Worse, the burden [generating energy] increasingly falls on the buyer [data center developers]
I don't believe this is worse, but appropriate. The grid is a shared resource, used by enterprise and individuals. If some class of consumers demand an outsized share of that resource, they should pay an outsized share of its maintenance and development. I don't see that happening.
It's as if trucking companies flooded the highways with so many trucks that people couldn't commute to work anymore.
When the first wave of these datacenter stories came out, the utilities complained that they couldn't finance the work because the data centers refused to commit to paying for them over the longer term, leaving other ratepayers on the hook for the investment.
This seems to further confirm that, though now the data centers are being portrayed as the victims in this.
The sentence immediately following your quote:
> securing grid-connected power now often requires developers to post substantial letters of credit, security deposits, or sign take-or-pay commitments to fund the generation built to serve their load.
The question is, if they couldn't commit to paying the utility, how can they finance building their own independant grid equivalent? Did they find some other sucker to take on the risk?
In the past major industrial electricity consumption facilities (like aluminum smelters) have been build near big power plants. Or dedicated power plants have been build to power major industrial electricity consumption facilities.
For example:
PS5 Combined Cycle Power Plant for Aluminium Bahrain (ALBA)
> It's as if trucking companies flooded the highways with so many trucks that people couldn't commute to work anymore.
Funnily, that's precisely what happened to America's rail network. Enterprise demanded unlimited right of way to the point that it's laughable to even consider transit by rail these days.
It's going exactly as well as you'd imagine. I expect our power infrastructure is going to be treated exactly the same as our rail infrastructure: legacy public assets to be exploited indefinitely[0] for free[1] profit, and never invested in until failure becomes an existential threat[2]
It's not "worse" per-say in that the buyer is crunching the numbers and deciding to go with on-prem generation for a lot of their jiggling electrons.
But holy shit is something broken if that sort of approach is more economically effective (middle of nowhere Alaska not withstanding) than having the people who not only specialize in pipes for jiggling electrons, but have all sorts of preferable treatment from regulators and expertise in financing these sorts of things, just enlarge the size of the pipe to your door.
Furthermore, the grid wants to be as big as it can for basically all the same inflow/outflow smoothy reasons a bank or an insurer wants to be. Something is very, very broken here if big, highly financed corporate interests are saying "no I'm not gonna get all my power from the grid" and the grid is saying "actually this is fine". They should be jumping at the chance to essentially have someone else at least partly subsidize their improvements.
Everyone wants to whine about future volatility, will the data centers be around to pay for what they ordered, etc, etc. But the energy industry has a long, long, long track record of financing this risk away. Oil and related energy infra has been boom and busting for over a century.
30yr ago it would have been unfathomable that anyone without an extremely niche case would go where the energy is rather than just ordering up power from the grid.
Not sure how you calibrate your "niche" scale, but foundries and aluminium smelting plants have commonly been sited near generators in Britain. I suppose that transmission losses matter at the GW scale; and also that a national grid is a complex system, with phase slump being a cause of failure even with sufficient generation capacity.
Every DC I've visited - probably ten - has on-site generators, for redundancy. To a certain type of mind, that's capital going to waste. I can imagine a bean-counter desire to sweat that plant.
I agree that today's economics is very broken, but co-siting generation and consumption is probably not a symptom.
Aside: Language is whatever people use, but, "per se" is Latin for "in itself". "Per-say" is an eggcorn.
>, but foundries and aluminium smelting plants have commonly been sited near generators in Britain.
IDK about Britain, especially historically when you would have been building aluminum smelters. It's not normal in the US more or less since the advent of the grid. I'm sure people will be along shortly to construe exceptions or special cases where the fossil fuels were extra cheap as the norm or heat was what they also needed and it was cheaper to capture that locally and make some electricity too.
>Every DC I've visited - probably ten - has on-site generators, for redundancy. To a certain type of mind, that's capital going to waste. I can imagine a bean-counter desire to sweat that plant.
But they don't use those for any semblance of "normal" load. Those are strictly for redundancy not "ah it's 3pm, power's getting expensive better fire up the genset for 4hr". At least that's normally how it is. It's not normal for it to be possible to out compete the utility even after fixed costs are amortized in. Bigger generation is more efficient, if you're buying fossil fuels then buying them in bulk is more efficient. Etc. etc. It should be cheaper to have the people who benefit from all that string you a wire.
In most places aluminum smelters have been located near power plants, because aluminum smelters have large electricity demand and transporting large amounts of electricity over large distances is expensive (high capacity power lines are expensive).
Electricity costs are between 40% - 50% of costs of the finished aluminum.
Also, reliability.
"Due to the nature of the process, power outages have the potential to cause damage to production cells as the molten liquids could solidify in absence of adequate current. For this reason, production facilities need to be near secure and reliable sources of energy."
Like in all these historical cases there's been some sort of high voltage transmission in between. Sure, not a hundred mile line through god knows how many jurisdictions, but still plenty, a far cry from the "this won't pencil out if we have go farther than across the street" that we're currently seeing.
Basically what I'm saying is that if datacenters with the closest thing you'll ever see to infinite piles of cash can't expand the grid then we're in for a very rough time.
> Basically what I'm saying is that if datacenters with the closest thing you'll ever see to infinite piles of cash can't expand the grid then we're in for a very rough time.
Isn't the problem that the grid operators/generators have been asking for them to commit part of this huge pile of cash before they start the expensive buildout process but the DCs are refusing to do so?
I don't think it says anything about having a rough time, the grid would expand if those requesting major expansions would commit paying for them. Instead the grid is being told "we need it, build it, and we will pay after" which for the kind of massive investment necessary to push extra dozens of GW is absurdly risky. If the grid starts major expansion and AI investments crumbles in 3-5 years then it's both a waste of resources while also requiring grid operators to charge more the current consumers to bridge the gap left by the AI DCs.
