IMO, stranded or overbuilt assets will become a huge problem in the next 10-20 years, but not from data centers, which will largely generate their own off grid power.
They will come from the wind and solar farms that have been vastly overbuilt during the decades of subsidies, only to be finally realized as a waste of money and not worth replacing as they wear out. The transmission lines and substations built out in no-mans' land will be underused and correspondingly inefficient. The cost of decommissioning renewables will leave many of the projects as abandoned eyesores, possibly for generations.
Agreed, and it's already happening with how they're treating coal retirements. It's just a matter of how bad it will get and whether data centers will exacerbate it.
AI - NetZero hypocrisy- the need for new generation for NetZero especially heating electrification is much larger and much more costly than AI. Heating is more costly due to its low load factor, its two times variability warm to cold years, and occurring only during the hours gas generation is expensive.
New England fights a 200MW data center behind the meter at the Millstone nuclear plant, but then plans to add 25,000 MW of heating load by 2050 (recently revised to 15,.000 as they've realized that heat pumps cost twice gas to operate).
And NetZero electrification jeopardizes our children's future. Increased computational capability is necessary for the survival of our existing businesses and for the businesses of the future. While electrification provides no new functions--just changes the energy type to one that typically doesn't work as well and is more costly.
AI is becoming a "cover-up" for the high costs that NetZero electrification is causing. Maine has added 130,000 heat pumps (800 MW?) and a federally funded $450M heat pump accelerator program to add 580,000 heat pumps (3600MW?) is underway. Recent (submitted December before Iran) bids for Jan 2027 Standard generation Service were very high: 32.5, 37.5, and 43.8 c/kWh for 10% full requirements slices. The heat pump customers don't pay the high costs they are imposing as the monthly prices are averaged, so the non-heat pump customer subsidizes the heat pump customer by $1.4 B per year for New England.
AI is also a cover-up for the massive amounts of generation and transmission that needs to be added for Net Zero: $150B for generation and $17-27B for transmission by 2050 for New England. .....And these numbers are likely low as they are based on national lab generation costs that are low today and forecasted to decline.
ERCOT is a great case study, and they are handling it the right way. Their large load requests sit at 439GW, over 2x vs December 2025. Maximum recent demand from the grid sits at 85GW. They have (are about to) put in place a whole new process for studying load in batches (Batch Zero) rather than sequentially. The process is designed to weed out speculative applications. How do they do that? Utilities in Texas recover the cost of building transmission through the electricity bill (generally), but they require the large load to backstop the spending on that infrastructure with a letter of credit. That mitigates the risk of socializing large capex to the retail rate base. Forecasts will take you only so far. Putting up private capital is the most honest measure.
Yes - I recall CREZ was $6.9B, recovery over 33 years. So that will be with us for a while. West Texas wind farms were grateful. Was that a good deal for the ratepayer?
No. Absolutely, no. There's a $.0013 per KWHr charge on my bill for every single KWHr to pay for those transmission lines. Actually an "ERCOT" charge, but I assume the bulk is for the lines. Even if wind is reaching 20% supply of Texas electricity, that means that every single KWHr of wind should have an additional $.06 per KWHr tacked onto the already inflated price.
This is another hidden cost of wind. Add in the tax funded subsidies, and the costs that "must take" force onto other generators, and wind costs at least twice what they charge for it.
Also consider the situation since 2008, which is about when 28 states implemented RPS (Renewable Portfolio Standards, strange how that happened in all those states simultaneously) the cost of nuclear and coal generated electricity has stayed about constant.
BUT...
The cost of natural gas has dropped from $8 - $13 per MMBTU to $2 - $5 per MMBTU. Ranges because the price always fluctuates.
The Operational Expense (OpEx) of a gas electric generator plant is 85% the cost of the gas. Texas generation is composed of more than 40% natural gas.
The cost of electricity in Texas should have declined by about 30% since 2008, not risen by 50%.
