It's important to ground the increase in raw numbers.
The total revenue for electricity generation was $514b in 2024. So this was a 4-5% increase in costs. And if it is being invested in better generation and our aging infrastructure, that seems fine.
This just isn't true. On balance, data centers are turning out to be more like the "anchor tenant" of the power grid, financing improvements for everyone.
> But what if the power company needs to upgrade the substation to handle the increased needs of the data center? Or secure additional sources of electricity? In these cases, the investments are part of the electricity grid that everyone uses. These costs will likely be shared among all customers.
Okay but this is a policy choice. It doesn’t have to be that way.
Where I live, if a developer wants to build a subdivision, they pay for the water and sewer lines. They pay to connect those lines to the city infrastructure. If the city infrastructure needs to be upgraded to handle the new volumes, they pay for at least a proportionate share of that too. The ongoing maintenance becomes part of the city's budget eventually, but not the costs of the build out.
All those costs go into the price of the houses built there.
And this is also part of why building "affordable" houses rarely happens. All the infrastructure costs the same whether the houses cost $100K or $1 million.
That's smart. Do they do it with roads also? That's a big one near me - developers buying hundred acre farms on unpainted 2 lane country roads and jamming in 2000 houses. Then inevitably the road becomes unusable until the city or county gets around to addressing it.
Always wondered why the county didn't require the road work, or money for it, up front.
That's negligence on the part of the county/township. Around here, every new development is required to pay for the traffic and utilities improvements that will be required once the thing is built. (And all the engineering that goes into figuring out the impact in the first place.)
In my hometown in Idaho in the 1990s and 2000s yes, this includes access roads and improvements to the surrounding area. A car dealership and Wal Mart both paid for road improvements and traffic lights as part of development.
The “affordable” housing thing seems like such a misdirection. I can’t help but daydream that some moneyed interest somewhere fans those flames, as it looks like a dead end that can absorb endless fervor.
You know what you do if you want an affordable car? You buy used. I think most people understand Ford is never going to build another a car that costs $10k brand new, and the last new car near that price barely sold because it was so stripped.
>The “affordable” housing thing seems like such a misdirection. I can’t help but daydream that some moneyed interest somewhere fans those flames, as it looks like a dead end that can absorb endless fervor.
Do you really need a conspiracy by "moneyed interest" when the general public is perfectly happy to support similarly bad policies like rent control?
In the US this isn't even true. The costs of grid upgrades are paid new capacity up front, which actually means there is a queue for grid connections twice the US total production, with no costs to ongoing customers.
I find it hilarious nobody ever mentions texas w/r/t this issue. They already required new Large Load Interconect Studies- and paying the cost of new grid infra. And, the hufe influx of study reqiests lead to new laws introducing batching.
I will note- the actual generation is left to market economics. I have much more faith in that working out equitably than regulating the grid. Even so, they've had significant consumer level grod connection fee increases- which I think reflects the end of various easy houshold effeciency gains (eg: incandecent -> CFL -> LED, P4 -> 14th gen, and HVAC. mainly lighting) and privatized profits more than anything.
> "concluded that expected power demand from data centers was _a_ primary reason for $23 billion in customer price increases "
Also, it's weird that he describes PJM as "the organization that monitors the PJM market" when they describe themselves as "a regional transmission organization (RTO) that coordinates the movement of wholesale electricity" [0]. So are they monitors of the market or are they the market themselves?
I don't know... maybe I'm being picky, but the article just seems off. The whole bit about how data centers could maybe game the system by using less power during peak times also doesn't make sense - that's when they also have the highest demand. Pointing to cryptominers just makes me think he doesn't get what they do, which is basically arbitrage. Of course they stop when power costs go up, it eats up all of their profits and they can simply start back up when the costs go down.
Right, you can't point to a year where price increases didn't happen, so "a primary reason" becomes meaningless. It was going to happen for other reasons if it didn't happen for this one.
by my research, everything has scaled almost linearly with population. But- there was a huge, undeniable, dip in cost of a kw/h of power when lighting started focusing on CFLs and LEDs. Grid expansion stalled, and now people are upset since it started growing again
I'm really tired of reading comments like this"this is a policy choice" when these companies are straight up bribing politicians or exploiting their conflicts of interest to make this their policy. The public doesn't have a choice when all our leaders are bought and paid for by oligarchs and their businesses.
Comments like this are even more tiring. Everyone knows that there are moneyed interests in politics. But decisions on these issues are made at a local level and so are the most addressable by local political actions.
Exactly. Over in Virginia 37 datacenters use close to 3GW of electricity. The power utilities and overarching transmitters are looking at projects to ship some of that electricity (which requires ~700kV transmission). Two projects in the works are at $18B between them.
