Essay by Eric Worrall
First published JoNova; I wonder how many fossil fuel plants and nuclear reactors are required to service 99GW of new energy demand?
Texas Needs Equivalent of 30 Reactors to Meet Data Center Power Demand
Demand on the Texas power grid is expected to expand so immensely that it would take the equivalent of adding 30 nuclear plants’ worth of electricity by 2030 to meet the needs.
Author of the article:
Bloomberg News
Naureen S. Malik…
The data centers “present a reliability risk to the Ercot system,” said Springer, who spoke on a panel at Infocast’s ERCOT Market Summit in Austin this week.
“We’ve never existed in a place where large industrial loads can really impact the reliability of the grid, and now we are stepping into that world.”
Risk of Grid Stress
Ercot said it’s gotten requests equal to 99 gigawatts for new connections to the grid from big power users, including data centers, bitcoin miners and hydrogen producers, according to an internal grid presentation Thursday. That’s up from 40.8 gigawatts last March.
…
Read more: https://financialpost.com/pmn/business-pmn/texas-needs-equivalent-of-30-reactors-to-meet-data-center-power-demand
It’s not just Texas. Across the world, new nuclear and fossil fuel capacity is being frantically constructed, by companies and governments who are frightened of falling behind.
No state is immune to this pressure. My prediction, even über green states like California will shortly cave and permit providing the gigawatts of cheap, dispatchable energy big tech requires to stay in the game, from any source available – fossil fuel, nuclear, imported, whatever they can get their hands on.
The USA isn’t the only player in town. China is going all in on artificial intelligence. Their disadvantage of being subject to sanctions is more than balanced by their enormous low cost computer chip industry and their energy surplus.
China announces high-tech fund to grow AI, emerging industries
By Nectar Gan and Juliana Liu, CNN
Published 4:30 AM EST, Thu March 6, 2025Hong KongCNN —
Fresh off the global success of DeepSeek’s latest artificial intelligence reasoning model, China’s top economic officials have vowed to set up a state-backed fund to support technological innovation.
The “state venture capital guidance fund” will focus on cutting-edge fields such as artificial intelligence, quantum technology and hydrogen energy storage, Zheng Shanjie, head of China’s state economic planner, told reporters Thursday on the sidelines of the annual gatherings of China’s rubber-stamp national legislature and advisory body.
The fund is expected to attract nearly 1 trillion yuan ($138 billion) in capital over 20 years from local governments and the private sector, added Zheng, chairman of the National Development and Reform Commission.
…
Read more: https://edition.cnn.com/2025/03/06/tech/china-state-venture-capital-guidance-fund-intl-hnk/index.html
India is also going big on AI;
India’s US$1.25 billion push to power AI
National funding to enhance computing infrastructure, deep-tech startups and data platform
India is ramping up its artificial intelligence (AI) ecosystem with a US$1.25 billion (Rs 10,372 crore) investment to expand its computing infrastructure and develop an accessible data repository.
Through public-private partnerships, IndiaAI Mission plans to set up a computing capacity or ‘compute’ of 10,000 or more Graphics Processing Units (GPUs). Over the next five years, it will build a platform stacked with non-personal data, such as environmental data, for Indian startups and researchers.
…
India’s AI market, growing at a compounded annual growth rate of 25-35% is expected to touch $US17 billion by 2027, fuelled by a growing AI workforce and an increase in generative AI investments.
…
Read more: https://www.nature.com/articles/d44151-024-00035-5
Europe is showing a flicker of interest, though they are so far behind in the cheap energy game, it is doubtful if they will ever catch up. France is the only European nation which appears to be taking AI seriously.
France tempts AI firms with its nuclear electricity
Tuesday, 11 February 2025
UK-based AI cloud provider Fluidstack has signed a memorandum of understanding with the French government to construct one of the world’s largest decarbonised AI supercomputers in France. Meanwhile, utility EDF has identified four sites on its own land that it will offer for data centres.
