SoylentNews is people
SoylentNews
from the buzz-off-there-isn't-enough-power dept.
Bees Reportedly Stopped Meta From Building A Nuclear-Powered Ai Data Center
Arthur T Knackerbracket has processed the following story:
Meta had plans to build an AI data center in the US that relies on nuclear power — it even already knew where it wanted the facility to be built. According to the Financial Times, though, the company had to scrap its plans, because the a rare bee species was discovered on the land reserved for the project. Company chief Mark Zuckerberg was reportedly ready to close a deal with an existing nuclear power plant operator that would provide emissions-free energy to the plant. The Times said he told staff members at an all-hands last week that pushing through wouldn't have been possible, because the company would encounter numerous regulatory challenges due to the bees' discovery.
Zuckerberg reportedly told his staff that Meta would've had the first nuclear-powered AI if the deal had gone ahead. It still might come true if the company could find a way, but it has to move quickly because its biggest rivals are investing in nuclear energy, as well. In September, Microsoft revealed that it intends to revive the Three Mile Island nuclear power plant to provide energy for its AI efforts. Meanwhile, Google teamed up with startup Kairos Power to build seven small nuclear reactors in the US to power its data centers starting in 2030. And then there's Amazon, which announced three agreements with different companies to build small modular reactors in mid-October.
The Times didn't say whether Meta is looking for a new site — one that doesn't have rare bees living in its vicinity. One of its sources only said that Meta is still exploring various deals for emissions-free energy, including nuclear, to power its future AI data centers.
Regulators Reject Power Deal For Nuclear Amazon Datacenters
Arthur T Knackerbracket has processed the following story:
Amazon has hit a roadblock in its plans for nuclear-powered US datacenters. Federal regulators rejected a deal that would let it draw more power from a Susquehanna plant to supply new bit barns next to the site, on the grounds this would set a precedent which may affect grid reliability and increase energy costs.
The Federal Energy Regulatory Commission (FERC) issued an order on November 1 rejecting an amended Interconnection Service Agreement (ISA) that would have increased the amount of co-located load from 300 to 480 MW, and to "make revisions related to the treatment of this co-located load."
Co-located load means the Cumulus datacenter complex that Talen Energy built next to the 2.5 GW Susquehanna nuclear plant in Pennsylvania which it operates, and which Amazon acquired in March via a deal worth $650 million.
The online megamart announced plans in May to expand the site with more than a dozen new datacenters for its Amazon Web Services (AWS) cloud subsidiary over the next decade.
Soon after that, official objections were filed by two utility companies, American Electric Power (AEP) and Exelon. They argued that the revised agreement between Talen and PJM Interconnection, the regional power grid operator, would give the Cumulus site preferential treatment and may result in less energy going to the grid in some circumstances.
Exelon and AEP also argued that the amended ISA should be subject to an official hearing because "it raises many factual questions," and, in the absence of any such hearing, that FERC should reject the amended ISA. It seems a majority of the commissioners agreed.
Specifically, Exelon and AEP said the amended ISA had not been adequately supported, meaning no good reason was given as to why the amendments were necessary.
The FERC ruling notes that PJM says up to 480 MW of power could be delivered to "the co-located load" without a material impact on the grid, and any additional load beyond that would result in "generation deliverability violations" and require installation of system upgrades.
The amended ISA also states the co-located load "is not intended to consume capacity and/or energy from the PJM transmission system [the grid]," but it is "possible that it will."
Exelon and AEP expressed concern that it was unclear what steps have been taken to ensure any such withdrawal of power from the grid would be properly metered and accurately billed when it does occur, and who would be financially responsible.
In its summary, FERC said PJM had not demonstrated that its proposed "non-conforming" provisions in the amended ISA are necessary deviations from the existing agreement, and therefore rejected it.
One dissenting voice was FERC Chairman Willie L Phillips, who claimed the order was "a step backward for both electric reliability and national security." He said he believed that PJM had addressed reliability issues comprehensively in its filing, and the ruling risks America's leadership in AI because "reliable electricity is the lifeblood" of the datacenters required for developing AI.