The only thing I can see showing a very rough time is the entitlement of these companies, just because they are surfing a humongous phase of investments they feel they don't need to play by the rules that always existed: if you need something massive you gotta pay for it.
> a far cry from the "this won't pencil out if we have go farther than across the street" that we're currently seeing.
I'm not sure what crying has to do with it but the losses across two streets are double those "across the street", across 4 streets - 4x loss, etc.
Public utilities minimize losses over a wider area with many customers, connecting a GW datacenter to the public grid isn't as simple or as safe as connecting another house, the best solution here is to keep large consumers separate to avoid grid instabilities.
> Basically what I'm saying is that if datacenters with the closest thing you'll ever see to infinite piles of cash can't expand the grid then we're in for a very rough time.
It's not true that they "can't expand the grid", they don't want to. They were eager to leach on the public grid but when asked to pay their fair share to be properly integrated by the public utilities, they got worried that some minuscule share of their "infinite piles of cash" may benefit the public - nope, you can't have that under any circumstances.
> The grid is a shared resource, used by enterprise and individuals. If some class of consumers demand an outsized share of that resource, they should pay an outsized share of its maintenance and development. I don't see that happening.
That makes sense up until you realize that the data centers aren't just burning watts for fun. They're providing a service that is in extremely high demand to a very high amount of those other electricity customers.
> It's as if trucking companies flooded the highways with so many trucks that people couldn't commute to work anymore.
And those trucks are delivering products to the stores those commuters shop at.
And we have a perfect instrument to express this relationship: money. If the service is really important to people they will pay money for it, and the companies can pay some of that money to utility providers. The economy works as intended, everyone is happy. Absolutely no need to invent subsidies or move the burden onto consumers
The problem is that the datacenters need grid investments. And if the utility makes those debt-backed investments but then the DC goes bankrupt, households are on the hook for the debt servicing.
Utilities requiring ten year of prepayment from the DCs makes sense.
As an (eastern) European this is to me a greater concern than the Russian invasion in terms of availability of gas for the winter, considering how so many of those places run on gas turbines.
Putin only has a certain amount of the resource on tap and sells it to someone eventually. Meanwhile data centers will eat up whatever you throw at them at any price, especially that in principle they don't need to be connected to the grid, so I imagine setting up their power supply must take months at the longest.
The issue is with the price - why bother shipping LNG half the planet when one can just sell it at a significant markup domestically?
When the price is right all kinds of contracts and agreements are usually thrown out the window - especially if the penalties are small enough to make it worth it.
That's the point - a huge part of the cost of LNG is the "L".
Way easier to just sell it closer to home and not bother with transporting it in a tanker which has to return empty. Demand is clearly there and not somewhere else.
Supply domestically outstrips demand, so no they can not sell it locally at that price. The US did not spend billions building LNG terminals for nothing.
Alternatively the crisis can be used to accelerate right development - quickly build up solar farms and transmission lines. Both of these can be done quickly if there were a political will. Yet, that will is missing.
32.5 gigawatts (GW) of solar power capacity, down 18% from last year.
11 GW of wind power capacity, up 19% from last year.
20 GW of thermal power capacity, up 414% from last year.
1.2 GW of hydro power capacity, down 36% from last year.
1.2 GW of nuclear power capacity. "
Edit: to the proponents of nuclear - i think nuclear is very damaging to society as it promotes corrupt crony-government capitalism instead of market forces.
Right off the top of my head, to power a 500MW Datacenter we need:
2.2-2.5 GWdc solar capacity which at 600W/panel amounts to 14-17sq mi plus additional 8-12GWh of storage to deal with nighttime, two-three days of cloud cover is not going to work with these numbers.
Ballpark 5-7BN$
Nuclear otoh, 1GW continuous - gives constant power, badly managed first of a kind (or first one after decades) build will be at around 10-15BN and that’ll cover two 500MW data centres.
There’s also the second/third degree order effects nuclear power stations have of creating jobs and industrial manufacturing demand. To run a nuclear power station you need to employ 1000 people (engineers plus support staff) - that’s a small town’s worth of adults. So you’ll need a town with a school, hospitals, stores - those need staffing as well.
Unfortunately building nuclear is not something that’s currently a feasible path as it requires patient capital and long term vision and planning.
I like how balanced their energy mix is. It is very obvious that China is optimizing for capacity and availability. There isn't really a push for clean energy sources for political, or climate, reasons. They are deployed when it makes sense, backed up by robust coal and nuclear sources.
In Europe, we approach energy generation as a political, or climate problem. We are building solar and wind power sources, not to make energy cheaper, or to make grid more resilient, but to fulfill an ideological goal.
The results are, not great, to be honest. The energy prices have increased substantially, and are now driving our chemical industry bankrupt.
Edit: I do not dispute the climate change. I am only highlighting impacts of current policy.
China is building all types of generation because they are rapidly growing their electric network (generation doubled from 2014 to 2024).
Europe's generation is roughly flat during the same period.
It doesn't make sense to build a lot of everything in a system without growth in generation. Replacing decommisioned generators would be enough. Growth in solar and wind generation (for ideological or economic reasons) means there's less reason to build new capacity of other types. There's complications there with firm vs intermittent capacity, but that's a different discussion.
US electrical generation was also flat from 2000 to 2020, but seems to be growing again since then, but not anywhere near China's growth rate.
China is still installing much more new coal power capacity then is decommissioning old coal power capacity.
Also the average age of Chinese coal power plants is much lower then the average age of US coal power plants. These new Chinese coal power plants could be used for decades.
What you said does not technically contradict what I said. We have to decide what it is that we wish to measure.
I do not pay attention to the "decommissioning" statistic, because a binary on/off switch is an arbitrary threshold.
Example -- China reduces the output of every coal plant to 1%, they would be "decommissioning" nothing, yet it would be fair to say they've abandoned coal.
