The latest figures from Ireland's Central Statistics Office (CSO) show that giant server farms now account for nearly a quarter of the country's metered electricity consumption.
Their share rose to 23 percent in 2025 after passing 20 percent in 2023 and 14 percent in 2021 – up from just 5 percent way back in 2015.
Luckily this will all be offset by the pot of gold at the end of the AI rainbow.
The optimal number of data centers is just enough so that my personal use is covered. No more. No less. Screw other people’s needs and demands because I know better.
Isn’t this exactly the type of thing that a market is designed to discover, which we’re seeing unfold right now?
Yes, maybe they’re building too many right now, who knows. It’s very likely that demand for computation will go up in the future, and EVs are also going to be consuming much more electricity, so all governments better start preparing for more (clean) electrical supply.
I doubt this. It's probably quite high but there is a limit to how much compute you can genuinely use. Just like there's only so much water you're gonna use even if you greatly enjoy Californian almonds.
However, there's probably no limit to energy/electricity we can usefully allocate. And therefore yes, we should in fact provide as much as possible, Dyson spheres and all.
I'd be more OK with it if prices weren't subsidized so much and people actually had to pay to ask opus how to pee, then maybe we would realize we don't need beefier models for everything
There isn't much evidence at all that inference is "subsidised" (and by whom?)
Training is quite expensive and it does look likely that the American providers have been doing that at a loss.
In any case, you can go buy a MacBook Pro M5 48GB or an AMD R9700 and run Qwen 3.6 35B-A3B (a very capable model) and the only "subsidy" is you plugging it in, and 140W is not exactly a huge amount of power (roughly 50¢ per day if you run it 24/7 at 100% load, which it is very unlikely you will).
> There isn't much evidence at all that inference is "subsidised" (and by whom?)
None of the big providers are profitable. It’s subsidised by overly enthusiastic VCs.
> In any case, you can go buy a MacBook Pro M5 48GB or an AMD R9700 and run Qwen 3.6 35B-A3B (a very capable model) and the only "subsidy" is you plugging it in
Right, people could. But they won’t, because that’s a bloody expensive computer and they don’t need that to ask ChatGPT. That war is lost already.
Subscription to the big players’ services would need to increase massively for that to happen. And the computational cost is only part of the problem; these models also eat a lot of storage and RAM, which is not exactly getting cheaper.
The typical "free" AI or cheap tiers are equivalent in power to a Qwen 3.6 model (which is also much cheaper to run in a hyperscaled situation than on my laptop or PC; a single H200 can host thousands of sessions of a typical chatbot user). There is no evidence the Chinese AI providers are being subsidised either.
You can look at API pricing on a service like OpenRouter (which isn't subsidised) and see pretty readily that it's not expensive to provide lower-tier inference. Higher-tier inference like GPT-5.6-Sol or Opus is expensive - $100 a month plan for realistic usage, and only up from there.
I agree with using smaller models, it's just that the majority of people I know feel like they need the biggest, beefier, behemoth model possible (with the longest thought setting) and consume much more than necessary when a flash or smaller model would be OK. I would also like to be able to use a smaller model, but given ram prices I would have to sell a kidney to buy ram now
Most people who use a free or $20 a month plan are already using smaller models, and the mainstream chatbot services will route requests to a smaller model often without really telling the end user.
You can run Qwen-3.6 on a 32GB card which will set you back about $1400, or $400 of just RAM if you want to run it on a CPU.
There is no evidence Chinese providers are getting such subsidised, and in fact apparently Jinping (who presumably knows what the CPP is doing) was surprised when DeepSeek and Qwen generated so much buzz. In China, AI inference is just viewed as another basic utility, much like an e-mail provider or a mobile phone network.
I'm not aware of how Musk/SpaceX are "raiding retirement funds"?
I get what you're saying, but I'm guessing that people asking how to pee is a drop in the bucket compared to the agentic loops being called to rename some variables across a project.