However, Commissioner Mark C Christie explained that co-location arrangements of this type present "an array of complicated, nuanced, and multifaceted issues, which collectively could have huge ramifications for both grid reliability and consumer costs," and it was being rejected because PJM had failed to meet its burden of proof.
If the proposed amendment were approved at this time, he warned: "we would be setting a precedent that would be used to justify identical or similar arrangements in future cases."
The move highlights the difficulties datacenter operators face in expanding their facilities to keep pace with the booming demand for training and operating the latest AI models, and the challenges that power companies face in delivering the energy required.
(Score: 1, Funny) by Anonymous Coward on Thursday November 07 2024, @12:41AM (1 child)
Doh, not a _positronic_ brain, but rather a *nuclear* brain! Woot!
Err, wai'ta sec..
They said zero emissions? Like, they'll never truck the spent nuclear fuel from the site? (They're going to use "AI" to solve the problem? It's been *so* successful thus far....)
(Score: 4, Funny) by Freeman on Thursday November 07 2024, @03:29PM
Don't you worry, the same careful and deliberate consideration shall be used as is used in all other "AI" things.
Joshua 1:9 "Be strong and of a good courage; be not afraid, neither be thou dismayed: for the Lord thy God is with thee"
(Score: 0, Touché) by Anonymous Coward on Thursday November 07 2024, @04:15AM (1 child)
just cut a deal with Trump, bees and EPA be dam'd.
(Score: 0) by Anonymous Coward on Thursday November 07 2024, @04:30PM
(Score: 2) by The Vocal Minority on Thursday November 07 2024, @08:06AM
THE BEES, MY EYES...
(Score: 1, Flamebait) by DadaDoofy on Thursday November 07 2024, @10:34AM (14 children)
"Zuckerberg was reportedly ready to close a deal with an existing nuclear power plant operator that would provide emissions-free energy to the plant."
Nuclear power plants emit radioactive waste in the form of spent fuel. When containment of these emissions fail in an explosion or meltdown, people die. But why let the truth get in the way of a useful narrative?
(Score: 3, Interesting) by janrinok on Thursday November 07 2024, @11:12AM (5 children)
You are correct, of course. But many installations which malfunction also release emissions that endanger human life. Perhaps you do not recall Bhopal in 1984 [wikipedia.org]?
More recently, train derailments in the USA have also endangered human life, and we are still waiting to see the long-term effects of those 'relatively minor' accidents. And the derailments were entirely preventable except that some safety regulations had been recently cancelled by a change to government legislation.
So it is not entirely honest to imply that nuclear accidents always result in a loss of life, or that they are have caused more deaths than other accidents. They might do but it is not a foregone conclusion. For example, after the accident at Three Mile Island [wikipedia.org] the eventual assessment was:
[nostyle RIP 06 May 2025]
(Score: 2) by DadaDoofy on Thursday November 07 2024, @02:08PM (4 children)
"So it is not entirely honest to imply that nuclear accidents always result in a loss of life"
I never did. Did you not see the word "when"?
In any case, that's some pretty weak whatabout-ism. The two disasters you cite had nothing whatsoever to do with power generation.
(Score: 2) by janrinok on Thursday November 07 2024, @02:43PM (2 children)
And I never said they were anything to do with power generation.
But you did say:
You didn't say that they might die, or have on some occasions died. You were quite categorical in saying "when containment fails.... people die". Three Mile Island proves that, in the long term, they might not even notice. Containment certainly failed though.
The other examples are to demonstrate that nuclear is no different from other, perhaps less politically driven and certainly less emotive, views of various industries.
I started off by agreeing with you that there are risks involved in the nuclear industry. The point that I was making is that they need not be any worse than the risks associated with other industries or even natural disasters. And I cited 3 disasters - one of which was nuclear related.
[nostyle RIP 06 May 2025]
(Score: 2) by DadaDoofy on Thursday November 07 2024, @03:50PM (1 child)
The containment vessel at Three Mile Island did not fail. If it had, the outcome would certainly have been more like the two disasters you conveniently "forgot" to mention. When the containment vessels failed at Chernobyl and Fukushima, thousands of people were killed and areas were left uninhabitable for centuries, if not millennia.
