By Jo Nova
The winds of change are howling through electricity grids
Since 2022, AI -related firms have stormed the S&P 500 market — growing by $12 trillion dollars.
The IEA just posted a whole report dedicated to AI. The demand from data-centers is so large in some places it is already rivaling the kind of monster consumption we are used to seeing from aluminum smelters. There are six states in the United States where data centers already consume over 10% of the electricity supply. In Ireland, data centers swallow about 20% of the electricity.
Currently, a normal data center consumes the same amount of electricity as 100,000 houses. But the new gargantuan data centers under construction will consume 20 times as much — equivalent to adding 2 million homes to the grid.
Data centers of the world are not spread evenly. In Virginia, the largest conglomeration of industrial data, their demand for power pulls in a quarter of the state’s electricity.
Australia is being left behind, because we won’t build coal plants in case we offend the UN, and we banned nuclear power as a fashion statement in 1998. The AI global race is on, but digital machines need reliable cheap electricity and lots of it.
Also not looking sparkling on the graph above — New Zealand, Norway, Sweden and Canada.
Sometime between now and 2030 (which is like ‘next week’) the world has to build a new network the size of Japan’s national grid — the fourth largest economy in the world.
Data centres accounted for around 1.5% of the world’s electricity consumption in 2024, or 415 terawatt-hours (TWh). The United States accounted for the largest share of global data centre electricity consumption in 2024 (45%), followed by China (25%) and Europe (15%). Globally, data centre electricity consumption has grown by around 12% per year since 2017, more than four times faster than the rate of total electricity consumption. AI-focused data centres can draw as much electricity as power-intensive factories such as aluminium smelters, but they are much more geographically concentrated. Nearly half of data centre capacity in the United States is in five regional clusters.
Data centre electricity consumption is set to more than double to around 945 TWh by 2030. This is slightly more than Japan’s total electricity consumption today.
There is no single factor driving up electricity demand more than AI at the moment:
In the United States, data centres account for nearly half of electricity demand growth between now and 2030. By the end of the decade, the country is set to consume more electricity for data centres than for the production of aluminium, steel, cement, chemicals and all other energy-intensive goods combined.
The IEA report, which was always a sop for “renewables” — suddenly isn’t that concerned about carbon emissions. This is an interesting shift. Almost like there is a sea-change in priorities of whoever it is that controls agencies like the IEA. Now it’s making excuses for industry…
“Concerns that AI could accelerate climate change appear overstated, as do expectations that AI alone will address the issue”
The widespread adoption of existing AI applications could lead to emissions reductions that are far larger than emissions from data centres – but also far smaller than what is needed to address climate change.
Just like private jets for billionaires, big new AI computers will save us from carbon emissions (maybe)
AI will find cost savings all over the place, but then again, if everyone has their own automated self-driving car, they won’t need to catch the bus will they? Oh the dilemma…?
AI applications in transport can improve efficiency and save costs, but they could also increase demand for personal mobility. AI applications are being used to manage traffic, optimise routes, predict maintenance needs and develop autonomous vehicles. The widespread adoption of AI applications across the transport sector could lead to energy savings equivalent to the energy used by 120 million cars. While autonomous vehicles operate more efficiently than conventional ones, they might also attract people away from public transport as costs fall and availability increases, leading to rebound effects.
In buildings, there is significant potential for AI-led optimisations to make heating and cooling systems more efficient and electricity use in buildings more flexible
Even the IEA admits we need affordable and reliable power
Countries with a record of reliable and affordable power will be best placed to unlock data centre growth, localise the computing power that is critical to homegrown AI development, and spur the IT industry more generally.
This graph with a microscopic font, compares how extensive those blackouts are around the world, with the First-World looking good (so far):
Beside that graph they have a very strange graph with microscopic fonts (which I expanded below) which has one solid square all by itself. This turns out to be the other emerging markets and developing economies they didn’t mention in the last graph.