$18B to provide redundancy and not have to require schools and local government to limit electricity use and provide a bit more slack in the powergrid is a burden that all the users get to share. Lucky them. Yes, all users benefit, but lucky break for those datacenters, getting all that redundancy for power, without a $500M/ea bill.
That's great to hear that we have at least somewhat decent voltages. I thought America was still sort of piddling around with mostly 500kV and under power distribution.
It's not a huge power multiplier (P = VI, so linear), but in principle I do love the idea that if you are going to have these massive transmissions lines we ought use the conductor well, at good high voltages.
Isn't this the classic overcapacity leads to lower prices that also represses investments that would increase capacity. But those lower prices also stimulate new demand that lead to higher prices...which then motivate investments that increase capacity?
Perhaps I'm just spoiled because I live in the PNW, where are best use for overcapacity was to ship power off to California. But in the past, cheap hydro attracted aluminum production that then attracted also attracted a whole airplane production industry.
I think most people are just debating whether the extra demand generated by AI is worth it, they weren't necessarily debating the same thing when it came aluminum or airplane production (albeit in the 1930s).
Isn't almost all of the datacenter build out for inference, rather than training? If so, what's the issue? If the electricity demand is coming from actual use, then why are people getting mad at datacenters or AI companies, rather than the actual people driving the demand? It's like getting mad at Amazon for how much they increased fuel prices, which they surely did, given all the fuel that their trucks/planes burn. Or getting mad at some global food conglomerate for making açaí berries[1] more expensive, but there's a global craze for them and the conglomerate is just catering to that demand.
[1] or whatever other "superfood" that explodes in popularity
What is the increase in jobs/GDP for those communities that have paid more in electricity? A lot of these data centers are built in places with declining population and zero economic prospects for locals within their communities so they're a huge boon.
None? There's like 12 jobs per datacenter and the revenue goes to the Valley, mostly. The utilities get some profit but I doubt that money stays very local either.
Data centers should at the very least build their own renewable energy generation. It would set the right incentives in place and encourage investment in clean energy generation solutions. It would also present a very compelling problem to direct all that new AI compute towards solving.
Isn't this the same as saying "utility regulators delaying connecting new power to the grid hiked electricity prices on the public by $23B?"
When my apples are expensive, I don't generally grumble about all the demand from pie makers. If they demand more apples, new suppliers should come in to restore the price, right?
Every tech company building out datacenters would pay in a heartbeat to add commensurate power to the grid. The cost is not an objection.
The problem is, there are insane and dumb regulatory barriers to adding power plants or interconnects. THESE ARE THE SAME PROBLEMS FACTORIES FACE WHEN RESHORING PRODUCTION, you should treat datacenters as the face of reindustrialization. Instead of complaining about using resources, we need to focus on solving our inability to provide infrastructure needed to support economic growth.
Given that datacenters seem to much more manpower efficient, that seems like a poor trade, to the point it might not be worth viewing them similarly. I'm seeing info reporting about 1 job per 5000 square feet of a datacenter after completion, and 1 job per 800-1500 square feet for a factory, depending on type.
More jobs is good, but if we're going to look at this through the lens of industry returning, it's a lousy return, even before factoring in that we probably lost a lot more factories than we're gaining datacenters.
The fact datacenters are standing up extremely expensive on-site natural gas turbines due to the immense delays for interconnect.
Pretty much no facility operator wants to also pay for and operate their own power plant. They are small and expensive to operate compared to combined cycle natural gas or other sources, and lack of access to a wider grid means even more additional expenses like a additional on-site redundancy.
If they could simply pay for grid interconnect that is ready by the time facility construction is completed they would do so, in the vast majority of cases.
The money is in getting things online ASAP. Builders are effectively throwing unlimited buckets of money at all aspects of these builds at people who can get shit done fast. Power interconnect would be no exception.
Elon (in)famously put gas turbines on pickup trucks to power the Colossus 1/2 datacenters because connecting to the grid was incredibly slow and there was limited supply. And got sued to remove them.
Absent regulation, every operator would happily do the same thing to make the problem go away.
It's important to ground the increase in raw numbers.
The total revenue for electricity generation was $514b in 2024. So this was a 4-5% increase in costs. And if it is being invested in better generation and our aging infrastructure, that seems fine.
This just isn't true. On balance, data centers are turning out to be more like the "anchor tenant" of the power grid, financing improvements for everyone.
Overview article with links to actual studies: https://cityjournal.substack.com/p/data-centers-arent-raisin...
> But what if the power company needs to upgrade the substation to handle the increased needs of the data center? Or secure additional sources of electricity? In these cases, the investments are part of the electricity grid that everyone uses. These costs will likely be shared among all customers.
Okay but this is a policy choice. It doesn’t have to be that way.