That's why it's a better metric to look at change in net coal usage (which is flat or going down in China last year), which factors in non-binary underutilization of old coal plants, as new coal and renewables come online.
You wrote "new coal is just replacing old coal", in terms if of coal power capacity this is incorrect, as many new coal power plants are build in China.
The utilization of coal power plant fleet is decreasing, the total amount of produced electricity from coal is decreasing.
"Coal usage" in China is different from "Coal usage for electricity production" in China, as the coal use as a feedstock for chemical and liquid fuels will increase in China.
In 2025 China mined more coal than in 2024, at the same time China produced less electricity from coal in 2025 than in 2024.
"China's coal output rose to a record in 2025, statistics bureau data showed on Monday, as lower domestic prices prompted buyers to cut imports and rebuild stockpiles with cheaper local supply, although the rise was limited by regulator efforts to curb production growth.
Production in 2025 reached 4.83-billion metric tons, up 1.2% from 2024, according to the National Bureau of Statistics."
Stats on wikipedia [1] stopped breaking out coal from other fossil fuel ('thermal') production in 2021.
But coal was 4.1 million GWh in 2014 and 5.0 million GWh in 2021; that's a lot of (relatively) new coal. Fossil fuel growth since 2023 is a lot less than earlier years, so maybe they have hit saturation for fossil fuel generation. I would expect high fossil fuel prices so far this year would drive usage lower, the stats for 2026 should be interesting.
Energy prices globally spiked in 2022 when Russia invaded Ukraine (again). I conclude that energy is tied more closely to fossil than solar. Happy to stand with you and throw rocks at stupid solar policy, though.
>I like how balanced their energy mix is. It is very obvious that China is optimizing for capacity and availability. There isn't really a push for clean energy sources for political, or climate, reasons. They are deployed when it makes sense, backed up by robust coal and nuclear sources.
yes, they don't seem to fall for the "solar is unstable" trap and recognize new reality - the solar + wind smoothed by nuclear/gas is the new baseline
To the comment on the coal below : not really. You don't need that much as a smoothing capacity.
If we build a terawatt of green energy capacity then turn around and immediately use it for datacenters rather than to replace fossil fuel generation, we haven’t gained anything. If anything we’ve lost out.
Proponents of solar power always like to equate 1 GW of solar power capacity with 1 GW of wind power capacity or 1 GW of thermal power capacity, ignoring capacity factors, ignoring need for reliable backup.
Sometimes they ignore the transportation costs, as for example: capital costs for transporting 1 GW of electricity over 1,000 km using High Voltage Direct Current (HVDC) technology costs more then the 1 GW peak solar power plant attached to the power line.
China builds everything solar, coal, wind, hydro, nuclear. So if you want follow the example of China, then you have to, like China, build everything and don't just pick one part of Chinas energy strategy.
Also I would not present China as an example of market capitalism, they have a very strong government capitalism. Probably the strongest government capitalism in the world.
Yeah and the richest person on the planet with full vertical integration (with partner) Elon musk and Tesla aka solar roof and power wall wants to sell DC in space and hasn't fixed any use of gas turbine for Colossus 1&2.
Without enforcement it will happen continuously with snail pace
You are correct in the sense that they can stop work in a way many generic server use cases can't (which is seen in lowering power supply reliability requirements as the article mentions), but running expensive servers at 50% utilization would dramatically affect the revenue generated per capital invested - IE you couldn't afford to buy the servers.
If it just costs more, similar to how it's possible to have a low water use datacenter, then in a perfect spherical cow economic universe then the lower cost option is better.
In the real world you'd need to add some cost to account for the externalities on water and GHG.
Do the numbers still work out for the gas powered plant? If not then you're just exploiting unaccounted for externalities.
If you invest into chips that deprecate in value really fast, not utilizing them to their full capacity because of power constraints would be counter productive.
I'm hoping the kind of gas turbines being installed on these datacenters are capable of quickly responding to a load change, meaning the primary source of energy could be solar during the day and complement when there isn't enough energy.
But I haven't looked into where these datacenter are being placed, I'm assuming that although solar is cheap now, the surface needed would make the purchase of nearby land probably not worth it. These new categories of datacenters are becoming very energy dense...
you're just incorrect. You probably missed 2 points :
- battery storage
- and in the article
"However, AI labs and some hyperscalers have relaxed those requirements as there is now a lower uptime tolerance applied to both inference and training, not just training. Many of Meta’s self-built AI datacenters, for example, target just two nines of uptime and forgo backup generators entirely, as detailed in our Industrials Model."
Economic development is great and I'm all for it, but it shouldn't be done at any cost. I quite like our atmosphere, ideally with less than 400ppm of CO2.
We're finally seeing signs of economic development decoupling from primary energy consumption, there are positive signs.
I find that it's just a shame to be so close to energy abundance - through renewables - to only be set back by the installation of "temporary" gas turbines. The economic incentive is so great that delaying the datacenters in order to power them cleanly isn't even considered: there's no incentive to do so. This may sound a little alarmist, but if it's GW of capacity in 2028, it's TW in the mid 30s.
I'd much rather live in a heavily compute-constrained world for the next 5 years or so than in a world that keeps increasing its gas consumption.
There's just no way to force our incentive to align with preserving our climate, and it's making me extremely depressed.
By your logic we shouldn't question any development, at any cost, including if it contributes to making our world uninhabitable for billions of people within our life's timescale?
And especially terrible for the communities that have to live with gas turbines or other local power generators as neighbors. Noise and air pollution constantly [1].
But fuck them, they are poor people so we don't care about them /s .
Additionally, people against data centers are accused of being paid by China [2]
Let's go back to have open air sewers while we are at it. I mean, if the goal is to bring back stuff that lowered life expectancy and reduce 50+ years of progress, I don't see why not. I really believe that as arguments go, yours is truly terrible.