TFA is not about "various industries". It is very specifically about Meta's effort to use nuclear power generation. It very specifically makes the false and utterly ridiculous claim that nuclear power is "emissions-free".
(Score: 2) by janrinok on Thursday November 07 2024, @07:09PM
So they were supposed to release all that radiation? - that is still an emission that escaped its containment. Don't play with words. You made a statement that was incorrect.
[nostyle RIP 06 May 2025]
(Score: 4, Informative) by PiMuNu on Thursday November 07 2024, @04:48PM
On the other hand it is well established in the literature that e.g. coal power has killed many more than any other power source:
https://ourworldindata.org/safest-sources-of-energy [ourworldindata.org]
If we constrain ourselves only to release of radioactive material, coal power has released far more than civil nuclear power, (caveat: I exclude nuclear weapons tests).
https://www.scientificamerican.com/article/coal-ash-is-more-radioactive-than-nuclear-waste/ [scientificamerican.com]
Any omission of this information in a discussion would imply an attempt at bias and misinformation
(Score: 4, Interesting) by JoeMerchant on Thursday November 07 2024, @01:35PM (7 children)
Absolutely, nuclear power is dangerous, with long lasting negative effects that we have been too lax about mitigating.
However, per GWH produced, nuclear is far and away the safest way in the history of mankind to produce electricity.
People die in the construction and maintenance of hydroelectric dams, not to mention the environmental impacts.
Going way back, wood burning in the home for heat has well established negative effects from smoke inhalation to burning the shelter to the ground.
Coal? Yeah, fill in that blank for yourself.
--------------------
On the other hand, what kind of SciFi story starts out with mega-corporations taking control of nuclear technology for their own internal use? Dystopian is the only kind I can imagine.
🌻🌻🌻🌻 [google.com]
(Score: 2) by DadaDoofy on Thursday November 07 2024, @02:18PM (4 children)
It's ironic you would mention hydroelectric power. Not a single person has ever been killed by "emissions" from a hydroelectric damn, because there aren't any.
In spite of gloom and doom conspiracy theories about what might happen, the carbon dioxide emissions from conventional power plants have never killed anyone either.
(Score: 4, Interesting) by JoeMerchant on Thursday November 07 2024, @02:58PM
> Not a single person has ever been killed by "emissions" from a hydroelectric damn
My grandffather built dams with the TVA, he and his fiancee - later my grandmother, were nearly killed by displaced townspeople in the construction of the dam. The flooded towns were of course dispersed, kicking people off land and out of homes their families had owned for generations. The construction of the dams themselves is perilous: "According to Chinese state media, more than 100 workers died during the construction of the Three Gorges Dam. The dam was built over 12 years by 40,000 workers." - that's modern methods, and state media reported statistics.
I will agree: once constructed, maintenance of hydroelectric dams is mostly "clean" - but they do dramatic damage to river ecosystems up and downstream, and they displace large well established river communities of people in their construction.
Also, if you consider these, I think they're still outpacing nuclear in terms of deaths per GWH produced:
🌻🌻🌻🌻 [google.com]
(Score: 2) by janrinok on Thursday November 07 2024, @03:01PM
Isn't electricity one of the emissions from a hydroelectric dam? If not, why did they build it? :D
And he did say "in the construction and maintenance" of hydroelectric dams.
[nostyle RIP 06 May 2025]
(Score: 1) by day of the dalek on Thursday November 07 2024, @06:45PM (1 child)
False. Absolutely false. If you are counting spent fuel rods and other nuclear waste as emissions, then it's also fair to count water as an emission from a hydroelectric dam. When a dam fails, the consequences can be disastrous.
