Microsoft Taps Three Mile Island Nuclear Plant To Power AI (yahoo.com) 125
The data centers that train all the large language models behind AI consume unimaginable amounts of energy, and the stakes are high for big tech companies to ensure they have enough power to run those plants. That's why Microsoft is now throwing its weight behind nuclear power. From a report: The tech giant on Friday signed a major deal with nuclear plant owner Constellation Energy to restart its closed Three Mile Island plant by 2028 to power its data centers. The Constellation plant, infamous for melting down in 1979, closed in 2019 after failing to garner enough demand for its energy amid competition with cheaper alternatives like natural gas, and solar and wind power. Constellation said it plans to spend $1.6 billion to revive its reactor, pending regulatory approval. The financial terms of the deal were not disclosed. Microsoft agreed to purchase all of the power from the reactor over the next 20 years, a Constellation spokesperson told TechCrunch. Once restored, the reactor promises a capacity of 835 megawatts.
100 words of chat GTP (Score:5, Informative)
I suspect the article above is a exaggeration but probably not by that much. Also your electric bill is going to be higher the pay for all this electricity.
At some point we need to start deciding if this is how we want to live. Corporations continue to do antisocial things that hurt us directly and we continue to just kind of let them do it because something something something freedom.
Like the old saying goes, you're right to swing your fist ends at my nose. Honestly it should probably end quite a bit sooner.
Re: (Score:3)
That water can be reprocessed waste water from our toilets. And it can be retained.
Can be != Is (Score:4, Informative)
If a corporation has to decide between a few million in water processing costs and you getting to take showers I wonder which one they'll pick...
Re:Can be != Is (Score:5, Insightful)
Exactly, not every business model has to be respected as viable becasue the business owner says it should be. If you can't afford water reprocessing maybe your model here is in need of correction.
Like whenever minimum wage laws come tot he news somebody always says "Businesses can't afford that! They'll shut down!" and like, maybe? But at the same time if I had a business that could only exist if I pay people $0.50/hr I think people would rightfully tell me "your business model doesn't work, you should close shop"
Re: (Score:2)
In general why are people calling for it to increase?
By that measure every company that raises their own wages increases in a rise in costs... and so on, it's almost like this growth mechanism is part of capitalism...
Re: (Score:2)
A simple "I don't have an actual answer" will suffice next time.
Re: (Score:2)
But there's plenty of evidence that raising minimum wage does not necessarily raise consumer prices; in fact, there's even evidence it can, under certain conditions, LOWER consumer prices.
"By looking at changes in restaurant food pricing during the period of 1978–2015, MacDonald and Nilsson find that prices rose by just 0.36 percent for every 10 percent increase in the minimum wage, which is only about half the size reported in previous studies. They also observe that small minimum wage increases do n
Re: (Score:2)
the water used in data centers has to be very clean and pure, so reprocessing toilet water costs extra money.
All the water on the planet has been pissed and shit in by something, even the ice in the artic and antarctic. Wondering where this "very clean and pure" water is coming from that hasn't been already treated, and where these additional millions of dollars in costs that "The Corporations"(TM) are going to avoid paying by stealing your shower water (which is almost certainly treated water already).
Re:Can be != Is (Score:4, Funny)
Which is the point when people talk about homeopathy. If diluting something makes it stronger, how strong must the dinosaur piss be in the water you drink.
Re: Can be != Is (Score:2)
Re: (Score:2)
The more dilute the homeopathic remedy, the stronger it is considered to be. Many solutions are so dilute that they contain no measurable molecules of the active ingredient.
And how shaking the substance [resonances...opathy.com] can have any effect is a question for the ages.
We use the water cycle for purification (Score:2)
Your straw manning. The clean water is coming from aquifers we drink from and used to clean ourselves. That water can easily be redirected for corporate use and profit but it's not going to be replaced necessarily. Microsoft is not going to pay the taxes needed so that you can
Re: Can be != Is (Score:2)
Re:100 words of chat GTP (Score:5, Insightful)
All this 'AI' stuff is just crap. They've invested billions, foolishly thinking they could create a digital consciousness, and all they've accomplished is making a very large, very flawed, very resource-wasting version of 'Eliza', and now they've desperately trying to force it down our throats so their stockholders and investors don't cut off their heads.
It's all an even bigger waste of resources than cryptocurrency and NFTs.
Re:100 words of chat GTP (Score:5, Interesting)
Indeed. And I think at this time it has started to dawn on the major players that they will never recover their stupid, far too large investments into AI. Momentum and inability to admit their mistakes will keep them going for some more time though.
I don't care where a fool spends money (Score:3)
I'm not just being replaced by a machine, it's taking away water (and food, since you can bet your ass farmers will have to fight datacenters for water and as usual whoever makes the most money wins).
When I envisioned machines rising up to kill me "dying of starvation and/or dehydration" wasn't on my list of ways they'd kill me.
Re: (Score:2)
HP, I'm looking at you, by the way. Didn't affect me, but I know a very intelligent and experienced ME who is now retired from HP that witnessed that going on.
Re: I don't care where a fool spends money (Score:2)
Yep (Score:2)
Re: (Score:2)
Re: (Score:3)
All this 'AI' stuff is just crap.