Additionally, you proved my point that the majority really doesn't care about the folks whose quality of life worsen because someone built a datacenter close to their homes. Or the fact that turbines still pollute the air with NOx emissions (which cause lung cancer and other respiratory problems).
You are basically saying that it is not so bad, but (I guess) it is because you aren't close to a datacenter not depending on the grid. And I trust more hearing about live accounts of people being affected, than a "trust me bro, it is not as bad". Are the accounts cherry-picked? Maybe, but they are also supported by videos that substantiate what they are saying.
Your entire life is subsidized by those wealthier than you. Get off your high horse.
Did you build your own house?
Did you build your own car?
Do you make your own power?
What really has been your contribution to society.
High horse? Man, my country was under communist dictatorship and I didn't see a banana until I was 8 years old. However, despite that, I am aware of how lucky I got in life and I try to be mindful of those that are not as much.
Also, I don't believe one bit my life has been subsidized. Wealthy people don't throw money around if there is no profit to be made. At best they subsidize stuff to get you hooked before raising prices and raking the money in.
Additionally, I paid people to build my own house with money that I earned, and same for my car. And I now live in a country where we pay living wages and not force people to do 2-3 jobs to stay afloat. So I didn't exploit anyone as those people received months of salaries for them and their families thanks to my house.
The only thing I can concede to your argument is that things like my phone or laptop are cheap because people in the world are exploited while mining and in Foxconn like factories and for that I am trying to limit my tech footprint.
Are you aware that public spending is a transfer from the public sector to the private, and that poor people spend approximately 100% of their income? Would you argue that stock buy-backs are a better use of public money?
At some point, probably in the 80s, the prophets of Mammon convinced policymakers that the liquidity of capital markets was the highest good, that no liquidity was enough, and that the financial class is inherently productive.
When we bailed out the banks in 2008, it amounted to a wealth transfer from the public purse to Jamie Dimon. That's your counterproductive welfare spending.
Texas had blackouts last year because free-market fundamentalism. Perhaps Texans should run their fridges on investment boom.
Yeah sure dogma. Capitalism is evil, yet it lifted more people out of poverty than anything in history.
Of the three economic powers, the one with the most social welfare and least capitalist is the one that never recovered from the 2008 crisis. Funny how that is. Strange when all the welfare gets spent in the economy.
The one with the unproductive capital buying back shares is building GW datacenters. The other with "productive" capital, is encouraging people not to buy ACs because the electric grid cant handle it and using energy is bad for the environment.
The difference is Europeans expect government to "fix" things. Americans expect companies to do that.
Eg, the disastrous energy policies all across Europe which have made price of energy insanely high with no hope of ever coming down (these high prices are all locked in with long term contracts).
America just builds lots of new power. Because of fewer market distorting policies the new power comes online at cheaper marginal rates so the price doesn't go up. And now Texas has some of the greener and cheapest energy in the West.
Texas's CO2 per KWh of electricity (which is roughly equal to the US as a whole) at about 330 is equidistant between China's (470) and EU's (190) it is not close to the greenest in the West.
I am interested to see how this plays out and what will happen to potentially stranded assets. Not even talking about outdated chips but all these gas turbines. I appreciate what this push can do to the built out of grids etc.
Developing world will soak them in a heartbeat. Unreliable grid is one of the problems they have that slow their industrialization and development pace. If you are foreign investor - LNG terminal + turbines may be more expensive, but if you bypass some of the bribes and incompetence it may be worth it.
"data centres operate 24/7 with an industry standard of “five-9s” reliability—99.999% uptime, equivalent to less than 5.3 minutes of downtime per year. This level of reliability requires access to a stable and highly reliable source of electricity."
Many kinds of renewables are day/night, weather dependent, you don't get this level of reliability without a fossil fuel backup. If you already need to build a dedicated gas power plant to power your data center, then the decision to invest in an additional renewables power plant depends on the costs of the natural gas. (Additional capital expenses of the renewables power plant vs reduction in consumption of the natural gas).
Interesting analysis. In it they say
I don't believe this is worse, but appropriate. The grid is a shared resource, used by enterprise and individuals. If some class of consumers demand an outsized share of that resource, they should pay an outsized share of its maintenance and development. I don't see that happening.It's as if trucking companies flooded the highways with so many trucks that people couldn't commute to work anymore.
When the first wave of these datacenter stories came out, the utilities complained that they couldn't finance the work because the data centers refused to commit to paying for them over the longer term, leaving other ratepayers on the hook for the investment.
This seems to further confirm that, though now the data centers are being portrayed as the victims in this.
The sentence immediately following your quote:
> securing grid-connected power now often requires developers to post substantial letters of credit, security deposits, or sign take-or-pay commitments to fund the generation built to serve their load.
The question is, if they couldn't commit to paying the utility, how can they finance building their own independant grid equivalent? Did they find some other sucker to take on the risk?
venture startups (nuclear) and alternative energy public equities
In the past major industrial electricity consumption facilities (like aluminum smelters) have been build near big power plants. Or dedicated power plants have been build to power major industrial electricity consumption facilities.
For example:
PS5 Combined Cycle Power Plant for Aluminium Bahrain (ALBA)
https://www.idom.com/en/project/ps5-combined-cycle-power-pla...
India
https://nalcoindia.com/business/operation/captive-power-plan...
Tiwai Point Aluminium Smelter, New Zealand
https://en.wikipedia.org/wiki/Tiwai_Point_Aluminium_Smelter
> It's as if trucking companies flooded the highways with so many trucks that people couldn't commute to work anymore.
Funnily, that's precisely what happened to America's rail network. Enterprise demanded unlimited right of way to the point that it's laughable to even consider transit by rail these days.
It's going exactly as well as you'd imagine. I expect our power infrastructure is going to be treated exactly the same as our rail infrastructure: legacy public assets to be exploited indefinitely[0] for free[1] profit, and never invested in until failure becomes an existential threat[2]
[0] until next quarter
[1] losses are the public's problem
[2] unless the government bails you out
It's not "worse" per-say in that the buyer is crunching the numbers and deciding to go with on-prem generation for a lot of their jiggling electrons.