Announced at the AI Action Summit in Paris, under the leadership of President Emmanuel Macron, the memorandum of understanding was signed by the French Minister of Economy, Finance, and Industrial and Digital Sovereignty, Eric Lombard, the French Minister for Industry and Energy, Marc Feracci, and Fluidstack co-founder and President, César Maklary.
Read more: https://www.world-nuclear-news.org/articles/france-tempts-ai-firms-with-its-nuclear-electricity
…
If you are interested in why AI requires so much energy, this article provides some answers – and you get to play with a real artificial intelligence.
There is no chance the AI push will fizzle. Certainly the current wave contains a lot of hype, in the next few years we shall see the usual bankruptcies as companies riding the wave but with nothing substantial to offer run out of shareholder’s cash. But the underlying technology is sound. Companies are already making enormous profits using artificial intelligence to perform accelerated drug discovery and repurposing, intelligence analysis, semantic analysis and collation of news and events, and bringing greater than human capabilities to forward planning, strategy and scientific research.
At least one political leader gets it.
In this race, there is no prize for second place.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Texas has a case of political cowardice, and is going for an “all of the above” approach. Further investments in wind and solar are useless, but have a constituency.
Currently Texas wind power can produce 38 Gigawatts if the wind is blowing and generally carries half of the States load for approximately 50% of the year. Wind is here to stay in Texas and is a considerable asset that has been tuned to compliment the standby gas turbine systems. When the wind is blowing the “finite” natural resource is being preserved.
Nonsense. The cost of running two systems when one would do is huge. Nor is there value in “preserving” gas.
Nonsense, you made a statement without any facts.
It happens a lot, congratulations.
LT3, I just looked up the ERCOT wind capacity factor. Last year, 33.5%. That means two thirds of the time, that capacity is supplied by natgas. That means two systems, one by definition underutilized by 1/3. Facts.
thanks rud, i was trying to find similar information but you beat me to it.
Sounds about right, pretty much how it should be, and does not negate any of my statements. They are not putting up too many windmills right now, but they will continue to put more, because it works here.
“The cost of running two systems when one would do is huge. Nor is there value in “preserving” gas.”
Does anyone support this statement. If you do you may as well back it up with “They didn’t leave me any Buffalo”.
They (Texas windmills) have been operating in the black for decades, we depend on them. You people bitch and complain and put any failure under a microscope and fail to understand that a renewable natural resource that supplies half of the Texas (the number 15th economy in the world) grid 30+% of the time is not looked upon as a failure by any civic leader, capitalist, engineer banker, etc…
Only whining people that are only capable of engaging in a conversation if there is the prospect of slinging mud at a topic that is tied to a green initiative.
Worse, Wind Turbines consume nearly 50% of the power they generate. and this “house” load is taken from the GRID, NOT the output power of the generated power. Even Worse WORSE, A Wind Turbine just like a NPP, NatGas, Coal power plant needs, requires grid power to startup from a cold position. No GRID, NO START. That is why NPPs have Diesel Generators.
He presented as many “facts’ as you did. The only difference is that his facts actually exist.
LT here are the facts. You are correct, Texas has about 38,000 KW of installed wind capacity. That said, your position that it is a good thing is not an intelligent position. In 2023, of the possible 8,760 hours in the year, 1,286 (14.68%) had a demand of 65,000 MW or more. Almost 60% of those hours, wind was providing less than 25.2% of its installed capacity. In other words, it was producing less than 10,000 KW. Almost 10% of the hours it was producing less than 3,900 (10%) of its installed capacity. You can download the spreadsheet from ERCOT that shows installed capacity, load demand, etc. Texas approach to all of the above is insanity and has introduced reliability issues onto our grid. It has also greatly increased the cost of electricity in Texas. Much of this increase is hidden from electric rates. For example, 2 years ago Texas created the Texas Energy Fund (TEF) and took $10B out of general revenue (tax funds) to help reliable, dispatchable energy producers to add generation capability to the grid. Did not increase electric rates at all but still takes money out of all 30 million Texan’s pockets. That works out to $333 per person. All because we still add more wind and solar to the grid and subsidize them. If we had an honest system, each source would not have subsidies and would have to stand on their economic feet.