Whatever could be the cause of the vast rise in electricity prices, while the cost of our primary generator declines and others stay constant?
This may be the only thing ERCOT has gotten right in years. Oncor's large loads have over 200 GW of unstudied requests... how does anyone reconcile that - you are right about putting up capital it is the only way to make them honest. We have to remember they are pushing for 765 kV transmission across the state and need to justify that build-out. Once the PUC figures out how they are going to allocate costs maybe the ratepayers won't get stuck with this enormous bill. Wind, solar and datacenters as well as oil and gas would all like to socialize those costs all in the name of ESG. Nope it hasn't gone away yet!
Fair comment. The only thing that flatters the wind generation is the high-gas price comparison. You could describe this as an expensive toll road built at public expense to a destination that has lobbied hard for the bond to be issued and is given a guaranteed supply of subsidized tourists, in part to justify building it in the first place. How could that 3600 miles of line have been better deployed for ratepayers benefit?
A timely subject EBBs. I thought this was a key point in the discussion...
"The result was decades of bill impacts tied to capacity that was barely used or never used at all. If data center projections miss badly, we could be in for a repeat..."
We are already experiencing something like this. With renewables invading the grid with their inherently low capacity factors, the capacity factor of a broad range of generating, transmission, and storage assets end up with low usage (as measured by capacity factors). More capital expenditure: poorer utility of the expenditures to providing customer service.
With regard to data centers specifically, I attended an Industrial Siting Council (ISC) hearing yesterday that was mainly about hoped-for amendments to permits already granted by the ISC. These included slow-downs and extensions to estimated completion dates, plus changes or modifications to permit ownership. The slowdowns are ostensibly the result of supply chain issues, but it was pretty obvious from the discussions that financing issues and difficulty matching plans to actual customers (inability to procure PPAs for instance) are occuring as well.
At one point an applicant was asking to modify the permit to accommodate a customer demand that battery storage be increased and the applicant emphasized and re-iterated that none of this particular project would have any impact on residential or commercial rates (it appears that the local utility is involved to simply wheel power). They are getting ahead of public concerns...
Pawlenty……. Follow the money and find people suddenly finding themselves in support of things they know not to be feasible or true. Minnesota a formerly mining and manufacturing powerhouse. Now an expensive case study in socialism and its failures. Incredible to see how much power the Twin Cities can project over an otherwise normal state.
Even in little ol' Wyoming we find people in support of things that can't happen -- the thought of making money distorts peoples' thinking about the future and what is possible.
An urban area, with it's denizens holding ideas utterly at odds about the role of government, how to divvy-up wealth, and "justice" from folks in the remainder of the state is a troublesome issue for all states. Idaho to New York. It is at heart our political divide.
Watch he Film from the 1950's, called the Forbidden Planet. Letting the genie out of the bottle is never a good idea. Except for my trust in the Lord Jesus I would be worried but because he told us not to worry I can be free of it. Amen
I don't remember the film, but agree with your trust in the Lord. My comment (above) sounds negative, but that's just in regards to wind and solar. I'm joyfully looking for the Lord's return.
Very good analysis, the one potential solution is to impose take or pay contract clauses for the data centers. That would make them less likely to overstate their power needs.
Incremental data expansion. In addition to the big players building the huge data centers I suspect there will be incremental expansion of existing data centers by corporations not only for AI but for the increasing need for non-AI automation. The big consulting firms seem to be adding AI staff and forming alliances develop this market. AI (or more traditional non-AI analysis) needs the appropriate database which many corporations have in house and which they may not want to take the risk of sharing with the big providers.
I don't know why corporations will opt for AI and answers that "you should check as they may be wrong" when you could engineer a system to provide the 100% right answer.
In any event there will be more corporate computational capability needed, AI or not. Any idea how much?
Correction, the AEP data center tariff is 85% of the forecast energy demand, not 85% of the forecast energy. You may want to post a note with a correction. See my piece on the data center tariff for more information.