But holy shit is something broken if that sort of approach is more economically effective (middle of nowhere Alaska not withstanding) than having the people who not only specialize in pipes for jiggling electrons, but have all sorts of preferable treatment from regulators and expertise in financing these sorts of things, just enlarge the size of the pipe to your door.
Furthermore, the grid wants to be as big as it can for basically all the same inflow/outflow smoothy reasons a bank or an insurer wants to be. Something is very, very broken here if big, highly financed corporate interests are saying "no I'm not gonna get all my power from the grid" and the grid is saying "actually this is fine". They should be jumping at the chance to essentially have someone else at least partly subsidize their improvements.
Everyone wants to whine about future volatility, will the data centers be around to pay for what they ordered, etc, etc. But the energy industry has a long, long, long track record of financing this risk away. Oil and related energy infra has been boom and busting for over a century.
30yr ago it would have been unfathomable that anyone without an extremely niche case would go where the energy is rather than just ordering up power from the grid.
Not sure how you calibrate your "niche" scale, but foundries and aluminium smelting plants have commonly been sited near generators in Britain. I suppose that transmission losses matter at the GW scale; and also that a national grid is a complex system, with phase slump being a cause of failure even with sufficient generation capacity.
Every DC I've visited - probably ten - has on-site generators, for redundancy. To a certain type of mind, that's capital going to waste. I can imagine a bean-counter desire to sweat that plant.
I agree that today's economics is very broken, but co-siting generation and consumption is probably not a symptom.
Aside: Language is whatever people use, but, "per se" is Latin for "in itself". "Per-say" is an eggcorn.
>, but foundries and aluminium smelting plants have commonly been sited near generators in Britain.
IDK about Britain, especially historically when you would have been building aluminum smelters. It's not normal in the US more or less since the advent of the grid. I'm sure people will be along shortly to construe exceptions or special cases where the fossil fuels were extra cheap as the norm or heat was what they also needed and it was cheaper to capture that locally and make some electricity too.
>Every DC I've visited - probably ten - has on-site generators, for redundancy. To a certain type of mind, that's capital going to waste. I can imagine a bean-counter desire to sweat that plant.
But they don't use those for any semblance of "normal" load. Those are strictly for redundancy not "ah it's 3pm, power's getting expensive better fire up the genset for 4hr". At least that's normally how it is. It's not normal for it to be possible to out compete the utility even after fixed costs are amortized in. Bigger generation is more efficient, if you're buying fossil fuels then buying them in bulk is more efficient. Etc. etc. It should be cheaper to have the people who benefit from all that string you a wire.
In most places aluminum smelters have been located near power plants, because aluminum smelters have large electricity demand and transporting large amounts of electricity over large distances is expensive (high capacity power lines are expensive).
Electricity costs are between 40% - 50% of costs of the finished aluminum.
Also, reliability.
"Due to the nature of the process, power outages have the potential to cause damage to production cells as the molten liquids could solidify in absence of adequate current. For this reason, production facilities need to be near secure and reliable sources of energy."
https://arcticecon.wordpress.com/2012/02/15/aluminium-smelti...
"near" vs "next door or closer".
Like in all these historical cases there's been some sort of high voltage transmission in between. Sure, not a hundred mile line through god knows how many jurisdictions, but still plenty, a far cry from the "this won't pencil out if we have go farther than across the street" that we're currently seeing.
Basically what I'm saying is that if datacenters with the closest thing you'll ever see to infinite piles of cash can't expand the grid then we're in for a very rough time.
> Basically what I'm saying is that if datacenters with the closest thing you'll ever see to infinite piles of cash can't expand the grid then we're in for a very rough time.
Isn't the problem that the grid operators/generators have been asking for them to commit part of this huge pile of cash before they start the expensive buildout process but the DCs are refusing to do so?
I don't think it says anything about having a rough time, the grid would expand if those requesting major expansions would commit paying for them. Instead the grid is being told "we need it, build it, and we will pay after" which for the kind of massive investment necessary to push extra dozens of GW is absurdly risky. If the grid starts major expansion and AI investments crumbles in 3-5 years then it's both a waste of resources while also requiring grid operators to charge more the current consumers to bridge the gap left by the AI DCs.
The only thing I can see showing a very rough time is the entitlement of these companies, just because they are surfing a humongous phase of investments they feel they don't need to play by the rules that always existed: if you need something massive you gotta pay for it.
> a far cry from the "this won't pencil out if we have go farther than across the street" that we're currently seeing.
I'm not sure what crying has to do with it but the losses across two streets are double those "across the street", across 4 streets - 4x loss, etc.
Public utilities minimize losses over a wider area with many customers, connecting a GW datacenter to the public grid isn't as simple or as safe as connecting another house, the best solution here is to keep large consumers separate to avoid grid instabilities.
> Basically what I'm saying is that if datacenters with the closest thing you'll ever see to infinite piles of cash can't expand the grid then we're in for a very rough time.
It's not true that they "can't expand the grid", they don't want to. They were eager to leach on the public grid but when asked to pay their fair share to be properly integrated by the public utilities, they got worried that some minuscule share of their "infinite piles of cash" may benefit the public - nope, you can't have that under any circumstances.
> The grid is a shared resource, used by enterprise and individuals. If some class of consumers demand an outsized share of that resource, they should pay an outsized share of its maintenance and development. I don't see that happening.
That makes sense up until you realize that the data centers aren't just burning watts for fun. They're providing a service that is in extremely high demand to a very high amount of those other electricity customers.
> It's as if trucking companies flooded the highways with so many trucks that people couldn't commute to work anymore.
And those trucks are delivering products to the stores those commuters shop at.