Mr. Wojick was 100% correct and you made a statement without any facts to back up your argument. Understand, when demand is highest wind sucks (it doesn’t blow) and will result in 3rd world conditions.
You’re welcome.
KW???…MW???
10,000 MW. Thanks for catching the erroneous label.
The problem is, you do not get to decide if the cost per person is worth it. Do you have grandchildren that will need natural gas left in the ground to cook their food and heat their home and water, without electricity?
Nothing you said has anything to do with what I am talking about, I am talking about not burning to make electricity, liquefying, selling to Mexico any more natural gas than is currently being consumed. It is an independent and disaster proven commodity that must be preserved for the infrastructure that is currently in place, that has nothing to do with making electricity.
Burning up all the natural gas to help ChatGPT allow for a better search experience is not something that is really sustainable, nor is it something that the citizens of Texas will put up with.
An in-person conversation would have gone much worse with Mr Wojo.
Agree, in a sane world wind and solar grid contamination would be illegal instead of subsidized.
It’s an “asset” most of the time in Spring and Fall, when demand is lowest. It’s a severe liability on most very hot summer days and very cold winter nights.
There are days when wind carries more than 50% of the load and days when it doesn’t carry its own weight.
question..when gas generation is not needed, are they completely shut down or burning gas at idle keeping the water warm?
Not only are the fossil fuel plants burning energy to “keep the water warm”, but when the wind isn’t blowing, wind turbines consume power. Both the keep the lubricant warm and to keep the blades moving so that flat spots don’t form on the bearings.
50% of the year? Not even close. Closer to 5%. As to saving fossil fuels, those savings are almost entirely imaginary, since the fossil fuel plants have to be run in warm to hot standby in order to handle the inevitable drops in wind velocity.
I lived through February 2021. Wind and solar do not function in freezing rain.
Wind is a diversion of investment.
I lived through it as well, a child died from hypothermia the 2nd night less than 10 miles from my home. Don’t get me wrong I am not an advocate of wind, we should have 50 next gen reactors distributed throughout the state (and will someday). Wind is not and should not be depended on to provide power during a freeze or no-wind period, it is merely a steppingstone that was forced upon the state by Bush Jr. The grid is balanced to handle that, the problem is that there was not sufficient insolation on the feed lines supplying the gas fired plants. Even the coal fired plants were having trouble staying online. If you would have told anyone responsible for winterizing any of the sites that failed during the freeze a year before that they need to be able to handle Arctic conditions for 3 days or children will die in their sleep, they would have laughed at you.
My point is only about the truth, and the truth comes from data. In the mix of energies in the Texas grid, wind is the 2nd most powerful entity by a large margin. On the graph of the last 366 days of power generation by energy source, anytime you see the green line going above the gold line, that means the wind is carrying more of a load than Natural Gas. There are a number of emotional responses to my comment above, but just because you feel something strongly, it has no real bearing on reality if you do not have anything other than your words. Everything is bigger in Texas, especially WIND.
The malign effect of wind is making other sources less economic, given the subsidies for wind, and not making the subsidy miners pay for their weather dependent supply. Backup is required, but it is not charged to the cost of wind and solar.
“Companies are already making enormous profits using artificial intelligence to perform accelerated drug discovery and repurposing, intelligence analysis, semantic analysis and collation of news and events, and bringing greater than human capabilities to forward planning, strategy and scientific research.”
I’m no Luddite, but sheesh- is all the disturbance required to massively increase energy production to satisfy AI really worth it? As for improved drug discovery- if only people lived better, that is more exercise, better diet, don’t smoke cigarettes and don’t drink much booze- everyone would be healthier – rather than spending trillions to get better drugs. Most people already take too much medication. As for semantic analysis and collation of news- do we really need it that much? No doubt there will be benefits to AI – but previous major advances in technology didn’t require the public to spend trillions to support it. Is AI just another scam like “green energy”?
AI is like any other break through technology. Lotsa hype and some amount of substance that won’t live up to the hype. Remember the .com bubble? Almost all the hyped up companies went bankrupt, but Amazon, E-Bay, Google and PayPal all came out of that bubble.