Duke and Dominion share another issue. Both serve in states (NC & VA) with legislated renewable generation requirements. Neither is on a reasonable path to meet the legislative requirements. Both have announced plans to add additional natural gas combined-cycle capacity and are meeting resistance. Data center developers have little interest in renewable plus storage generation to power their sites.
Great article. This is a critical question with a lot of diverse perspectives at the moment.
The landscape for large customer loads has evolved significantly over the past few years. 3-5 years ago large loads (data centers) received a very low rate with absolutely no strings attached. The biggest issue was that the load ramp almost always trailed the customer forecast.
Now, in addition to asking the existing customers to sign contracts, the new customers are asked to sign contracts with substantial protections for the utility and customers. They include long term contracts, minimum loads (take or pay), funding for new resources, advanced payment, and substantial collateral.
The end result is that regular customers should benefit from the large loads absorbing some of the fixed cost, and they should be protected if the large load customers fail / default.
This being said, there remains the risk that large load customers will fail over the next few years leaving utilities or customers will excess generation….at a time when there’s significant rate stress.
Despite the hype, AI and its data center crew haven't budged national economic performance by one hundredth of a per cent. What keeps the bubble inflated is what economists call rent-seeking. Better mousetraps have been built. This is fine as long as you're not a mouse.
One thing I don't get is the idea that data centers will be built all over the place. Building one at a retired power plant and running the plant for power could work most places. Why in the world would someone build a data center using the grid for power in a high electric rate location when ther are so many options with lower prices.
In Upstate New York there is talk of plunking one in a suburb of Syracuse. What with RGGI and the Climate Act guarantees for even higher prices that makes no sense to me.
IMO, stranded or overbuilt assets will become a huge problem in the next 10-20 years, but not from data centers, which will largely generate their own off grid power.
They will come from the wind and solar farms that have been vastly overbuilt during the decades of subsidies, only to be finally realized as a waste of money and not worth replacing as they wear out. The transmission lines and substations built out in no-mans' land will be underused and correspondingly inefficient. The cost of decommissioning renewables will leave many of the projects as abandoned eyesores, possibly for generations.
Agreed, and it's already happening with how they're treating coal retirements. It's just a matter of how bad it will get and whether data centers will exacerbate it.
AI - NetZero hypocrisy- the need for new generation for NetZero especially heating electrification is much larger and much more costly than AI. Heating is more costly due to its low load factor, its two times variability warm to cold years, and occurring only during the hours gas generation is expensive.
New England fights a 200MW data center behind the meter at the Millstone nuclear plant, but then plans to add 25,000 MW of heating load by 2050 (recently revised to 15,.000 as they've realized that heat pumps cost twice gas to operate).
And NetZero electrification jeopardizes our children's future. Increased computational capability is necessary for the survival of our existing businesses and for the businesses of the future. While electrification provides no new functions--just changes the energy type to one that typically doesn't work as well and is more costly.
AI is becoming a "cover-up" for the high costs that NetZero electrification is causing. Maine has added 130,000 heat pumps (800 MW?) and a federally funded $450M heat pump accelerator program to add 580,000 heat pumps (3600MW?) is underway. Recent (submitted December before Iran) bids for Jan 2027 Standard generation Service were very high: 32.5, 37.5, and 43.8 c/kWh for 10% full requirements slices. The heat pump customers don't pay the high costs they are imposing as the monthly prices are averaged, so the non-heat pump customer subsidizes the heat pump customer by $1.4 B per year for New England.
AI is also a cover-up for the massive amounts of generation and transmission that needs to be added for Net Zero: $150B for generation and $17-27B for transmission by 2050 for New England. .....And these numbers are likely low as they are based on national lab generation costs that are low today and forecasted to decline.
Great points. Thanks for sharing, bob
Those numbers also do not include the storage ultimately required to make the renewables dispatchable.