And we have a perfect instrument to express this relationship: money. If the service is really important to people they will pay money for it, and the companies can pay some of that money to utility providers. The economy works as intended, everyone is happy. Absolutely no need to invent subsidies or move the burden onto consumers
The problem is that the datacenters need grid investments. And if the utility makes those debt-backed investments but then the DC goes bankrupt, households are on the hook for the debt servicing.
Utilities requiring ten year of prepayment from the DCs makes sense.
What an absolute ecological disaster. If bubble there is, now would be a great time for it to pop.
As an (eastern) European this is to me a greater concern than the Russian invasion in terms of availability of gas for the winter, considering how so many of those places run on gas turbines.
Putin only has a certain amount of the resource on tap and sells it to someone eventually. Meanwhile data centers will eat up whatever you throw at them at any price, especially that in principle they don't need to be connected to the grid, so I imagine setting up their power supply must take months at the longest.
The US has incredible gas reserves. Much more that can be exported with ships.
The issue is with the price - why bother shipping LNG half the planet when one can just sell it at a significant markup domestically?
When the price is right all kinds of contracts and agreements are usually thrown out the window - especially if the penalties are small enough to make it worth it.
No they can not sell it with a markup domestically. LNG price is much higher.
That's the point - a huge part of the cost of LNG is the "L".
Way easier to just sell it closer to home and not bother with transporting it in a tanker which has to return empty. Demand is clearly there and not somewhere else.
Supply domestically outstrips demand, so no they can not sell it locally at that price. The US did not spend billions building LNG terminals for nothing.
Alternatively the crisis can be used to accelerate right development - quickly build up solar farms and transmission lines. Both of these can be done quickly if there were a political will. Yet, that will is missing.
So, we're talking 40GW. Lets see what China does:
https://energyandcleanair.org/china-energy-and-emissions-tre...
"In the first two months of 2026, China added:
32.5 gigawatts (GW) of solar power capacity, down 18% from last year. 11 GW of wind power capacity, up 19% from last year. 20 GW of thermal power capacity, up 414% from last year. 1.2 GW of hydro power capacity, down 36% from last year. 1.2 GW of nuclear power capacity. "
Edit: to the proponents of nuclear - i think nuclear is very damaging to society as it promotes corrupt crony-government capitalism instead of market forces.
Right off the top of my head, to power a 500MW Datacenter we need:
2.2-2.5 GWdc solar capacity which at 600W/panel amounts to 14-17sq mi plus additional 8-12GWh of storage to deal with nighttime, two-three days of cloud cover is not going to work with these numbers.
Ballpark 5-7BN$
Nuclear otoh, 1GW continuous - gives constant power, badly managed first of a kind (or first one after decades) build will be at around 10-15BN and that’ll cover two 500MW data centres.
There’s also the second/third degree order effects nuclear power stations have of creating jobs and industrial manufacturing demand. To run a nuclear power station you need to employ 1000 people (engineers plus support staff) - that’s a small town’s worth of adults. So you’ll need a town with a school, hospitals, stores - those need staffing as well.
Unfortunately building nuclear is not something that’s currently a feasible path as it requires patient capital and long term vision and planning.
So gas turbines it is at around ~1BN$
> is not something that’s currently a feasible path as it requires patient capital and long term vision and planning
HN arguing in favour of a planned economy - see, we aren't free-market cultists!
I like how balanced their energy mix is. It is very obvious that China is optimizing for capacity and availability. There isn't really a push for clean energy sources for political, or climate, reasons. They are deployed when it makes sense, backed up by robust coal and nuclear sources.
In Europe, we approach energy generation as a political, or climate problem. We are building solar and wind power sources, not to make energy cheaper, or to make grid more resilient, but to fulfill an ideological goal.
The results are, not great, to be honest. The energy prices have increased substantially, and are now driving our chemical industry bankrupt.
Edit: I do not dispute the climate change. I am only highlighting impacts of current policy.
China is building all types of generation because they are rapidly growing their electric network (generation doubled from 2014 to 2024).
Europe's generation is roughly flat during the same period.
It doesn't make sense to build a lot of everything in a system without growth in generation. Replacing decommisioned generators would be enough. Growth in solar and wind generation (for ideological or economic reasons) means there's less reason to build new capacity of other types. There's complications there with firm vs intermittent capacity, but that's a different discussion.
US electrical generation was also flat from 2000 to 2020, but seems to be growing again since then, but not anywhere near China's growth rate.
They're not really building anything other than renewables. Nuclear is a rounding error and new coal is just replacing old coal.
China is still installing much more new coal power capacity then is decommissioning old coal power capacity.
Also the average age of Chinese coal power plants is much lower then the average age of US coal power plants. These new Chinese coal power plants could be used for decades.
Last available data from 2025.
https://www.statista.com/chart/36007/power-capacity-of-new-d...
https://globalenergymonitor.org/projects/global-coal-plant-t...
What you said does not technically contradict what I said. We have to decide what it is that we wish to measure.
I do not pay attention to the "decommissioning" statistic, because a binary on/off switch is an arbitrary threshold.
Example -- China reduces the output of every coal plant to 1%, they would be "decommissioning" nothing, yet it would be fair to say they've abandoned coal.
That's why it's a better metric to look at change in net coal usage (which is flat or going down in China last year), which factors in non-binary underutilization of old coal plants, as new coal and renewables come online.
From ZeroGravitas in the parallel thread: https://ember-energy.org/data/electricity-data-explorer/?ent...
You wrote "new coal is just replacing old coal", in terms if of coal power capacity this is incorrect, as many new coal power plants are build in China.
The utilization of coal power plant fleet is decreasing, the total amount of produced electricity from coal is decreasing.
"Coal usage" in China is different from "Coal usage for electricity production" in China, as the coal use as a feedstock for chemical and liquid fuels will increase in China.
https://www.reuters.com/commentary/reuters-open-interest/chi...