Its too early to tell what AI will actually be good for in the long run. I use it instead of Google now, particularly for anything complicated where the answers I get prompt additional questions that need to be answered in the context of what I already got. Google of course has their own AI that they are trying to integrate with their search tool, but as far as I am concerned, the traditional google search is obsolete.
There will be a lot of bankruptcies and research dead ends with AI but there will be things that change the world too. My prediction is that anyone who is predicting what the big successes will actually be, is probably wrong.
It is worse than that: LLM/AI is simply a shit business. the technology has no scalability, it does not produce reliable results and it is horrifically expensive. LLM/AI is literally selling $1 for 30 cents right now and this does not improve with scale, unlike say search or even Amazon nationwide distribution and delivery.
Ai = ASS.
Artificial simple stupidity.
Yes, good points. The generation of new drug candidates and clinical trials still has to be done by real people making the candidates, and real human beings undergoing the testing (not to mention real animals where the toxicology is tested). Supposed in silico toxicology was and always will be a joke. I suppose AI will make the data analysis faster.
Repurposing doesn’t need long clinical trials because those trials to establish safety have already happened.
Somewhere out there are drugs which cure cancer, treat mental illness, perhaps even delay old age, which are currently in use for blood pressure, weight loss, treating headaches. Their additional qualities are completely unknown.
Discovering those additional capabilities using AI is generating huge profits.
Yes, I know this, I’m in the field. What’s a good example of a drug. that was repurposed by AI and is now (present tense) generating huge profits?
I love AI, use it every day and know it will be very beneficial in human medicine. The rate-limiting step in getting a new drug approval is the laborious clinical trial process and that includes anything AI can generate.
I made a drug that was approved for human use when I was a grad student (got lucky) and I now have two in clinical trials at a company I work with. I actually want to repurpose one of them for another indication, so point me at a reference where AI will help me, please.
I call bullshit on your assertion that “AI is generating huge profits”.
OpenAI lost 5 billion dollars last year. They will lose twice that this year.
The entire industry is a gaping black hole for money.
Subprime 2 coming fast.
Just pump and dump.
Valid questions.
Let the debate continue!
Agree 100% Joseph. Personally, I think all this computer stuff needs to slow down rather than speed up, but then some do call me a Luddite–ever since reading Player Piano (Vonnegut) back in the mid 60’s
AI – at least this version which is LLM – is bullshit.
The numbers also are skewed to represent AI being bigger on power demand than it actually is, at this moment.
For example: Microsoft is leading the LLM/AI charge. Microsoft was the original partner for OpenAI starting in 2016; OpenAI/ChatGPT represents at least 75% of the space.
Microsoft had 4 GW total in all of its data centers as of April 2024 ramping to 7.5 GW by the end of 1H2025 (i.e. has not happened yet). It is safe to say that the entire cloud data center space as represented by the big US tech companies in the US is at/below 20 GW right now. This means existing build, not future.
But but but LLM/AI is the future! Well, Microsoft doesn’t apparently think so. Tyler Cowen just came out with a report 3 weeks ago showing that MSFT has cancelled over 1 GW of future data center build. Microsoft also changed its partnership agreement with OpenAI last year to allow OpenAI to get cloud compute from Oracle. Keep in mind MSFT has the most insight into OpenAI’s operations because of its multiyear partnership and the literal billions of GPU compute credits which MSFT has either given OpenAI or has sold at a massive discount (estimated at 66% off list price).
The latest, biggest sign that the LLM/AI space is not going well is the twin news of Stargate and Softbank. Softbank has been the stupid money bagholder for literally every tech scam in the last decade; Stargate is using federal money to build the GPU compute. Now why would any self respecting, private sector GPU compute data center provider want a subsidized federal government competitor?
They would not, if there was real money to be made.
Long story short: LLM/AI is big on hype, small on profit; long on bullshit, short on delivery.
What a hilarious dilemma for Big Tech.
They HAVE to invest in AI to stay competitive.