GLM 5.2 isn't quite modern Opus tier, as seen in this comparison where Opus 4.5 scores 4/5 on some coding tasks where GLM 5.2 scores 0/5: https://www.tryai.dev/blog/gpt-5.6-build-off-12-models But yes, GLM 5.2 is cheap.
But the real standout on price is DeepSeek V4 Flash, which competes, more or less, with models in between Sonnet and Haiku. From third-party providers, it costs around $0.09/M, $0.18/M out, compared to $3M/in, $15M/out for Sonnet and $1M/in, $5M/out for Haiku. To get the price of DSv4 (Flash and Pro) so low, DeepSeek did a lot of innovative optimization work that will likely show up in other open weight models in the future.
For a context: France relies heavily on automotive transport, plus it's a home to enormous agricultural sector, tractors are literally everywhere in the country during the summer. To a certain degree, structurally it resembles USA a lot.
However they also quite famously rely on a majority of nuclear power for their electric grid. Great for France, but that makes them an already-low carbon emitter compared to many others and an ungenerous comparison.
legality of the datacenters aside, I wonder why countries don't at least demand that they're totally carbon neutral or free. it's possible today. it's not like it's sci-fi.
USA is partly a petrostate so regulatory capture is a problem. To mitigate the climate catastrophe it would be important to ramp down fossil fuel production in a big hurry, but it's not in sight.
In Europe this is covered by the emission trading system (EU ETS) and datacenters have to share the same shrinking emissions quota as other industries.
I think most ways of obtaining carbon neutrality are a little bit BS, that's why.
An alternative is what Google is theoretically aiming for: being carbon-free. But they've already started using language describing it as a moonshot or idealistic goal so seems likely they'll abandon that
There are laws in the EU that will save the planet, like drinking from soggy paper straws instead of plastic and requiring caps stay attached in plastic bottles.
And just to make sure, at least in Poland they now charge you $0.10 if you buy anything plastic until you bring it back to the grocery store empty.
But... Datacenters don't burn anything, right? Powerplants do and we try to switch all the transport and heating and whatever to be electric.
So the answer is to build the damb nuclear power and a lot of it and price CO2 emissions at the actual cost of sucking the thing back out if the atmosphere
They do have a growing amount of Scope 1 emissions (emissions from their on site sources) which originally was primarily on site diesel but due to grid interconnect delays have been growing number of on site gas turbines.
This certainly wouldn’t be necessary with adequate generation and transmission capacity.
This is true, but I'm pretty firmly of the opinion that these data centres shouldn't be built, or at least allowed to operate until/unless they can be powered cleanly and without cornering the market and driving out existing consumers of power.
If they're so keen to build that they're willing to fund power generation (e.g. on site gas generators) then it should be clean/renewable (solar, wind, small modular reactors, full scale nuclear plants, whatever).
Degrowth is bad but so is ignoring the planet, the environment, and people's health to get ahead faster in business.
The need for memes knows no bounds. In short order the majority of power usage worldwide will be for compute and newer generations will wonder how it took so long.
Which is why things like nuclear power plants, grid upgrades, hydroelectric projects, and intelligently placed wind/solar (instead of placing it due to subdisies or political concerns) should have been done a long time ago.
> price CO2 emissions at the actual cost of sucking the thing back out if the atmosphere
This is the only relevant bit actually. The rest will follow from there. And in principle, at least in Europe, we already have some mechanisms to do this. We'd "just" have to up the prices.
BUT of course with the right wing on the advance, and with them having identified basic physics (i.e. climate change) as a culture war terrain, this keeps being watered down... Oh well... This is why we can't have nice things... like a future...
If enough climate systems collapse, lots of existing farms will no longer be viable. That means famine and migration, which means war, which means lots of death. I don't know anyone who thinks we'll see extinction (outside of possible "hothouse earth" scenarios, where we become a second Venus) but societal collapse is definitely on the table. Saying this is "not scientific" would just be you not understanding the science.
"Climate conspiracy"? Like you, mean, the conspiracy of climate scientists to publish facts to the best of their understanding?