In 2025 China mined more coal than in 2024, at the same time China produced less electricity from coal in 2025 than in 2024.
"China's coal output rose to a record in 2025, statistics bureau data showed on Monday, as lower domestic prices prompted buyers to cut imports and rebuild stockpiles with cheaper local supply, although the rise was limited by regulator efforts to curb production growth.
Production in 2025 reached 4.83-billion metric tons, up 1.2% from 2024, according to the National Bureau of Statistics."
https://www.miningweekly.com/article/china-2025-coal-output-...
Stats on wikipedia [1] stopped breaking out coal from other fossil fuel ('thermal') production in 2021.
But coal was 4.1 million GWh in 2014 and 5.0 million GWh in 2021; that's a lot of (relatively) new coal. Fossil fuel growth since 2023 is a lot less than earlier years, so maybe they have hit saturation for fossil fuel generation. I would expect high fossil fuel prices so far this year would drive usage lower, the stats for 2026 should be interesting.
[1] https://en.wikipedia.org/wiki/Electricity_sector_in_China
Stats on additional generation for the last few years:
https://ember-energy.org/data/electricity-data-explorer/?ent...
Energy prices globally spiked in 2022 when Russia invaded Ukraine (again). I conclude that energy is tied more closely to fossil than solar. Happy to stand with you and throw rocks at stupid solar policy, though.
>I like how balanced their energy mix is. It is very obvious that China is optimizing for capacity and availability. There isn't really a push for clean energy sources for political, or climate, reasons. They are deployed when it makes sense, backed up by robust coal and nuclear sources.
yes, they don't seem to fall for the "solar is unstable" trap and recognize new reality - the solar + wind smoothed by nuclear/gas is the new baseline
To the comment on the coal below : not really. You don't need that much as a smoothing capacity.
You mean + coal, they don't have a meaningful amount of nuclear (just under 5% of electricity mix and 2% of energy). 50% of electricity is coal.
If we build a terawatt of green energy capacity then turn around and immediately use it for datacenters rather than to replace fossil fuel generation, we haven’t gained anything. If anything we’ve lost out.
You gained the datacenter, which is providing a highly desired service... Are you European?
Personally I prefer a livable climate and affordable power over yet another datacenter for AI or mining crypto.
Proponents of solar power always like to equate 1 GW of solar power capacity with 1 GW of wind power capacity or 1 GW of thermal power capacity, ignoring capacity factors, ignoring need for reliable backup.
Sometimes they ignore the transportation costs, as for example: capital costs for transporting 1 GW of electricity over 1,000 km using High Voltage Direct Current (HVDC) technology costs more then the 1 GW peak solar power plant attached to the power line.
China builds everything solar, coal, wind, hydro, nuclear. So if you want follow the example of China, then you have to, like China, build everything and don't just pick one part of Chinas energy strategy.
Also I would not present China as an example of market capitalism, they have a very strong government capitalism. Probably the strongest government capitalism in the world.
Yeah and the richest person on the planet with full vertical integration (with partner) Elon musk and Tesla aka solar roof and power wall wants to sell DC in space and hasn't fixed any use of gas turbine for Colossus 1&2.
Without enforcement it will happen continuously with snail pace
solar farms are worst kind of power source for constant loads like datacenters running AI training
Solar power still can take a certain amount of load from any other source and saving money and CO2 while doing so.
And grid battery works and is cheap enough now.
Why? Unlike loads involving a real physical process there is absolutely no need for AI-training to be constant.
You are correct in the sense that they can stop work in a way many generic server use cases can't (which is seen in lowering power supply reliability requirements as the article mentions), but running expensive servers at 50% utilization would dramatically affect the revenue generated per capital invested - IE you couldn't afford to buy the servers.
If it just costs more, similar to how it's possible to have a low water use datacenter, then in a perfect spherical cow economic universe then the lower cost option is better.
In the real world you'd need to add some cost to account for the externalities on water and GHG.
Do the numbers still work out for the gas powered plant? If not then you're just exploiting unaccounted for externalities.
If you invest into chips that deprecate in value really fast, not utilizing them to their full capacity because of power constraints would be counter productive.
I'm hoping the kind of gas turbines being installed on these datacenters are capable of quickly responding to a load change, meaning the primary source of energy could be solar during the day and complement when there isn't enough energy.
But I haven't looked into where these datacenter are being placed, I'm assuming that although solar is cheap now, the surface needed would make the purchase of nearby land probably not worth it. These new categories of datacenters are becoming very energy dense...
but you can't trust especially hyperscalers with securely sealed HEUs in shipping containers
you're just incorrect. You probably missed 2 points :
- battery storage
- and in the article
"However, AI labs and some hyperscalers have relaxed those requirements as there is now a lower uptime tolerance applied to both inference and training, not just training. Many of Meta’s self-built AI datacenters, for example, target just two nines of uptime and forgo backup generators entirely, as detailed in our Industrials Model."
That's why we are getting clean, beautiful coal the likes of which nobodys ever seen before!
By that logic almost any kind of economic development is an ecological disaster because it uses power.
Economic development is great and I'm all for it, but it shouldn't be done at any cost. I quite like our atmosphere, ideally with less than 400ppm of CO2. We're finally seeing signs of economic development decoupling from primary energy consumption, there are positive signs.
I find that it's just a shame to be so close to energy abundance - through renewables - to only be set back by the installation of "temporary" gas turbines. The economic incentive is so great that delaying the datacenters in order to power them cleanly isn't even considered: there's no incentive to do so. This may sound a little alarmist, but if it's GW of capacity in 2028, it's TW in the mid 30s.
I'd much rather live in a heavily compute-constrained world for the next 5 years or so than in a world that keeps increasing its gas consumption.
There's just no way to force our incentive to align with preserving our climate, and it's making me extremely depressed.
By your logic we shouldn't question any development, at any cost, including if it contributes to making our world uninhabitable for billions of people within our life's timescale?