They CANNOT run their AI data centers on solar and wind.
They HAVE to provision nuclear or gas for generation.
So they not only have to abandon their precious virtue signalling solar and wind, they have to abandon their refusal to use nuclear and/or gas too. They will no doubt paper this over by having some solar and/or wind for show in every install and when one does the numbers it will turn out its cheaper to just not hook them up.
I said two decades ago that the climate nonsense will continue until economic imperatives outweigh the narrative. This wasn’t quite what I had in mind, I was more thinking about people starving or hospitals without power, but hey, I will take the win!
EDIT … sorry, hit the wrong reply button but, while I’m here, yes totally agree.
Recursive iteration requires huge data centers, which requires a huge and reliable energy supply to maintain operations.
Solar, wind and battery storage are not reliable. So big tech cant use them going forward with any sensible business plan.
Yep. I saw it coming 8 years ago, smells like victory now it is here.
The greens do have one last card to play, stoking fear of malevolent AI to try to squash the revolution under the weight of public panic. Expect to see lots more “terminator” style movies in the near future.
Sorry. I cannot do that, Dave.
One of the teen agers I used to tutor in math got me into video games. One day I picked up the controller and the Xbox said in a very monotone voice “Sorry, I can’t do that for you Dave”. I have to admit, one of the better practical jokes anyone has ever pulled on me.
I really did come to hate that movie. For at least a YEAR afterward, it didn’t matter who was calling to ask for something, they would all go “Sorry, I can’t do that Dave” and howl with laughter.
Up in Smoke was worse. A full year of answering the phone with “Hi, Dave here” would be responded to with “Dave’s not here man” and gaols of laughter.
“Dave: How would you account for this discrepancy between you and the twin 9000?
HAL: Well, I don’t think there is any question about it. It can only be attributable to human error. This sort of thing has cropped up before, and it has always been due to human error.
Frank: Listen HAL. There has never been any instance at all of a computer error occurring in the 9000 series, has there?
HAL: None whatsoever, Frank. The 9000 series has a perfect operational record.
Frank: Well of course I know all the wonderful achievements of the 9000 series, but, uh, are you certain there has never been any case of even the most insignificant computer error?
HAL: None whatsoever, Frank. Quite honestly, I wouldn’t worry myself about that.”
(From 2001, A Space Odyssey)
“In this race, there is no prize for second place.”
What, no participation trophies? Won’t that upset an awful lot of people who expect to be rewarded for showing up?
This is, after all, a trans gender energy identity issue!
Fire up all fossil fuel and nuclear generators. Build new fossil fuel and nuclear generators. Remove all wind and solar from the grid.
We have lots of available freed up labor thanks to DOGE. We can form a partnership between government and renewable companies for the purpose of removing wind and solar farms and properly disposing of them.
Some may question that I include government. Two reasons they have the power to force the renewable companies to join in and two if there is one thing the government should be able to handle it is to destroy things. Start with the old decrepit eyesores and work your way to the newer ones.
Anyone who wants to buy a wind or solar farm is welcome to with the understanding that you are responsible for the operation, maintenance and eventual dismantling. No hiding behind corporate badges, individuals will be held responsible. You can apply to be connected to the grid but it is unlikely your power will be needed or wanted.
What the heck is “decarbonised AI supercomputer?”
Curious minds want to know?
Nuclear power. Wind and solar are no good for AI.
One that only operates 30% of the time?
“What the heck is “decarbonised AI supercomputer?”
None of the vacuum tubes use a carbon filament? 😎
Is that a 2/3 increase in ERCOT capacity? Is that even possible by 2030?
Yes. And No.
These companies that need this giant supply of electricity to function should build and pay for their own power plants. Ratepayers should not have to subsidize these companies. They have plenty of money and can generate their own electricity.
Otherwise, everyone’s electric bill is going to go higher.
Trump says he favors these companies building their own power plants to supply themselves, and a couple of investors said they wanted to build their own power plants.
Trump declared an energy emergency, so he can waive a lot of regulations that would have increased costs and increased time to completion.
Don’t put the onus on the rate-paying public. Our electric bills are already too high.