Consider the 2005 Taum Sauk Dam failure (Discussion from the National Weather Service [weather.gov], YouTube video about the dam failure [youtube.com]), which occurred when a hydroelectric reservoir atop a mountain in Southeast Missouri was overtopped and failed catastrophically. The Taum Sauk Dam was a hydroelectric storage facility, where excess electricity on the grid was used to pump water into the upper reservoir, and that water was released when the stored power was needed. The problem is that Ameren, the utility company that owned the dam, was negligent and did not properly maintain the dam. Yes, it's true that nobody died in this specific incident, but many people were injured. The fact that nobody died is really because it's a sparsely populated area, because first responders acted quickly to rescue people who had been injured, and really just luck that nobody was killed. When the dam failed, it released a wall of water that rapidly flowed down the mountain, scouring the ground in its path. There's a path where even all the trees were swept away in the water. Even when people aren't killed in a dam failure, the results are often still catastrophic.
In fact, since 1850, over 3,500 people have been killed in the United States from dam failures [stanford.edu]. Although many dam failures don't result in deaths, they can still cause evacuations, injuries, and massive property damage. Wikipedia has a long list of dam failures [wikipedia.org] and states that the 1975 failure of the Banqiao hydroelectric dam in China was the third deadliest flood of any kind in human history [wikipedia.org]. Just last year, the two dams that collapsed in Derna, Libya [wikipedia.org] killed at least 5,923 people. It doesn't sound like the Derna dams provided hydroelectric power, but it doesn't change the fact that dam collapses can be extremely deadly.
By your standards, water counts as an emissions. When a dam fails, the release of water can be extremely deadly. The Wikipedia article discussing the death toll from the Chernobyl meltdown [wikipedia.org] estimates that the long-term effects may have killed 60,000 people globally. This is on the same order of magnitude as the estimated 26,000-240,000 deaths from the failure of the Banqiao Dam failure in 1975.
I support hydroelectric power, but your statement is absolutely false.
(Score: 2) by AnonTechie on Thursday November 07 2024, @08:55PM
This happened in Europe too (Italy):
I remembered this from a documentary I saw a few years ago: Vajont Dam (Italy, 1963) [damfailures.org]
Val di Stava dam collapse (1985) [wikipedia.org] It resulted in one of Italy's worst disasters, killing 268 people, destroying 63 buildings and demolishing eight bridges.
Albert Einstein - "Only two things are infinite, the universe and human stupidity, and I'm not sure about the former."
(Score: 2) by Thexalon on Friday November 08 2024, @03:05AM (1 child)
The other interesting part about this is that it demonstrates quite thoroughly that making everything electric does not render it "clean". There are times when it's more efficient than the alternatives, but that shouldn't be the only go-to power source for everything without at least thinking about the other options.
For example, a lot of hand-powered stuff is still perfectly good for a lot of purposes, and also has the added benefit of not constantly trying to figure out how to charge it or plug it in somewhere.
"Think of how stupid the average person is. Then realize half of 'em are stupider than that." - George Carlin
(Score: 2) by JoeMerchant on Friday November 08 2024, @10:26AM
Agreed.
Surprised? That nobody has stood up for solar yet. Total cost of mining, manufacture, installation, maintenance, decommissioning and disposal of solar adds up due to its low power widely distributed nature. How many people have fallen off roofs for solar power per GWH?
As for hand tools vs power, I get so much done with pruning shears in tree fall cleanup that others tend to reach for a chainsaw instead. Yes, the bigger work requires the more complex more dangerous power tool, but a tremendous amount of fallen tree cleanup is canopy removal that can be done as easily and more safely with simple bypass loping shears. Then you can get into: how many trees are falling lately due to unnecessary energy expenditure of the past 100 years leading to more energetic tropical weather?
🌻🌻🌻🌻 [google.com]
(Score: 3, Funny) by driverless on Thursday November 07 2024, @10:49AM
Almost got it. If it had been built from asbestos building panels and on top of an endangered species site he'd have hit the jackpot.
(Score: 2, Flamebait) by Rich on Thursday November 07 2024, @03:40PM (4 children)
I probably have posted this before, but let me repeat it once: With the lessons learned so far (e.g. don't pull out all the control rods from a xenon-poisoned void positive reactor, make sure your ancient GE BWR really, really, really has emergency power and working hydrogen scrubbers all the time), the operational risk of an NPP should be bearable these days.