Welcome to European thought
And especially terrible for the communities that have to live with gas turbines or other local power generators as neighbors. Noise and air pollution constantly [1].
But fuck them, they are poor people so we don't care about them /s .
Additionally, people against data centers are accused of being paid by China [2]
[1] https://www.theguardian.com/environment/2026/feb/13/elon-mus... [2] https://fortune.com/2026/06/10/kevin-oleary-trump-administra...
It wasn't long ago that every city had a coal power station right in the center of it. Things aren't nearly as bad as you suggest.
Let's go back to have open air sewers while we are at it. I mean, if the goal is to bring back stuff that lowered life expectancy and reduce 50+ years of progress, I don't see why not. I really believe that as arguments go, yours is truly terrible.
Additionally, you proved my point that the majority really doesn't care about the folks whose quality of life worsen because someone built a datacenter close to their homes. Or the fact that turbines still pollute the air with NOx emissions (which cause lung cancer and other respiratory problems).
You are basically saying that it is not so bad, but (I guess) it is because you aren't close to a datacenter not depending on the grid. And I trust more hearing about live accounts of people being affected, than a "trust me bro, it is not as bad". Are the accounts cherry-picked? Maybe, but they are also supported by videos that substantiate what they are saying.
Your entire life is subsidized by those wealthier than you. Get off your high horse. Did you build your own house? Did you build your own car? Do you make your own power? What really has been your contribution to society.
High horse? Man, my country was under communist dictatorship and I didn't see a banana until I was 8 years old. However, despite that, I am aware of how lucky I got in life and I try to be mindful of those that are not as much.
Also, I don't believe one bit my life has been subsidized. Wealthy people don't throw money around if there is no profit to be made. At best they subsidize stuff to get you hooked before raising prices and raking the money in.
Additionally, I paid people to build my own house with money that I earned, and same for my car. And I now live in a country where we pay living wages and not force people to do 2-3 jobs to stay afloat. So I didn't exploit anyone as those people received months of salaries for them and their families thanks to my house.
The only thing I can concede to your argument is that things like my phone or laptop are cheap because people in the world are exploited while mining and in Foxconn like factories and for that I am trying to limit my tech footprint.
It's called progress. People like clean air and rivers that don't glow.
Sacrificing it all for making tech bros richer seems like a ridiculously bad play.
If their property is damaged they can sue. We have a system for this.
you can't put an ADU, yet, you can put in 200MW generating capacity.
This is a defeatist EU attitude that leads to the worse outcomes. Expensive power and a failing economy.
There is money now for a huge investment boom, which is what is happening in the US and China. With EU failing completely.
>There is money now for a huge investment boom
Not in western Europe, because they spent almost all their excess economic output over the past few decades on welfare/transfer payments.
Are you aware that public spending is a transfer from the public sector to the private, and that poor people spend approximately 100% of their income? Would you argue that stock buy-backs are a better use of public money?
At some point, probably in the 80s, the prophets of Mammon convinced policymakers that the liquidity of capital markets was the highest good, that no liquidity was enough, and that the financial class is inherently productive.
When we bailed out the banks in 2008, it amounted to a wealth transfer from the public purse to Jamie Dimon. That's your counterproductive welfare spending.
Texas had blackouts last year because free-market fundamentalism. Perhaps Texans should run their fridges on investment boom.
Public spending is dead unproductive capital. Giving poor people money just traps them into a cycle of poverty.
Spain also had blackouts.
Yes, I'm aware of free-market-fundamentalist dogma. Are you?
Yeah sure dogma. Capitalism is evil, yet it lifted more people out of poverty than anything in history.
Of the three economic powers, the one with the most social welfare and least capitalist is the one that never recovered from the 2008 crisis. Funny how that is. Strange when all the welfare gets spent in the economy.
The one with the unproductive capital buying back shares is building GW datacenters. The other with "productive" capital, is encouraging people not to buy ACs because the electric grid cant handle it and using energy is bad for the environment.
That was my point. We in the EU are destroying ourselves. The US is racing ahead and investing in all kinds of power generation technologies.
The difference is Europeans expect government to "fix" things. Americans expect companies to do that.
Eg, the disastrous energy policies all across Europe which have made price of energy insanely high with no hope of ever coming down (these high prices are all locked in with long term contracts).
America just builds lots of new power. Because of fewer market distorting policies the new power comes online at cheaper marginal rates so the price doesn't go up. And now Texas has some of the greener and cheapest energy in the West.
America has a lot of cheap fracking gas. Europe imports this gas from US in form of LNG at much higher prices.
The Russian gas is cheaper, but paid by Ukrainian blood.
Texas's CO2 per KWh of electricity (which is roughly equal to the US as a whole) at about 330 is equidistant between China's (470) and EU's (190) it is not close to the greenest in the West.
I am interested to see how this plays out and what will happen to potentially stranded assets. Not even talking about outdated chips but all these gas turbines. I appreciate what this push can do to the built out of grids etc.
Developing world will soak them in a heartbeat. Unreliable grid is one of the problems they have that slow their industrialization and development pace. If you are foreign investor - LNG terminal + turbines may be more expensive, but if you bypass some of the bribes and incompetence it may be worth it.
why are they burning gas when supposedly renewables are the cheapest option?
"data centres operate 24/7 with an industry standard of “five-9s” reliability—99.999% uptime, equivalent to less than 5.3 minutes of downtime per year. This level of reliability requires access to a stable and highly reliable source of electricity."
https://www.oxfordeconomics.com/resource/the-rising-challeng...
Many kinds of renewables are day/night, weather dependent, you don't get this level of reliability without a fossil fuel backup. If you already need to build a dedicated gas power plant to power your data center, then the decision to invest in an additional renewables power plant depends on the costs of the natural gas. (Additional capital expenses of the renewables power plant vs reduction in consumption of the natural gas).
[flagged]