I don't know what exact strawman you're arguing against, although I'm sure you can always find some idiots saying something like what you say. But scientific consensus has long been that it will lead to increasingly extreme weather and mass extinction, which we seem to be on track for. Of course, we can't know for sure what the consequences are untill we do the experiment, which in this case means potentially destroying large sections of the biosphere and living with increasingly destructive weather patterns. Surely that risk is worth at least legislating that hyperscalers need to spend some of their billions on solar panels?
It’s anxiety about tipping points more than conspiracy theories. If you look at the figures in the various scenarios without even factoring them in: hundreds of millions to billions face food and water insecurity. You may be insulated enough from the direct effects but what about what they trigger? Already even our wealthy societies are being strained by rising food prices, extreme weather, and dysfunctional migration.
We don’t know exactly what the tipping points would lead to but we do know it would be some degree of a more severe and abrupt decline across the board.
Most conspiracy theories think food grows in supermarkets and because they have a job which pays them a good global income they can just pay a bit more.
> The left wing version of climate conspiracy is that climate change will end humanity itself.
That is unrealistic, but also not what a conspiracy is. It’s also a red herring, as nobody serious is claiming that. They are talking about civilisation changes, which we are already seeing.
At some point you have to engage with the arguments besides shouting "no, it’s you".
The solution is simple: require datacenters to overprovision solar panels and grid-scale batteries for themselves, and use that capacity to strengthen the grid and transition off of hydrocarbons.
You can’t get a grid tie for those panels in most of the US right now. The process for connecting to the grid is done serially, and requires a large study for any new generation.
I imagine it'd be a lot easier to get a giant datacenter through the queue and connected if said datacenter also generated more than it's own needs with solar and grid scale battery. You'd essentially be asking the grid to act as a backup and sink for spare generation. Avoiding a need for central generation capacity buildout.
Sure you can. The datacentre builders just don't want the (fairly modest) extra expense to do so properly. Obviously some preparatory work is required before dumping a lot of extra capacity into the grid.
My state allowed the power utilities to charge a modest fee ($10,000 to $100,000, depending on project size) before a yet-to-be-built data centre could demand a large amount of electricity. The amount of planned data centres went down by an order of magnitude. The truth is that most of the data centre builders (not all) do not want to be responsible citizens and are simply extracting value and wealth from other people, including from the power utilities and grid operators.
In related current news:
Irish datacenters now guzzle 23% of the country's electricity https://www.theregister.com/on-prem/2026/07/11/irish-datacen...
Luckily this will all be offset by the pot of gold at the end of the AI rainbow.23 percent is a bit fucked actually
It's really fucked in the country with the most expensive kWh in all of Europe...
Tax breaks matter more than electricity prices.
One more datacenter bro one more datacenter bro please I swear bro one more datacenter and everything will be OK please bro
The optimal number of data centers is just enough so that my personal use is covered. No more. No less. Screw other people’s needs and demands because I know better.
Pluralism? What’s that?
So infinite? There’s no amount of compute that would satisfy everyone’s needs and demands.
Isn’t this exactly the type of thing that a market is designed to discover, which we’re seeing unfold right now?
Yes, maybe they’re building too many right now, who knows. It’s very likely that demand for computation will go up in the future, and EVs are also going to be consuming much more electricity, so all governments better start preparing for more (clean) electrical supply.
Building data centers sounds like a good bet to make some money in the next decade or so.
That’s not the same as conceding that “everyone should get as much compute as they demand”.
That’s all fine as long as the market covers all externalities. It doesn’t
I doubt this. It's probably quite high but there is a limit to how much compute you can genuinely use. Just like there's only so much water you're gonna use even if you greatly enjoy Californian almonds.
However, there's probably no limit to energy/electricity we can usefully allocate. And therefore yes, we should in fact provide as much as possible, Dyson spheres and all.
How do you figure that? And if it exceeds what we can generate, a very big number is as unsatisfiable as infinity
Dyson spheres are science fiction. Science fiction is fiction.
Work without effort and thinking machines were also once fiction.
I know that market, and people for that matter don't care, but the environmentalist in me questions the word "needs" in the context of using AI.
