Tech Industry Wants to Lock Up Nuclear Power for AI (wsj.com) 70
Tech companies scouring the country for electricity supplies have zeroed in on a key target: America's nuclear-power plants. From a report: The owners of roughly a third of U.S. nuclear-power plants are in talks with tech companies to provide electricity to new data centers needed to meet the demands of an artificial-intelligence boom. Among them, Amazon Web Services is nearing a deal for electricity supplied directly from a nuclear plant on the East Coast with Constellation Energy, the largest owner of U.S. nuclear-power plants, according to people familiar with the matter. In a separate deal in March, the Amazon subsidiary purchased a nuclear-powered data center in Pennsylvania for $650 million.
The discussions have the potential to remove stable power generation from the grid while reliability concerns are rising across much of the U.S. and new kinds of electricity users -- including AI, manufacturing and transportation -- are significantly increasing the demand for electricity in pockets of the country. Nuclear-powered data centers would match the grid's highest-reliability workhorse with a wealthy customer that wants 24-7 carbon-free power, likely speeding the addition of data centers needed in the global AI race. But instead of adding new green energy to meet their soaring power needs, tech companies would be effectively diverting existing electricity resources. That could raise prices for other customers and hold back emission-cutting goals.
The discussions have the potential to remove stable power generation from the grid while reliability concerns are rising across much of the U.S. and new kinds of electricity users -- including AI, manufacturing and transportation -- are significantly increasing the demand for electricity in pockets of the country. Nuclear-powered data centers would match the grid's highest-reliability workhorse with a wealthy customer that wants 24-7 carbon-free power, likely speeding the addition of data centers needed in the global AI race. But instead of adding new green energy to meet their soaring power needs, tech companies would be effectively diverting existing electricity resources. That could raise prices for other customers and hold back emission-cutting goals.
Real headline: (Score:5, Insightful)
"tech" "industry" has latched on brute force in an effort to ride the most recent hype wave, trying to sell worthless buzzwords to the unwashed investor masses.
Re: (Score:2)
All that matters is my stocks in AI-related companies keep going up. However it's done is fine by me.
Re: (Score:2)
It's not "Woosh!", its Poe's law.
Re: (Score:2)
"But we need this, this is the next revolution, the era of artifici..."
"No"
"Imagine all the possibilities we would unlock, and artificial mind, 24/24 7/7 unlimited..."
"No"
"OK we pay you a bazillion dollars, to start a new plant"
"No"
"If we do not do it, China will! They will enslave..."
"No"
"This is Rebecca, she wants to talk to you"
"No"
starts jumping up and down in anger, breaks through the floor and dies.
"No"
Re: (Score:2)
Investors don't give a shit about carbon fantasies beyond regulatory necessity ... it's just a mass delusion among managers.
Re: (Score:2)
Throw in a bit of greenwashing too. Nuclear is only comparable to renewables for carbon emissions in certain circumstances, some of which, such as long timer waste storage, have yet to be resolved. It's the classic tech industry scam of quoting the theoretical best case, rather than the actual reality users will experience.
Re: (Score:2)
Long time waste storage is mostly a red herring. It would be resolved if the ridiculous bans on reprocessing and development of reprocessing technologies are lifted. After reprocessing, the amount of "waste" that needs to be taken care of is reduced to nearly nothing. The situation with development of breeder reactors is also similar.
Both would require significantly smaller investment than fiction like "fusion power", and have real benefits. Neither will happen because of fear of "proliferation".
Re: (Score:2)
The main reason why we don't do much reprocessing is because it costs more than just making new fuel. Nuclear is already very expensive and they aren't going to waste more money on reprocessing when they can just dig up more.
Economics are nuclear power's Achilles' heel.
Re: (Score:2)
The main reason why you don't do much reprocessing is that you have nothing to do with nuclear power.