I had read something about Google doing something !!! It was probably lamestream media, so I didn’t commit it to memory. So, I just asked perplexity.ai and it came up with some up-to-date stuff. Here are a couple of interesting bullets (from many):
Meanwhile, the all-in Net Zero countries are looking at these to meet their additional AI energy loads –
(one issue to be resolved first though, is that all the operating instructions are in Tibetan 🙁 )
This is seriously wrong. The data centers will not be built if they cannot be powered. They have to apply for service then wait. The real story is how long AI will have to wait. For much of this “demand” ten years looks good.
China is scaring the pants off US Government insiders right now, at least those who aren’t on the Chinese payroll – they are advancing way too fast in this space. I expect it to mirror the space race.
Time for us to build modern coal fired power.
There is no doubt AI has utility. How much has yet to be seen thru all the startup hype. Anthropic valuation ~$65 billion. xAI valuation $53 billion. Haven’t seen stuff this crazy since the .com bubble and Sun Microsystems—‘we put the dot in dotcom.’
There is also no doubt AI electricity demand will put strains on existing electricity grids. I previously looked up an older ChatGPT version power consumption during a months long single training session. On the order of 40 MW continuous.
Expanding grid capacity takes time, so will be a significant constraint on AI development for at least a while. For example, the new 2.4GW CCGT power station at Port Everglades (replacing a very old resid fired 2.2GW tear down) still took three years despite reusing all the pre-existing grid infrastructure.
Microsoft exploring restarting old 3 Mile Island unit 1 for a proposed AI data center shows how desperate the grid capacity situation is becoming.
In consequence, states and Nations worried about ‘going green’ will be significantly AI disadvantaged. For sure, Europe and Australia. California may have AI startups, but it won’t have their data centers.
Rud, there’s a saying among Californians that describes how their business guys think: Take it easy, but take it. I think that will start applying here.
regarding regulation of nuclear, coal and natural gas for base load, i don’t think they can build base load generation fast enough to meet the demand for these data centers. it would not surprise me that legislation, under the term of national security, is enacted to give priority to these data centers and what ever juice is leftover is supplied to the common rate payer.
I doubt the AI craze will live very long. Certainly not as long as so called green tech, which is now on its way out
In light of this, I think it’s a new way to transfer money from taxpayers to corporations, the elite and the political class
I also think the new European spending on defence could be the same, since the writing is definitely on the wall in Europe as far as net zero is concerned
This current craze looks like repurposed machine learning and data science. They were similar crazes that investors and businesses soon got bored of. Even in academia, machine learning has been renamed AI to make it sound different (novel) for the purposes of papers and funding. The people doing this almost invariably understand neither of these areas
The data science part is primarily related to collecting, cleaning, and storing the data and the machine learning part is developing algorithms that can learn from this data and perform specific tasks, there is little or no AI as far as I can see.
The data centres are tied to all the fancy ideas like Internet of things and block chain tech that have been talked about forever but have never materialised
Aside from a few R&D heavy companies like pharmaceuticals, I don’t really see a big take up of these services. Not on a scale that would justify the investment and running energy costs.
Even the pharmaceutical companies and others would prefer their own in house equivalents, which is perfectly feasible and already the case for most of their projects
I don’t give it more than a few years
Yes, a lot of major facets of just “biotechnology” are now labeled “synthetic biology”. I do it myself.
BTW, I don’t know how long exactly you’ve been posting. I’ve noticed your posts in the past couple of weeks, and not only do a lot of them correspond to exactly how I think, but you manage to say a lot in your posts. Do keep posting please.
Interestingly, my prediction vis-à-vis your last sentence differs from yours. Perhaps you just mean the AI moniker? I’m a firm Ray Kurzweil fan and I think that computing will just keep on expanding, possibly logarithmically, until it hits the Singularity. Then … choose your own ending to this sentence !!!!
Thanks for the compliments Phil.
I don’t honestly know where it will top out. I think the key to a leap in hardware performance depends on quantum computing
I haven’t looked into it so I don’t know how feasible it is – is it another cold fusion…..