I'd be more OK with it if prices weren't subsidized so much and people actually had to pay to ask opus how to pee, then maybe we would realize we don't need beefier models for everything
There isn't much evidence at all that inference is "subsidised" (and by whom?)
Training is quite expensive and it does look likely that the American providers have been doing that at a loss.
In any case, you can go buy a MacBook Pro M5 48GB or an AMD R9700 and run Qwen 3.6 35B-A3B (a very capable model) and the only "subsidy" is you plugging it in, and 140W is not exactly a huge amount of power (roughly 50¢ per day if you run it 24/7 at 100% load, which it is very unlikely you will).
> There isn't much evidence at all that inference is "subsidised" (and by whom?)
None of the big providers are profitable. It’s subsidised by overly enthusiastic VCs.
> In any case, you can go buy a MacBook Pro M5 48GB or an AMD R9700 and run Qwen 3.6 35B-A3B (a very capable model) and the only "subsidy" is you plugging it in
Right, people could. But they won’t, because that’s a bloody expensive computer and they don’t need that to ask ChatGPT. That war is lost already.
Subscription to the big players’ services would need to increase massively for that to happen. And the computational cost is only part of the problem; these models also eat a lot of storage and RAM, which is not exactly getting cheaper.
The typical "free" AI or cheap tiers are equivalent in power to a Qwen 3.6 model (which is also much cheaper to run in a hyperscaled situation than on my laptop or PC; a single H200 can host thousands of sessions of a typical chatbot user). There is no evidence the Chinese AI providers are being subsidised either.
You can look at API pricing on a service like OpenRouter (which isn't subsidised) and see pretty readily that it's not expensive to provide lower-tier inference. Higher-tier inference like GPT-5.6-Sol or Opus is expensive - $100 a month plan for realistic usage, and only up from there.
I agree with using smaller models, it's just that the majority of people I know feel like they need the biggest, beefier, behemoth model possible (with the longest thought setting) and consume much more than necessary when a flash or smaller model would be OK. I would also like to be able to use a smaller model, but given ram prices I would have to sell a kidney to buy ram now
Most people who use a free or $20 a month plan are already using smaller models, and the mainstream chatbot services will route requests to a smaller model often without really telling the end user.
You can run Qwen-3.6 on a 32GB card which will set you back about $1400, or $400 of just RAM if you want to run it on a CPU.
> There isn't much evidence at all that inference is "subsidised" (and by whom?)
Well... why else would the major providers now tighten the screws on per-token pricing?
Because they thought they could. Turns out the Chinese and Elon had other plans.
The Chinese providers are just as much getting subsidies from the CCP, and Musk/SpaceX is (indirectly) raiding retirement funds to fund the bonanza.
There is no evidence Chinese providers are getting such subsidised, and in fact apparently Jinping (who presumably knows what the CPP is doing) was surprised when DeepSeek and Qwen generated so much buzz. In China, AI inference is just viewed as another basic utility, much like an e-mail provider or a mobile phone network.
I'm not aware of how Musk/SpaceX are "raiding retirement funds"?
The truly cheap Chinese models are usually the open weights ones so while there may be a training subsidy, the inference prices reflect real costs.
I get what you're saying, but I'm guessing that people asking how to pee is a drop in the bucket compared to the agentic loops being called to rename some variables across a project.
The fact is, you wouldn’t be okay with the real prices as well. All indications point to opus not being subsidised but having huge margins.
GLM 5.2 is not subsidised - it is an Opus tier model that costs a small fraction. I doubt you would be okay and all problems would suddenly vanish.
GLM 5.2 isn't quite modern Opus tier, as seen in this comparison where Opus 4.5 scores 4/5 on some coding tasks where GLM 5.2 scores 0/5: https://www.tryai.dev/blog/gpt-5.6-build-off-12-models But yes, GLM 5.2 is cheap.
But the real standout on price is DeepSeek V4 Flash, which competes, more or less, with models in between Sonnet and Haiku. From third-party providers, it costs around $0.09/M, $0.18/M out, compared to $3M/in, $15M/out for Sonnet and $1M/in, $5M/out for Haiku. To get the price of DSv4 (Flash and Pro) so low, DeepSeek did a lot of innovative optimization work that will likely show up in other open weight models in the future.