The High Outage Country class of 2025 has more like 700 hours of average system interruption a year. It’s “off the charts” bad.
Tacitly, it suggests that the AI revolution won’t be moving to South Africa or Cuba.
Is there any industrial complex in the universe that works better on a part time random basis than it does on a predictable, reliable schedule?
History will show that countries with energy to spare will take over the world, and maybe the solar system.
REFERENCES
I don’t understand how the Australian Government wants Australia to become an “AI and data centre superpower” as well as a “renewables superpower”.
You can’t have both. The requirement of massive amounts of cheap reliable energy for AI and data centres make the two mutually exclusive.
Pick one.
190
>Pick one
Plus recharging a national fleet of net zero electricity powered buses trucks and motor cars.
110
If “government’, especially this one, is involved, the result will be “None of the above”.
As intended
70
The answer is obvious. The AI factories will just load shed when the grid is running down.
So if you are planning a big day on the AI band wagon, then you need to get in around lunchtime. On cloudy days, AI is going to be taking a nanny nap or two.
40
The Australian Government has a dilemma.
As the country descends into a totalitarian state under the Labor mandate, the Government will need increasing data centre and AI capacity so that every aspects of the lives of non-Elites can be fully traced, tracked and analysed including every thought you have on social media and every cent you spend and with whom as well as facial recognition tracking of you everywhere you go.
All required for the eventual establishment of a Chicomm-style social credit scheme.
To run the vast data processing requirements for this, massive Government AI and data centres will have to be implemented which will need vast amounts of energy which the country no longer has.
So they will have to either build coal, gas or nuclear power stations to power it all or do without. The latter might be preferable.
130
We are becoming East Germany, witness the interrogation from Which Bank when you want to withdraw some of your OWN money.
80
With possibly another 6 years of labor, unreliables get the nod so data centres are off the agenda.
It’s going to be like NINE years of Whitlam or Krudd.
60
The “Data Centres” will be on the “essential grid”, along with the ADF, sundry police services and “enforcement agencies”.There will NOT be enough “juice’ or “infrastructure to hold up the Hospitals and other health srvices.
A bit more “right-sizing”.
If it all goes dark for an extended period, then the STASI will be reduced to pencils and manual typewriters, (until the paper runs out), a la “Brazil”, (the movie); “Buttle / Tuttle”)
20
” You will have no power and you will be happy”.
00
“It’s going to be like NINE years of Whitlam or Krudd.”
Well it’s a blessing Dutton was axed by his constituents as he, aided by a squabbling ex-Coalition, would have wrecked the country by a week on Thursday.
03
Once the realization of this inconvenient truth penetrates the skulls of the dimwits in charge of the economy, surely then it must mean the death knell for the inefficient ruinables.
20
We can only hope.
Black-Out Bowen is doing his job.
Stuff up EVERY portfolio he has ever “run”.
70
What is it about Bowen who is incompetent in everything he does but still allows him to be put into powerful political positions which are also ruinous for the country?
Oh, I think I just answered my own question.
30
Not too complex a question really. He is an example of a politician who is easily manipulated by the bureaucracy.
And isn’t that all of them?
10
Their skulls are too thick for anything to penetrate.
00
Nett Zero is too STUPID to be STUPID. Bottom line is the sheeple of Australia voted for it. Don’t ever think you will VOTE these idiots out.
20
66% of Australians DIDN’T vote for this.
We can blame the elites in the capital cities and the dole bludgers.
Mustn’t forget the AEC in all of this either.
50
Yes Net Zero was introduced into Australia by Prime Minister Scott Morrison and Minister for Industry, Energy and Emissions Reduction Angus Taylor in October 2021 released Australia’s Long Term Emissions Reduction Plan (the Plan), to deliver net zero emissions by 2050. No wonder it is too stupid to be stupid
02
I wrote an article on data centres.
https://www.siliconchip.com.au/Issue/2025/January/Data+Centres+%2526+Cloud+Computing
Not only do data centres and AI use vast amounts of power they can also use vast amounts of water.