The main reason why US doesn't do it is because it has been prohibited by policy since the 70s, somewhat like this:
1976
In an October 28 nuclear policy statement, President Ford announced his decision that
the reprocessing and recycling of plutonium should not proceed unless there is sound
reason to conclude that the world community can effectively overcome the associated risks
of proliferation
With that announcement, agencies of the executive branch were directed to delay
commercialization of reprocessing activities in the United States until uncertainties were
resolved.
1977
In an April 7 press statement, President Carter announced, “We will defer indefinitely the
commercial reprocessing and recycling of plutonium produced in the U.S. nuclear power
programs.”
1990
In the National Defense Authorization Act for Fiscal Year 1991 (P.L. 101-510, Sec. 3142),
Congress declared under Findings and Declaration of Policy that
[a]t the present time, the United States is observing a de facto moratorium on the production
of fissile materials, with no production of highly enriched uranium for nuclear weapons
since 1964.
1993
President Clinton issued a policy statement on reprocessing stating that “[t]he United States does
not encourage the civil use of plutonium and, accordingly, does not itself engage in plutonium
reprocessing for either nuclear power or nuclear explosive purposes.
The main reason it doesn't happen almost everywhere in the rest of the US-allied world is that US is strongly discouraging it, as an extension of their domestic policy.
The main reason it doesn't happen in other nations that develop their own nuclear power, like France or Russia - oops, it happen
Not surprising (Score:3)
Re: Not surprising (Score:2)
Microsofthas underwater data centers in Scotland. Seems like we could do them here along the Gulf coast. Especially on the shallower shelf around Texas and Alabama. Run power and data cables to nearby major metropolitan areas like Dallas.
Re: (Score:2)
Re:Not surprising (Score:5, Insightful)
Sure, the nukes are already built, but they're already being used for baseload duties. Just taking that output and hooking it up to new data centers means you still have to make up the difference by building new plants of some kind. Or possibly running inefficient load following plants more.
So the AI industry is striking deals that will enable it to claim it's carbon neutral while *forcing everyone else to burn more fossil fuels*.
Re: (Score:2)
Re: Not surprising (Score:2)
I'm reliably informed that wind and solar are cheaper than new fossil fuel plants and are perfectly suited to reliably supplying the grid. I'm sure therefore that all new sources will be quick to ramp up and will work perfectly fine as well as being emission-free.
Re: (Score:2)
We're almost there. We just need a couple more years to work on grid storage.
Re: (Score:2)
The green line says otherwise.
https://transmission.bpa.gov/b... [bpa.gov]
Re: (Score:2)
Re: (Score:2)
Sure, the nukes are already built, but they're already being used for baseload duties. Just taking that output and hooking it up to new data centers means you still have to make up the difference by building new plants of some kind. Or possibly running inefficient load following plants more.
So the AI industry is striking deals that will enable it to claim it's carbon neutral while *forcing everyone else to burn more fossil fuels*.
Exactly. They simply want the power and don't really care where it comes from as long as they get what they need. In reality, they're getting power from whatever source is best for the grid at the time, just as someone paying extra for green power may actually be getting theirs from a coal plant. If needed, the data centers will push to reopen old coal plants and eventually build their own and pay the power company to run it.
Re: (Score:2)
The already built nuclear plants are getting old. They don't last forever and there is only a limited amount of maintenance you can do on stuff that is highly radioactive, like the reactor vessel. Over time the spent fuel and other waste products build up too.
Speaking of water, nuclear power requires a lot of it.
Re: (Score:2)
The already built nuclear plants are getting old. They don't last forever and there is only a limited amount of maintenance you can do on stuff that is highly radioactive, like the reactor vessel. Over time the spent fuel and other waste products build up too..
.However, they are a here and now solution, even if an interim one It's not just nuclear, it's really any source that can provide them with large amounts of uninterruptible power, including recommissioning coal plants.
Speaking of water, nuclear power requires a lot of it.
True, but it's use is already there, a data center represents a new demand on what in some areas is already a very limited resource and in others significantly increase the demand on an aging system.