ERCOT is a great case study, and they are handling it the right way. Their large load requests sit at 439GW, over 2x vs December 2025. Maximum recent demand from the grid sits at 85GW. They have (are about to) put in place a whole new process for studying load in batches (Batch Zero) rather than sequentially. The process is designed to weed out speculative applications. How do they do that? Utilities in Texas recover the cost of building transmission through the electricity bill (generally), but they require the large load to backstop the spending on that infrastructure with a letter of credit. That mitigates the risk of socializing large capex to the retail rate base. Forecasts will take you only so far. Putting up private capital is the most honest measure.
Yes - I recall CREZ was $6.9B, recovery over 33 years. So that will be with us for a while. West Texas wind farms were grateful. Was that a good deal for the ratepayer?
No. Absolutely, no. There's a $.0013 per KWHr charge on my bill for every single KWHr to pay for those transmission lines. Actually an "ERCOT" charge, but I assume the bulk is for the lines. Even if wind is reaching 20% supply of Texas electricity, that means that every single KWHr of wind should have an additional $.06 per KWHr tacked onto the already inflated price.
This is another hidden cost of wind. Add in the tax funded subsidies, and the costs that "must take" force onto other generators, and wind costs at least twice what they charge for it.
Also consider the situation since 2008, which is about when 28 states implemented RPS (Renewable Portfolio Standards, strange how that happened in all those states simultaneously) the cost of nuclear and coal generated electricity has stayed about constant.
BUT...
The cost of natural gas has dropped from $8 - $13 per MMBTU to $2 - $5 per MMBTU. Ranges because the price always fluctuates.
The Operational Expense (OpEx) of a gas electric generator plant is 85% the cost of the gas. Texas generation is composed of more than 40% natural gas.
The cost of electricity in Texas should have declined by about 30% since 2008, not risen by 50%.
Whatever could be the cause of the vast rise in electricity prices, while the cost of our primary generator declines and others stay constant?
Then after typing all that out, I see this:
https://comptroller.texas.gov/economy/economic-data/energy/2023/wind-snap.php
Time to vote Glen Hegar out of office if he's still there. He's clearly incapable of doing a competent analysis.
This may be the only thing ERCOT has gotten right in years. Oncor's large loads have over 200 GW of unstudied requests... how does anyone reconcile that - you are right about putting up capital it is the only way to make them honest. We have to remember they are pushing for 765 kV transmission across the state and need to justify that build-out. Once the PUC figures out how they are going to allocate costs maybe the ratepayers won't get stuck with this enormous bill. Wind, solar and datacenters as well as oil and gas would all like to socialize those costs all in the name of ESG. Nope it hasn't gone away yet!
Fair comment. The only thing that flatters the wind generation is the high-gas price comparison. You could describe this as an expensive toll road built at public expense to a destination that has lobbied hard for the bond to be issued and is given a guaranteed supply of subsidized tourists, in part to justify building it in the first place. How could that 3600 miles of line have been better deployed for ratepayers benefit?
A timely subject EBBs. I thought this was a key point in the discussion...
"The result was decades of bill impacts tied to capacity that was barely used or never used at all. If data center projections miss badly, we could be in for a repeat..."
We are already experiencing something like this. With renewables invading the grid with their inherently low capacity factors, the capacity factor of a broad range of generating, transmission, and storage assets end up with low usage (as measured by capacity factors). More capital expenditure: poorer utility of the expenditures to providing customer service.
With regard to data centers specifically, I attended an Industrial Siting Council (ISC) hearing yesterday that was mainly about hoped-for amendments to permits already granted by the ISC. These included slow-downs and extensions to estimated completion dates, plus changes or modifications to permit ownership. The slowdowns are ostensibly the result of supply chain issues, but it was pretty obvious from the discussions that financing issues and difficulty matching plans to actual customers (inability to procure PPAs for instance) are occuring as well.
At one point an applicant was asking to modify the permit to accommodate a customer demand that battery storage be increased and the applicant emphasized and re-iterated that none of this particular project would have any impact on residential or commercial rates (it appears that the local utility is involved to simply wheel power). They are getting ahead of public concerns...