On the face of it, according to the foundational postulates of quantum mechanics, there is an uncertainty on the quantum scale. Bodies exist in a superposition (linear combination) of possible states, which are fixed only upon observation.
This is fundamentally different from the deterministic 0-1 binary logic of computing today. I don’t know how it is possible to deal with this, but I suppose the experts are trying to figure it out
For now, the next big thing seems to be data, masses and masses of it
Machine learning (not AI, which deals with all aspects of intelligent systems including sensing and mechatronics for hardware) is like any other set of methods, it has its limitations. I don’t forsee any game changing advances in the underlying methods
It seems to me to be that the only way to improve accuracy currently is these massive data sets, and more complexity in the transformer based LLMs by adding more layers. More data means more degrees of freedom/parameters can be added
It boils down to optimising some loss function or empirical risk over these parameters and the more data the more accurate – the curse of dimensionality links data set size to input space size n, the former growing approx according to a some number to the power n for a fixed accuracy
Excellent. Thank you. WUWT moves so fast, this will probably get lost.
Sometimes that’s a good thing for me, ha ha. I do like to fire at will at the trolls, while also getting an education on here, and sometimes contributing.
Here’s around 6GW of opportunity for reactivation of coal fired generation (source wikipedia, so it MUST be right). Texas can easily delay closure or contine indefinitely on about 10 GW more of capacity currently on the phase-out list. Speeding up NG conversion of several other transitioning or recently closed plants could also help.
Texas is also going for the world’s third largest desalinization capacity, as described today, right here on WUWT https://wattsupwiththat.com/2025/03/07/texas-is-going-big-on-desalination/ That will also require reliable electrical power.
But the good part about the AI demand and water desalinization production is that computer and water production time can be scheduled around power load times for other industrial demand. What this means is that with proper scheduling and some creative connectivity upgrades, the reliable portions of the existing generator grid can be run at a high capacity factor and get some partial good for AI and water while permanent upgrades to power generation are under construction. Likewise, temporary or permanent capacity installed for AI or desalinzation can be switched over to service general demand on relatively short notice — so AI generation can offer some peaker plant capacity to meet demand for weather events or other high demand conditions.
Small scale, semi-portable gas turbine technology is very well developed in support of hydraulic fracturing. Look for more portable, modular generation stations like those being built in Tennessee, Louisiana, Virginia and many other states.
99GW needed within five years?
Wind takes 4-8 years to construct smaller turbines. 7-12 years for larger structures.
Solar Farms 8-18 months. No 24×7 reliability.
Nuclear 10+ years to construct. SMRs not available until 2030+.
Natural Gas 12-18 months.to construct
Oil and coal not desirable
Which type of power plant would you recommend?
BTW, that’s over 100 average NG power plants.
I hate wind and wish it did take 4-8 years but I’ve witnessed onshore at <2 years, but for a lot of other reasons here in the U.S with natural gas at <$5/MCF wind and solar are both off the chart more expensive.
I wonder if Texas will just plug their desal plants into the grid or use dedicated CCGT, surely not wind and or solar. The only desal plant I have ever even seen is The Jubail Desalination Plant located in the Eastern province of Saudi Arabia. It has an installed gas turbine capacity of approximately 2,745 MW, and the plant can produce around 800,000 cubic meters of desalinated water per day. . It was built in the 80’s.
I drove past it numerous times back in the 90’s, there are visible emissions from the natural gas used for power (no wind or solar in the area, chuckle). Meanwhile, Saudi Aramco, in the Eastern province has built a 20MW solar desal plant in 1998 and is scheduled to start up a new 970 MW gas fired desal plant this year. If Texas wants to go big, they have some work to do. BTW the Saudi CCGT/Solar ratio looks about right, (4,000 MW CCGT vs 20 MW solar).
Nuclear power is the only realistic option, and even then it’s a massive implementation.
100GW is 30 massive nuclear power plants or at least 200 SMRs
Even then I am not sure Texas has the ability to cool that much – that is probably around 200GW of heat that needs to go somewhere.