Where I live, if a developer wants to build a subdivision, they pay for the water and sewer lines. They pay to connect those lines to the city infrastructure. If the city infrastructure needs to be upgraded to handle the new volumes, they pay for at least a proportionate share of that too. The ongoing maintenance becomes part of the city's budget eventually, but not the costs of the build out.
All those costs go into the price of the houses built there.
And this is also part of why building "affordable" houses rarely happens. All the infrastructure costs the same whether the houses cost $100K or $1 million.
That's smart. Do they do it with roads also? That's a big one near me - developers buying hundred acre farms on unpainted 2 lane country roads and jamming in 2000 houses. Then inevitably the road becomes unusable until the city or county gets around to addressing it.
Always wondered why the county didn't require the road work, or money for it, up front.
That's negligence on the part of the county/township. Around here, every new development is required to pay for the traffic and utilities improvements that will be required once the thing is built. (And all the engineering that goes into figuring out the impact in the first place.)
In my hometown in Idaho in the 1990s and 2000s yes, this includes access roads and improvements to the surrounding area. A car dealership and Wal Mart both paid for road improvements and traffic lights as part of development.
The “affordable” housing thing seems like such a misdirection. I can’t help but daydream that some moneyed interest somewhere fans those flames, as it looks like a dead end that can absorb endless fervor.
You know what you do if you want an affordable car? You buy used. I think most people understand Ford is never going to build another a car that costs $10k brand new, and the last new car near that price barely sold because it was so stripped.
>The “affordable” housing thing seems like such a misdirection. I can’t help but daydream that some moneyed interest somewhere fans those flames, as it looks like a dead end that can absorb endless fervor.
Do you really need a conspiracy by "moneyed interest" when the general public is perfectly happy to support similarly bad policies like rent control?
It's almost like "affordable hosuing" are the vacancies that the people moving into shiny new homes leave behind.
In the US this isn't even true. The costs of grid upgrades are paid new capacity up front, which actually means there is a queue for grid connections twice the US total production, with no costs to ongoing customers.
Still waiting for canadian cell phone bills to come down after all those "upgrades" that supposedly happened.
Build more power!
Precisely this. Incredibly annoying headline
I find it hilarious nobody ever mentions texas w/r/t this issue. They already required new Large Load Interconect Studies- and paying the cost of new grid infra. And, the hufe influx of study reqiests lead to new laws introducing batching.
I will note- the actual generation is left to market economics. I have much more faith in that working out equitably than regulating the grid. Even so, they've had significant consumer level grod connection fee increases- which I think reflects the end of various easy houshold effeciency gains (eg: incandecent -> CFL -> LED, P4 -> 14th gen, and HVAC. mainly lighting) and privatized profits more than anything.
It doesn't even match what's in the article.
> "concluded that expected power demand from data centers was _a_ primary reason for $23 billion in customer price increases "
Also, it's weird that he describes PJM as "the organization that monitors the PJM market" when they describe themselves as "a regional transmission organization (RTO) that coordinates the movement of wholesale electricity" [0]. So are they monitors of the market or are they the market themselves?
I don't know... maybe I'm being picky, but the article just seems off. The whole bit about how data centers could maybe game the system by using less power during peak times also doesn't make sense - that's when they also have the highest demand. Pointing to cryptominers just makes me think he doesn't get what they do, which is basically arbitrage. Of course they stop when power costs go up, it eats up all of their profits and they can simply start back up when the costs go down.
[0] https://www.pjm.com/about-pjm
Right, you can't point to a year where price increases didn't happen, so "a primary reason" becomes meaningless. It was going to happen for other reasons if it didn't happen for this one.
by my research, everything has scaled almost linearly with population. But- there was a huge, undeniable, dip in cost of a kw/h of power when lighting started focusing on CFLs and LEDs. Grid expansion stalled, and now people are upset since it started growing again
I'm really tired of reading comments like this"this is a policy choice" when these companies are straight up bribing politicians or exploiting their conflicts of interest to make this their policy. The public doesn't have a choice when all our leaders are bought and paid for by oligarchs and their businesses.
Comments like this are even more tiring. Everyone knows that there are moneyed interests in politics. But decisions on these issues are made at a local level and so are the most addressable by local political actions.
Exactly. Over in Virginia 37 datacenters use close to 3GW of electricity. The power utilities and overarching transmitters are looking at projects to ship some of that electricity (which requires ~700kV transmission). Two projects in the works are at $18B between them.
$18B to provide redundancy and not have to require schools and local government to limit electricity use and provide a bit more slack in the powergrid is a burden that all the users get to share. Lucky them. Yes, all users benefit, but lucky break for those datacenters, getting all that redundancy for power, without a $500M/ea bill.
That's great to hear that we have at least somewhat decent voltages. I thought America was still sort of piddling around with mostly 500kV and under power distribution.