Re: (Score:2)
In that sense then yes, this might be a decent use for old nuclear plants that are struggling to integrate into the new and more dynamic grid. But then you come to the question of who gets to pay for it. One of the advantages of not having data centres be guaranteed loads for nuclear plants is that it encourages their owners to invest in better distribution, so they can export that power to further away when the local area is saturated with renewable energy.
Re: (Score:2)
In that sense then yes, this might be a decent use for old nuclear plants that are struggling to integrate into the new and more dynamic grid. But then you come to the question of who gets to pay for it. One of the advantages of not having data centres be guaranteed loads for nuclear plants is that it encourages their owners to invest in better distribution, so they can export that power to further away when the local area is saturated with renewable energy.
While the story focuses on nuclear while the real story is data centers want large amounts of power and don't care where it comes from as long as they get it. Localities like them because, well, jobs and tech; and leave it to the utility to figure out what to do with a huge load that may very well take all their new generation and thus thrown the growth plans into a mess. That new plant that as meant to handle say 5-10 years of regional growth is now dedicated to one new customer who has priority over mos
Re: (Score:2)
(I'm all for nuke power, tho. Always have been, always will be.)
It's still one of the safest careers to go into in the US.
Re: (Score:3)
That kind of thing works better with a few concrete examples.
But Bill Gates, for one, was a co-founder of Terra-power [wikipedia.org] in 2006.
Whether he is not a tech-bro or rather the OG-tech-bro I will leave to the jury.
Re: (Score:2)
Greener would be better, but Nuke is not bad. (Score:2)
Nuke powerplants have one major issue - the fear of people being affected by an accident. If we build a new nuke plant in the middle of nowhere by a server 'city', that would make a lot of sense.
Of course, the article is entirely correct that building server farms by variable power supplies - such as solar or wind - and designing them to raise activity when power generation is high and reduce activity when power generatio is low makes more sense.
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
Those sound very bad and scary... until you look at coal power plants.
The average coal power plants not only give massive CO2 emissions, but also emit more radiation than TMI, Simi Valley or Windscale ever did. Thorium, you know, is found in coal. Along with Mercury to add that ever present risk of insanity, but that is another story.
Basically, coal is the worst thing to burn, followed by oil, with Natural Gas bringing up the rear. The middle category of "better than Fossil Fuels: include Nuclear and Hy
The marketing for Fallout has gone a bit too far (Score:2)
This is what wind power actually produces (Score:4, Informative)
The UK has about 29 GW of wind installed, about half off- and half on-shore. Its a useful test case for these kinds of ideas. This is what it produced, day averages:
2023
minimum: 0.071 GW
maximum: 17.485 GW
average: 7.192 GW
2024
minimum: 0.16 GW
maximum: 16.713 GW
average: 7.451 GW
The suggestion in the piece is that AI customers should install this to run their servers? Its obviously impossible, this is not a usable or saleable product, and would never get started if not for heavy government subsidies, compulsory purchases etc.
Solar? Even more useless. Hardly any production between November and February, and what there is is sharply peaked and is delivered between about 11am and 1pm.
They are buying nuclear because it delivers what they need, dispatchable power. Their other alternatives would be gas or coal, but those at present are politically impossible.
The ironic thing about the piece is that its objecting to the AI companies buying dispatchable power on the grounds that it takes away the reliable power and leaves everyone else with... the unreliable power which is increasingly coming from wind and solar. And then they propose that the solution is for the AI people to start relying on...? The unreliable power from wind and solar, while leaving the reliable from nuclear etc for existing users.
The plain facts that pieces like this refuse to consult are simple. One, wind and solar are intermittent. Two, intermittent electricity cannot run industrial processes. Or much of anything, actually.
Re: (Score:2)
And by the way, someone will propose storage. The UK is once more a test case on this. The Royal Society estimated how much storage is needed for the UK to deliver net zero with its demand peak of 45GW.
The answer is about one third of annual demand, or 100 TWh.