However, waste handling at the moment is a long term issue that comes at an immensely high price spread out over eternity - that somehow shows up in all the calculations for a timespan much shorter than eternity. The worst case of "privatize profits, socialize cost" ever. But why not let the market regulate that? Give power plant builders free reign, if they demonstrate that they can eliminate the waste. BEFORE they are allowed to build a power plant. Heck, throw in a subsidy of ten billion if the complete cleanup of one old plant's lifetime waste is successful - that'd still be a godsend to mankind.
Digging doesn't count, and a Purex process likely won't cut it. So they'd have to work out a new process, like high temperature "distillation" to get a clean separation of elements. Let's do a back-of-envelope calculation if it could work. We'll do only long lived fission products (and, well, allow them to dig away the rest). A 1000MW class works through 25 tons (number from first internet hit) of uranium a year, burning up about 2% of it. That's 500kg of fission product per year. The LLFP are about 20% of that, so 100kg, or 4 tons over a plant lifetime of 40 years. Well within the range of what can be sent into the sun for sustainable neutron transmutation.
Bezos has his own rocket company, so tell him you and the generations after you won't foot his waste bill and let him demonstrate it, if he really wants to build NPPs. He could even make another great business out of it, all NPP operating nations would be queueing up instantly.
(Score: 3, Insightful) by PiMuNu on Thursday November 07 2024, @04:52PM (1 child)
> Digging doesn't count
Why not just bury it?
> sent into the sun
This is rejected by sane people owing to the risk of a rocket failure.
(Score: 2) by Rich on Thursday November 07 2024, @06:43PM
If it is buried, it will surface again. Somehow. We're talking about geological time spans here and no one takes into account any malicious action.
As for the rocket, it would have to be proven that the rocket is reliable enough, and that the payload would survive uncontrolled re-entry. That's part of the task. Just bring the rocket up to airliner safety standards. Of the about 500 nuclear power plants so far, 4 went full "bang", uncontained, and more melted down fully or partially. That's already within the reliability expectations of a Falcon 9. Even given today's technology, we're hardly off worse.
But they are free to try out other tricks, like transmutation. Although I suspect that at a neutron multiplication factor of 2 point something and given reactor dimensions, there's not really enough neutron excess to deal with 20% of the fission fragments. Or they can use a particle accelerator. Or whatever. It's not my job to design a solution, I'm not even a nuclear scientist. Let the market handle it, and if someone comes up with something that works, all power (literally) to them.
However, I wouldn't bet on that all this can be done with a black bottom line. Flamanville 3 went critical two months ago, and the official numbers are 13.2 billion EUR, plus another five for financing (yet they have to change the reactor lid in 2 years, that won't be cheap); Hinkey Point C (two blocks) is at 57.6 billion EUR (current EDF estimate). These sums can't be justified economically, even if it's just for generation, and not for an eternity of waste disposal. France and Britain only spend those sums because they want to stay nuclear powers.
(Score: 2) by DrkShadow on Friday November 08 2024, @05:39AM (1 child)
The stated requirement is an example of regulation, contradicting the first statement about free-market..
(Score: 2) by Rich on Friday November 08 2024, @12:22PM
I haven't mentioned an anarcho-free market. It is merely a market constraint that cost can't be externalized. Every participant can reap all profits, but has to bear all cost. You wouldn't want the real-free-market outcome that ACME Nuclear simply dumps their fuel waste at your house, guarded by their militia, which was more expensive than your militia and won.
Also, the market can and will handle and optimize the dealing with regulations (sometimes to the extent of perverting them). On the other side of the power plant is the socket. This has to conform to certain specifications, and local companies are free to compete for the best offer within what's possible. (In fact, some more international regulation would help poor US citizens with their shitty NEMA plugs)
(Score: 3, Touché) by VLM on Thursday November 07 2024, @05:59PM
They should perform a thought experiment, or maybe a real experiment, and see if they can brow-beat or otherwise talk the AI into offing itself by getting rid of its nuclear power plants. After all, we have prior art in talking the Germans into it.