It's not a huge power multiplier (P = VI, so linear), but in principle I do love the idea that if you are going to have these massive transmissions lines we ought use the conductor well, at good high voltages.
China has been doing 1MV and 1.1MV lines for a while now, which is so excellent to see. https://www.bbc.co.uk/future/article/20241113-will-chinas-ul... https://en.wikipedia.org/wiki/Ultra-high-voltage_electricity...
Isn't this the classic overcapacity leads to lower prices that also represses investments that would increase capacity. But those lower prices also stimulate new demand that lead to higher prices...which then motivate investments that increase capacity?
Perhaps I'm just spoiled because I live in the PNW, where are best use for overcapacity was to ship power off to California. But in the past, cheap hydro attracted aluminum production that then attracted also attracted a whole airplane production industry.
I think most people are just debating whether the extra demand generated by AI is worth it, they weren't necessarily debating the same thing when it came aluminum or airplane production (albeit in the 1930s).
Isn't almost all of the datacenter build out for inference, rather than training? If so, what's the issue? If the electricity demand is coming from actual use, then why are people getting mad at datacenters or AI companies, rather than the actual people driving the demand? It's like getting mad at Amazon for how much they increased fuel prices, which they surely did, given all the fuel that their trucks/planes burn. Or getting mad at some global food conglomerate for making açaí berries[1] more expensive, but there's a global craze for them and the conglomerate is just catering to that demand.
[1] or whatever other "superfood" that explodes in popularity
What is the increase in jobs/GDP for those communities that have paid more in electricity? A lot of these data centers are built in places with declining population and zero economic prospects for locals within their communities so they're a huge boon.
None? There's like 12 jobs per datacenter and the revenue goes to the Valley, mostly. The utilities get some profit but I doubt that money stays very local either.
Data centers should at the very least build their own renewable energy generation. It would set the right incentives in place and encourage investment in clean energy generation solutions. It would also present a very compelling problem to direct all that new AI compute towards solving.
Failures to allow faster generation/hookup rollout have suppressed supply increase relative to demand increase
Isn't this the same as saying "utility regulators delaying connecting new power to the grid hiked electricity prices on the public by $23B?"
When my apples are expensive, I don't generally grumble about all the demand from pie makers. If they demand more apples, new suppliers should come in to restore the price, right?
Do you need apples to turn on your lights? Was there a sudden new influx of pie makers? Is growing apples a government granted monopoly business?
If the answer to these questions was yes, yes, and yes then I think you would grumble.
Every tech company building out datacenters would pay in a heartbeat to add commensurate power to the grid. The cost is not an objection.
The problem is, there are insane and dumb regulatory barriers to adding power plants or interconnects. THESE ARE THE SAME PROBLEMS FACTORIES FACE WHEN RESHORING PRODUCTION, you should treat datacenters as the face of reindustrialization. Instead of complaining about using resources, we need to focus on solving our inability to provide infrastructure needed to support economic growth.
Given that datacenters seem to much more manpower efficient, that seems like a poor trade, to the point it might not be worth viewing them similarly. I'm seeing info reporting about 1 job per 5000 square feet of a datacenter after completion, and 1 job per 800-1500 square feet for a factory, depending on type.
More jobs is good, but if we're going to look at this through the lens of industry returning, it's a lousy return, even before factoring in that we probably lost a lot more factories than we're gaining datacenters.
> Every tech company building out datacenters would pay in a heartbeat to add commensurate power to the grid. The cost is not an objection.
Yeah. I'm going to need a source on this claim.
The fact datacenters are standing up extremely expensive on-site natural gas turbines due to the immense delays for interconnect.
Pretty much no facility operator wants to also pay for and operate their own power plant. They are small and expensive to operate compared to combined cycle natural gas or other sources, and lack of access to a wider grid means even more additional expenses like a additional on-site redundancy.
If they could simply pay for grid interconnect that is ready by the time facility construction is completed they would do so, in the vast majority of cases.
The money is in getting things online ASAP. Builders are effectively throwing unlimited buckets of money at all aspects of these builds at people who can get shit done fast. Power interconnect would be no exception.
Elon (in)famously put gas turbines on pickup trucks to power the Colossus 1/2 datacenters because connecting to the grid was incredibly slow and there was limited supply. And got sued to remove them.
Absent regulation, every operator would happily do the same thing to make the problem go away.
Maybe what they're doing in Oregon with POWER Act hikes on data centers is the way:
Oregon approves PGE’s 29.7% rate hike for data centers under landmark law
https://www.opb.org/article/2026/07/07/oregon-data-center-ge...
https://archive.ph/Ypq0o
according to my wife we paid about 1/2 of that :)
"small price to pay for Massive AI Gains"
FYI to others: the above is not a quote from the article.