I wrote:
50
“About 2 /3 of the water used by Amazon data centres evaporates; the rest
is used for irrigation.”
What are they irrigating, lawns and gardens ??
10
According to the reference I used, feed crops and pastures. The water is discharged into irrigation canals.
20
Irrigating?
Dope, opium and poppy farms, more likely.
10
Explains some of the dopes in govt.
00
Harvesting all that waste heat would be an opportunity to generate power with an Organic Rankine Cycle generator, and it would reduce the cooling water losses.
10
Not hot enough to make any significant energy. Remember Carnot, energy out is proportional to the temperature difference of the source and the sink. That’s why the power stations use very hot steam.
Hot water, on a bad day, may only be 30C above ambient. Compare that to steam, which will have a delta T of around 300, even on a hot day.
20
What’s the probability that the AI label is just a Trojan horse for CBDCs? Blockchain tech is not very efficient for storing data but in theory it does give an immutable record of txs. It also gives a great way of controlling not only YOUR money but also your whole identity. Think East Germany. The Stasi (sp?) had massive underground data stores – this is what they do!
40
Stasi had something like 8 kilometers of racking to store paper spy data, spying on their own people.
20
About one in six East Germans were informers for the Stasi.
We see the emergence of that here.
Here is a poster on Victoriastan public transport. SEE LINK.
No one is saying that the described activity is nice or appropriate but you are being asked to report something to police with no knowledge of what a person might be looking at. They might be staring into space, or something behind the person, for example.
https://images.app.goo.gl/tiAc5yChH1z77X6U7
We truly are living in an Orwellian society ALREADY.
10
Plus see my post #2.
10
I am sick of Climate Change. Utterly sick of it. Humans have not and cannot change CO2 levels. End of argument.
But it would be nice to talk about real developments like AI and its uses. Only the regressive socialist states are still talking Climate Change.
In Victoria we should be looking at every possible way to generate energy. Fracking, coal, gas. I do not include solar and wind because they are completely unreliable and inadequate, non commandable. The only certainty is that there will be zero solar at night.
And our politicans lie to us. Routinely. Massive carbon charges buried in every bill. And they pretend to hand back money to assist the poor who didn’t have a problem until electricity became unaffordable.
In the Liberal party, it is split along Susan Ley’s passion, Climate Change. No wonder Peter Dutton could not attack. And she would rather wreck the joint than give up on Climate Change. The Nationals are better off not tied to a losing position.
And in the US, I have not read of Climate Change this year! Gas prices are lower. Egg prices are lower. 300,000 uninvited Venezuelans may be going home. Tren De Aragua gang members are being put in high security jails in El Salvador, despite the protestations of the Democrats who see these violent tatooed monsters as farm assistants or food delivery people. Which is as ignorant as it is tragic.
Trump’s approval rating is an amazing 55% just after 100 days of promises made, promises kept. Our weasel self serving leaders, planning their comfortable early retirements on other people’s money say one thing and do another. Especially Howard, Turnbull, Morrison,Dutton. There is no Conservative party. And in the UK a woman is in jail for a comment on social media and Keir Starmer is undoing BREXIT, regardless of what the people want.
This is the new lying political animal. Quoting Adam Bandt, “tell them what they want to hear and when we get power, we do what we like”. The Tories and Australia’s Liberals will soon be history too.
As for AI, it is as exciting as it is frightening. It is the ultimate software, a program which programs itself, adapting to the situation. But will it become another Adam Bandt?
90
I don’t think so.
See my post #7 in the Wednesday thread.
https://joannenova.com.au/2025/05/wednesday-107/#comment-2849703
10
Littleproud then is just as bad as the rest. We may as well have AI politicians.
10
Just a minor bit of mind reading Tdef.
