year before.
Climate scientists have shown concerns as Big Tech giants such as Google, Amazon and Microsoft continue to invest billons of dollars into AI.
Sadly it’s tricky to separate the two.
Say if hypothethically we have a data center that is not connected to the grid, and is entirely running on solar power and battery storage.
If the grid still generates (part of) its electricity need using fossil fuels, those same solar panels and batteries could instead have been used to (further) decarbonize the grid.
While using solar power is good, increasing the overall unnecessary electricity consumption is still not great.
But you can measure how much of the power of a grid is generated with fossil fuels at a particular place and time. For example, if they have more data centers where energy is cheap like from hydro or geothermal, then the carbon footprint will be less than if they were just using average power statistics.
That is assuming that those data centers are necessary. If the data center is doing something that is not really needed then it is in effect wasting power that could have been used for other purposes. (e.g. using surplus power to make steel or aluminium for instance)
While I do think that AI-tools can be increadibly useful, the current hype surrounding it very much looks like a bubble akin to the DotCom bubble to me. Companies left and right are jumping on the AI bandwagon for the sake of using the buzzword “AI” in their marketing speech.
I don’t consider that kind of use of datacenters to be necessary.
isn’t steel still primarily made through the coking process? Or is that transitioning to more efficient means, last i heard the barrier was efficient hydrogen generation.
You are correct, but that cooking process doesn’t have to be done with fossil fuels. Hydrogen (like you mentioned) is an alternative and you can create hydrogen using water and electricity.
In the NL we have a pretty polluting steel mill that is currently still coal fired. They are working on a transition plan where they adapt it to be gas fired instead, with the ability down the line to make it hydrogen fired when hydrogen production capacity is up to speed.
https://www.ad.nl/economie/tata-steel-stopt-met-kolen-binnen-tien-jaar-over-op-waterstof~a801e791/
(Translated headline: “Tata Steel stops with coal: Transition to hydrogen within ten years”)
there are almost certainly heuristics you can use, but these are going to be heuristics the size of the national US grid, with physics similar to how water flows through pipes. Except these pipes are dynamic and significantly less restrictive.
Plus source generation is very sparse, CCG gas plants for example generally only run when peaking, and solar only works during the day, generally, and nuclear power runs 24/7 around the clock, so it’s not quite trivial to calculate. More than likely what the heuristic they’re using here is that they consume 43% more power as a corpo, and thusly, produce 43% more CO2.