I bet when hot and cold running water first came out you would have questioned whether people "needed" that as well.
Plainly it was not a "need" since people managed for thousands of years without it.
This is what progress looks like.
It literally has only downsides except some convenience for people who don't want to think and don't want to work.
For a context: France relies heavily on automotive transport, plus it's a home to enormous agricultural sector, tractors are literally everywhere in the country during the summer. To a certain degree, structurally it resembles USA a lot.
However they also quite famously rely on a majority of nuclear power for their electric grid. Great for France, but that makes them an already-low carbon emitter compared to many others and an ungenerous comparison.
legality of the datacenters aside, I wonder why countries don't at least demand that they're totally carbon neutral or free. it's possible today. it's not like it's sci-fi.
USA is partly a petrostate so regulatory capture is a problem. To mitigate the climate catastrophe it would be important to ramp down fossil fuel production in a big hurry, but it's not in sight.
In Europe this is covered by the emission trading system (EU ETS) and datacenters have to share the same shrinking emissions quota as other industries.
I think most ways of obtaining carbon neutrality are a little bit BS, that's why.
An alternative is what Google is theoretically aiming for: being carbon-free. But they've already started using language describing it as a moonshot or idealistic goal so seems likely they'll abandon that
https://sustainability.google/reports/247-carbon-free-energy...
Carbon offsets are absolutely a scam but you could easily force data centers to provide their own renewable energy.
Not to worry.
There are laws in the EU that will save the planet, like drinking from soggy paper straws instead of plastic and requiring caps stay attached in plastic bottles.
And just to make sure, at least in Poland they now charge you $0.10 if you buy anything plastic until you bring it back to the grocery store empty.
We are safe.
We don't really need the French on the other hand, how could we live without AI?
Yeah, we are not talking about ecological impact of France enough
Yann Le Cun enters the chat...
Pretty small if you consider the value they provide, honestly.
And they'll ride the transition to green energy for "free".
Is this value in the room with us?
Do you know what site you're on?
But... Datacenters don't burn anything, right? Powerplants do and we try to switch all the transport and heating and whatever to be electric.
So the answer is to build the damb nuclear power and a lot of it and price CO2 emissions at the actual cost of sucking the thing back out if the atmosphere
They do have a growing amount of Scope 1 emissions (emissions from their on site sources) which originally was primarily on site diesel but due to grid interconnect delays have been growing number of on site gas turbines.
This certainly wouldn’t be necessary with adequate generation and transmission capacity.
This is true, but I'm pretty firmly of the opinion that these data centres shouldn't be built, or at least allowed to operate until/unless they can be powered cleanly and without cornering the market and driving out existing consumers of power.
If they're so keen to build that they're willing to fund power generation (e.g. on site gas generators) then it should be clean/renewable (solar, wind, small modular reactors, full scale nuclear plants, whatever).
Degrowth is bad but so is ignoring the planet, the environment, and people's health to get ahead faster in business.
Make all DCs provide their own power and suddenly they’d be large amounts of solar and battery in cheap scrubland
A typical US DC costs about $35b per GW. Solar and battery would increase that to about $45b.
Then during summer it would generate so much excess power that it would run all the domestic air conditioning you could need.
The need for memes knows no bounds. In short order the majority of power usage worldwide will be for compute and newer generations will wonder how it took so long.
Which is why things like nuclear power plants, grid upgrades, hydroelectric projects, and intelligently placed wind/solar (instead of placing it due to subdisies or political concerns) should have been done a long time ago.
> price CO2 emissions at the actual cost of sucking the thing back out if the atmosphere
This is the only relevant bit actually. The rest will follow from there. And in principle, at least in Europe, we already have some mechanisms to do this. We'd "just" have to up the prices.
BUT of course with the right wing on the advance, and with them having identified basic physics (i.e. climate change) as a culture war terrain, this keeps being watered down... Oh well... This is why we can't have nice things... like a future...