I think what you are sick of is not climate change but you (and I) are sick of the climate scam.
30
There is no Climate Change. It was degrees colder in 1870. But that happens. Climate change is Siberia to Spain. Frankly warmer is always better. Yes, it’s a scam invented by Al Gore and the UN in 1988. And never proven.
10
Yes, I was just a little surprised to see you write that you were sick of climate change.
Keep up the good work. In particular your recent insights on varying radiation received from the sun are most informative.
10
The growth in AI processing centers will be limited to the growth in reliable generating capacity. Projections of AI growth do not seem to realize this.
60
Australia is going to need a bigger international fibre connection to where the data centres are (will be).
00
I call BS!
” AI applications are being used to manage traffic, optimise routes, predict maintenance needs and develop autonomous vehicles…In buildings, there is significant potential for AI-led optimisations to make heating and cooling systems more efficient and electricity use in buildings more flexible”
I haven’t seen anything AI has invented, all it seems to be is a just a slightly faster computer doing the same stuff any computer could do, which is the same stuff any person could do, but faster. Manage traffic? Pointsmen do it. Optimise routes? Dispatchers do it. Predict maintenance? Mechanics do it..
Heating and cooling? “HAL, turn up the heating, I’m cold” …”Dave, you’re not cold, the heating is optimised and your sensations of coldness are illusory. The shortage of electricity caused by AI centres have meant there is not enough left for heating more than optimised..”
All this is crap, the only reason for AI is to sift through large amounts of data and sort it, perfect for spying on millions of people, but really another complete waste of human capital, like ruinables. The only advantage I see is that AI companies will have to build their own nuclear power stations and may sell the over-supply to the local community. Work for IBM, get electricity in your IBM-owned apartment block on campus.
20
I’m suspecting that once these AI centers are up and operating the workforce to maintain them are minimal. So, the electricity supplied is way out of kilter to people employed. That the people are only employed in the building phase. In the past, industries using vast amounts of electricity tended to employ a lot of people. Even at a TESLA gigafactory a significant amount of people are employed. Strange mix. Economic benefits to a regions economy where these AI centers are built, would appear likely to be quite minimal.
20
According to Goolag AI:
Where’s this electricity going to come from? Certainly not the sun, wind or Unicorn or Chrissy Bowen flatulence.
And the water?
And Goolag AI also says:
Note to Leftoids: I am very proud to say that that Goolag AI search produced lots of wonderful greenhouse (sic) gases.
Goolag AI is shy about telling me it’s “greenhouse” (sic) gas production but for a similar AI, ChatGPT:
Per AI query 68g of CO2 is produced and a simple calculation shows 68g of CO2 gas occupies 34.5 litres.
Isn’t that wonderful?
Beautiful, life-giving CO2 (what Leftoids call carbon”.
40
Just do one AI query and pump put 34.5 litres of CO2.
Just think how much you could produce sitting at your keyboard all day doing random searches.
00
So the ideal place to build these AI centres would be close to electricity generation that has a good water supply. Gee, I wonder where there are things that already supply lots of dependable cheap electricity and utilise a dependable local water supply?
00
My crystal ball has been a bit cloudy lately, but I wonder whether the cognitive dissonance induced by this issue might just prompt some learning within the bureaucracy. Highly unlikely I know so I present…
My guess is that the world will end up with multilevel electrical grid. High level users such as data centres will have power supplies collocated or nearly so. Low level users such as suburban homes will be stuck with solar panels and intermittent energy requiring vast unprofitable grids totally disconnected from the data centres.
Entire continents such as Australia will simply dangle at the end of overseas fibre optic connections. Cloudless daytime only of course.
Oh well. As a nation we’ve still got sheep, wheat and stuff which can be dug up and exported. Don’t we?
10
Except the Government is at war against the farmers.
And Australia doesn’t have a monopoly on useful rocks in the ground.