I don't suppose, if the AI companies went out to tender and specified they wanted dispatchable supply, any wind or solar operators would tender. Could they actually deliver it in the first place? Probably not. And for sure they could never make a
Re: (Score:2)
You are presupposing that the entire country has to run on battery, but this is about powering data centers.
Re: (Score:2)
https://www.aurecongroup.com/insights/data-centres-the-worlds-greatest-energy-guzzlers#:~:text=Data%20centres%20consume%20up%20to,strategy%20is%20key%20to%20success
Re: (Score:3)
>> Hardly any production between November and February
The UK is at a latitude farther north than Maine so its no surprise that there isn't much insolation. All of the southern US gets copious sunlight all year.
The data centers should install or secure access to electricity from battery storage. There are lots of different kinds these days and they can be built much faster than nuke plants.
Re: (Score:2)
"All of the southern US gets copious sunlight all year."
Do they? Heavy cloud cover can cut the input to 7% of nameplate. That's real data by the way.
Re: (Score:2)
Yeah, well sunset cuts the input way down too but so what? The southern US does get very good insolation, and particularly in comparison with the UK.
Re: (Score:2)
The point you missed is that the cloud cover will prevent you from recharging the batteries you need to get through the night.
Or you can just fire up the natural gas power ICE generator. But that's a whole different set of economics. Where is the trade off between fuel costs for the engines vs the PV panels. Either way you have capital sitting idle whether it's ICE generators not running in the daytime or the PV not doing anything useful at night or on cloudy days. Justifying batteries gets real difficult i
Re: (Score:2)
I don't see anyone claiming that solar can be the sole source of grid electricity at present so why are you belaboring this? Solar plus batteries reportedly do power individual residences with little or no input from the grid. Clearly it is feasible up to a point.
Meanwhile the EU has achieved about 47% electricity from renewables.
https://cleantechnica.com/2024... [cleantechnica.com]
Re: (Score:2)
Because unlike you and me, most people seem to think their solar panels always put out the rated power any time the sun is up. And even fewer take into account the costs of the other generator they will need when renewables fail to carry the load.
Re: (Score:2)
There is also wind power and various kinds of battery storage. The EU gets about 47% of electricity from renewables and climbing even with relatively poor insolation.
Re: (Score:2)
Techfarming...
Re: (Score:2)
They are "buying" nuclear because they have convinced themselves that promises about emissions somehow impress investors and consumers.
When the nuclear plant has some downtime they will be sipping from the grid all the same ...
Re: (Score:2)
Where did you get these numbers from? I searched and the only result matching them is your post.
The numbers for 2024 are definitely wrong. Here's the stats for right now: https://grid.iamkate.com/ [iamkate.com]
17.22GW from wind, higher than your claimed maximum, and today isn't particularly windy, not like back in the winter months.
Also the official government figure for installed wind capacity was 30GW back in 2022, so unless it's somehow fallen since then (it hasn't, new wind has come online) your 29GW figure is wrong
Obvious outcome (Score:3)
Re: Obvious outcome (Score:3)
Might be cheaper to force relocate tech bros to a Mars dome colony or Venus balloon city. And let us meek people inherit the Earth.
Re: (Score:2)
Re: (Score:2)
2) AI begins running society
3) SKYNET!
Re: (Score:2)
5) Profit! (sorry, Cowboy Neal made me do it)
Re: (Score:2)
Re: (Score:2)
Greenwashing Coal (Score:4, Insightful)
Allow big-tech to build their own nuclear reactors (Score:2)
The current nuclear in the US is like the old space programs run by NASA, LM and Boeing. Always delayed and over budget. We need a SpaceX for nuclear power.
Re: (Score:2)
Great Scott! (Score:2)
One point twenty one jiga watts!
Fear Not (Score:1)
Anything that provides jobs for Americans will fail.
See, if an American gets a job he'll be able to afford a place to fuck, and that might lead to pregnant women.
Can't have any of that now.
Waste heat (Score:2)
Next step (Score:1)