The left wing version of climate conspiracy is that climate change will end humanity itself. This is not based on science.
Often repeated everywhere as a trump card to get what they want - crush technological and economic progress.
If enough climate systems collapse, lots of existing farms will no longer be viable. That means famine and migration, which means war, which means lots of death. I don't know anyone who thinks we'll see extinction (outside of possible "hothouse earth" scenarios, where we become a second Venus) but societal collapse is definitely on the table. Saying this is "not scientific" would just be you not understanding the science.
"Climate conspiracy"? Like you, mean, the conspiracy of climate scientists to publish facts to the best of their understanding?
I don't know what exact strawman you're arguing against, although I'm sure you can always find some idiots saying something like what you say. But scientific consensus has long been that it will lead to increasingly extreme weather and mass extinction, which we seem to be on track for. Of course, we can't know for sure what the consequences are untill we do the experiment, which in this case means potentially destroying large sections of the biosphere and living with increasingly destructive weather patterns. Surely that risk is worth at least legislating that hyperscalers need to spend some of their billions on solar panels?
End humanity, maybe not. Causing very painful social changes, however, is absolutely on the cards.
More than on the cards, it is already happening.
Fair point! I meant "vastly more painful change than already is happening".
> Causing very painful social changes, however, is absolutely on the cards.
Worse than voluntarily inviting masses of incompatible cultures into western countries like we're already doing?
Yes.
And while the populist right love importing incompatible cultures en mass, that will have to stop.
It’s anxiety about tipping points more than conspiracy theories. If you look at the figures in the various scenarios without even factoring them in: hundreds of millions to billions face food and water insecurity. You may be insulated enough from the direct effects but what about what they trigger? Already even our wealthy societies are being strained by rising food prices, extreme weather, and dysfunctional migration.
We don’t know exactly what the tipping points would lead to but we do know it would be some degree of a more severe and abrupt decline across the board.
Most conspiracy theories think food grows in supermarkets and because they have a job which pays them a good global income they can just pay a bit more.
> The left wing version of climate conspiracy is that climate change will end humanity itself.
That is unrealistic, but also not what a conspiracy is. It’s also a red herring, as nobody serious is claiming that. They are talking about civilisation changes, which we are already seeing.
At some point you have to engage with the arguments besides shouting "no, it’s you".
*raise the prices
Related:
Microsoft latest report shows 25% emissions raised due to AI data centers
https://news.ycombinator.com/item?id=48870229
Man, we are cooked, literally
The solution is simple: require datacenters to overprovision solar panels and grid-scale batteries for themselves, and use that capacity to strengthen the grid and transition off of hydrocarbons.
You can’t get a grid tie for those panels in most of the US right now. The process for connecting to the grid is done serially, and requires a large study for any new generation.
Down. So they are self sufficient in winter and in summer they use excess energy to create green hydrogen.
No idea what you're talking about. My local utility lit up 100MW of solar over the last year alone. Everywhere I look is doing the same.
He's talking about the interconnection queue. You see the 100MW of solar they're wiring up, but not the hundreds of gigawatts in the queue.
I imagine it'd be a lot easier to get a giant datacenter through the queue and connected if said datacenter also generated more than it's own needs with solar and grid scale battery. You'd essentially be asking the grid to act as a backup and sink for spare generation. Avoiding a need for central generation capacity buildout.
Sure you can. The datacentre builders just don't want the (fairly modest) extra expense to do so properly. Obviously some preparatory work is required before dumping a lot of extra capacity into the grid.
My state allowed the power utilities to charge a modest fee ($10,000 to $100,000, depending on project size) before a yet-to-be-built data centre could demand a large amount of electricity. The amount of planned data centres went down by an order of magnitude. The truth is that most of the data centre builders (not all) do not want to be responsible citizens and are simply extracting value and wealth from other people, including from the power utilities and grid operators.
No, wait! The increased productivity will lead to a decoupling of the economy from resources consumption and GHGs emission. Just one more data center.
/s
It’s such a tiring narrative isn’t it ?