The biggest consumers like the Chicomms are and have established colonies in Africa to obtain the required minerals and slave labour. Other large consumers like the United States are either self-sufficient or have alternative sources of supply apart from Australia.
I doubt whether the aluminium industry will last much longer. It was run on cheap coal power which is being shut down. Despite heavy taxpayer subsidies to keep them here I think they will move to more favourable places to do business.
The Government plan for aluminium smelting is to run it on “renewables”, an absurd proposition even with massive taxpayer subsidies.
00
Hmm! Well worth another look.
I’ve been using the U.S. EIA (Energy Information Administration) data since I started out back in 2008. It’s a mammoth site with so much information, and it’s probably on the scale of an AI Centre as well.
I haven’t ‘been there’ for a while. It ‘morphs’ every so often, so the data you go looking for now gets ‘hidden in plain sight’, so one needs to go looking, but in reality, there’s nothing about power generation you can’t find. All it requires is correct wording in its search engine. So that’s why it has become more time consuming when using the site, and then knowing what to ‘click on’ to find what you need.
They do a two Monthly Update of all power generation, and you can spend huge time just looking at that data of itself.
However, one of the entries there is the total power consumed (hence, generated) in the U.S. and that’s a (seemingly) four thousand plus Terawatthours. (TWH)
It’s actually been around that figure since I started back in 2008. Some years it goes (fractionally) higher and some years (fractionally) lower.
Now, having not gone to the site for a while, I looked specifically for that total, with Joanne’s latest post here in mind ….. the rise in power usage by those AI centres.
So just in the last two years alone, power usage seems to have spiked, keeping in mind what I said above ….. some years up a bit and some down.
Power generation rose in 2024 by around 2.7%, and then in 2025 by around that same percentage, 2.7%. Now here they use a rolling base for comparison, so it can be accurate till the end of the most recent recording period, and with the next one not due for a coupla weeks, (end April) this rolling basis is February to February, if you can understand what I mean here. (and here, see how so much text is needed to provide context)
OK, so, a 2.7% rise may not sound like much.
That 2.7% increase in each of these last two years is actually in the vicinity of 120TWH of power generation. For comparison, just the AEMO coverage here in Oz generates around 214TWH this last year, (2024) so just that rise is U.S. power generation is half of Australia’s total.
So, that recent year on year 2.7% rise is indeed quite a large amount, considering it hovered around that 4,000 mark for so long. Here, they only keep data for the most recent ten years, so back in 2008 when I first checked the site, it went back to 1998, so in effect that 4000TW dates now at 25 years, not including this rise across the most recent two (rolling) years.
Now, because power generation hovered around that earlier mark, I can only guess the extra power need coming down to the need for extra power needed for AI, as that has only ‘boomed’ in the last two or so years. So, hovering around 4,000TWH since 1998, the most recent rolling 12 Month number is 4344TWH
An interesting exercise all in all.
Huh! Now I’m probably going to spend hours and days looking at the site.
Incidentally, when I started in 2008, I also found, after looking for so long, similar data coming out of China, and there, (in China) power generation was way lower than in the U.S. It took until 2012/13 for power generation in China to rise to the same level as in the U.S. At the end of 2022, power generation in China was ….. DOUBLE that of the U.S. That data now coming out of China is now is now commensurately more difficult to locate. A few years back, they changed their original site from what seemed a ‘rudimentary’ basis to something more akin to the U.S. EIA site, only way way way more huge, and now that data is almost impossible to find. At the start, during the changeover, that data was still relatively easy to find, now, not so much.
Also incidentally, Australia does something similar, only a yearly report, a hundred and more pages long, and that gets released around August or so with data from the previous year, so almost a year out of date. I use the OpenNEM site for (somewhat rudimentary) data and the better Aneroid site, which is easier to navigate, once you know what to look for, Both of these sites are ‘real time’, and depending on what I’m looking for, I use both sites.
Tony.
00