Pawlenty……. Follow the money and find people suddenly finding themselves in support of things they know not to be feasible or true. Minnesota a formerly mining and manufacturing powerhouse. Now an expensive case study in socialism and its failures. Incredible to see how much power the Twin Cities can project over an otherwise normal state.
Even in little ol' Wyoming we find people in support of things that can't happen -- the thought of making money distorts peoples' thinking about the future and what is possible.
An urban area, with it's denizens holding ideas utterly at odds about the role of government, how to divvy-up wealth, and "justice" from folks in the remainder of the state is a troublesome issue for all states. Idaho to New York. It is at heart our political divide.
It’s really sad. Love Wyoming BTW! The Cowboy State!
Watch he Film from the 1950's, called the Forbidden Planet. Letting the genie out of the bottle is never a good idea. Except for my trust in the Lord Jesus I would be worried but because he told us not to worry I can be free of it. Amen
I don't remember the film, but agree with your trust in the Lord. My comment (above) sounds negative, but that's just in regards to wind and solar. I'm joyfully looking for the Lord's return.
Very good analysis, the one potential solution is to impose take or pay contract clauses for the data centers. That would make them less likely to overstate their power needs.
Incremental data expansion. In addition to the big players building the huge data centers I suspect there will be incremental expansion of existing data centers by corporations not only for AI but for the increasing need for non-AI automation. The big consulting firms seem to be adding AI staff and forming alliances develop this market. AI (or more traditional non-AI analysis) needs the appropriate database which many corporations have in house and which they may not want to take the risk of sharing with the big providers.
I don't know why corporations will opt for AI and answers that "you should check as they may be wrong" when you could engineer a system to provide the 100% right answer.
In any event there will be more corporate computational capability needed, AI or not. Any idea how much?
Correction, the AEP data center tariff is 85% of the forecast energy demand, not 85% of the forecast energy. You may want to post a note with a correction. See my piece on the data center tariff for more information.
Duke and Dominion share another issue. Both serve in states (NC & VA) with legislated renewable generation requirements. Neither is on a reasonable path to meet the legislative requirements. Both have announced plans to add additional natural gas combined-cycle capacity and are meeting resistance. Data center developers have little interest in renewable plus storage generation to power their sites.
Great article. This is a critical question with a lot of diverse perspectives at the moment.
The landscape for large customer loads has evolved significantly over the past few years. 3-5 years ago large loads (data centers) received a very low rate with absolutely no strings attached. The biggest issue was that the load ramp almost always trailed the customer forecast.
Now, in addition to asking the existing customers to sign contracts, the new customers are asked to sign contracts with substantial protections for the utility and customers. They include long term contracts, minimum loads (take or pay), funding for new resources, advanced payment, and substantial collateral.
The end result is that regular customers should benefit from the large loads absorbing some of the fixed cost, and they should be protected if the large load customers fail / default.
This being said, there remains the risk that large load customers will fail over the next few years leaving utilities or customers will excess generation….at a time when there’s significant rate stress.
" Predictions are hard, especially about the future.", Yogi Berra, American Philosopher
Your continuing practical environmentalism is a welcomed relief from the Greenwishing gone mad.
Despite the hype, AI and its data center crew haven't budged national economic performance by one hundredth of a per cent. What keeps the bubble inflated is what economists call rent-seeking. Better mousetraps have been built. This is fine as long as you're not a mouse.
Thank you this was very informative.
One thing I don't get is the idea that data centers will be built all over the place. Building one at a retired power plant and running the plant for power could work most places. Why in the world would someone build a data center using the grid for power in a high electric rate location when ther are so many options with lower prices.
In Upstate New York there is talk of plunking one in a suburb of Syracuse. What with RGGI and the Climate Act guarantees for even higher prices that makes no sense to me.
https://rossboulton1.substack.com/p/the-cloud-that-doesnt-make-rain?r=2leuaj&utm_campaign=post-expanded-share&utm_medium=web