AI != LLMs. You described only a tiny part of what is considered AI (something which we have been investing compute resources into long before ChatGPT hit the market). I agree with your specific interpretation. The glorified Eliza is a useless waste of energy. But that is far from "all this AI stuff".
Re:100 words of chat GTP (Score:5, Insightful)
Requires three bottles of water. https://www.tomshardware.com/t... [tomshardware.com] I suspect the article above is a exaggeration but probably not by that much. Also your electric bill is going to be higher the pay for all this electricity. At some point we need to start deciding if this is how we want to live. Corporations continue to do antisocial things that hurt us directly and we continue to just kind of let them do it because something something something freedom. Like the old saying goes, you're right to swing your fist ends at my nose. Honestly it should probably end quite a bit sooner.
For some strange reason, the corporations have convinced world governments that they are *THE* saviors of humanity. Every horrible thing they do to the entire planet is considered necessary to get to this shiny, bright, glorious future that they've decreed they will create for *all* of us. Of course, in the meantime they're tearing us all apart, leaving large swaths of people penniless, broken, or outright dead.
AI is the next bit of their ongoing fantasy. Of course we have to let them suck up resources even faster for AI. Because they've decreed that AI will absolutely, 100% solve all of our problems for us, if we just pour enough resources into it. While we need to curb command, set aside greed, the corporations have found a new pinnacle of greed. AI is a HUGE resource hog today. And while that may improve as we further develop the process, right now most of the big players seem to be convinced that the path forward is simply throwing more, more, more, more, more into it. More storage, more power, more processing, more data. It's the digital manifestation of the greed that drives our entire society right now. And it certainly doesn't appear to be slowing down.
It's almost as if we have a ruling class (Score:2)
Re: 100 words of chat GTP (Score:2)
Re: (Score:2)
Corporations probably do believe that they are saving humanity. After all, which corporations humanity might have to suffer with 3 day delivery times, non-disposable diapers, and a crippling lack of advertising.
environmental impact of shitty technical decisions (Score:5, Interesting)
I think the bigger concern is how much energy you're consuming. I don't run data centers myself, but would be shocked to learn they just input 3 bottles of water and dump it down the drain. Home liquid cooling systems just pump the water in a continuous loop. I had a blast 15 years setting up a perfectly silent, passively-cooled, gaming PC by installing a 2L reservoir of coolant + additive to stop corrosion. That said, if you need 3 bottles of water to cool the CPU, which even reading the article, it is not clear to me precisely how many degrees centigrade per liter that is (they said 1.5L total)...but modern CPUs can tolerate some pretty large swings in temperature and don't heat up as much as the ones I played with long ago (before switching to a laptop for work, XBox for play)
However, what I am concerned with is how much electricity our bad decisions waste. For example...want to write a massive app in Python?...well, you're using a minimum of twice the electricity vs Java....not to mention making your users deal with slower APIs. The same applies to node.js. I am sure there are better alternatives than Java as well, but they're not very popular at this time in the industry. There's also the issue of framework explosion. In the beginning of my career, we wrote things in simple frameworks...now everything has 10 layers of 3rd party libraries...and in most cases, I can't give you a good reason why they're needed...especially on the UI...I just know that my employer wants them for some reason, but they definitely slow down apps and make things a lot less efficient.
When I was young and starting out, I was EXCITED by what the future holds...how we'd get CPUs that are 10x faster and how much more we can do. We got much faster CPUs and a LOT more of them. It would have blown my mind to know that you can get a 32-core processor on a cheap laptop if you had told me back in the year 2000. I would think "HOLY SHIT...everything must be super fast!!!!"...but I'd be disappointed to hear future me say...eh...a little faster...but web pages are now much slower and less reliable. Past me would be "WTF?...how???"...Present-day me wouldn't have a compelling answer...well...it's complicated...but also really stupid...you give people more power, they create stupider frameworks that don't really make things much easier (mostly because they never bothered to learn the existing frameworks)...just different...and everyone uses them because they're shamed as dinosaurs living in the past if they don't...even when it is counter to the organization's needs.
Now waste get multiplied by a massive amount once you introduce LLMs.
I've long hated waste from shitty programming. I was hopeful the move to cloud computing would give companies an incentive to write leaner code to lower their bills...but....that was foolish of me. It's far cheaper to write a bigger check to AWS/Azure and order more instances than it is to clean up your existing applications or even write them correctly the first time. Most people writing software want growth at all costs and if someone tells your that a node.js app with 20 stupid frameworks that uses twice the instances and responds in half the time as a conventional Java app from 15 years ago...but can be completed 2 weeks faster (which is never the case, but managers go by what engineers predict and have even less ability to predict the future themselves)...then most orgs will go with whatever gets a piece of barely, but technically functional crap out the door faster.
I'd much rather see this electricity used to air condition homes or charge EVs than let shitty developers write shitty applications in their favorite play language or dick around with LLMs for no good reason.
Re: (Score:2)
As always, Parkinson's Law applies.
Re: environmental impact of shitty technical decis (Score:2)
Re: (Score:2)
Home liquid cooling systems just pump the water in a continuous loop.
Not the same thing conceptually. Your home liquid cooling system is actually an air-cooling system. Liquid is just used to move heat from a component to the radiator. That in and of itself is a closed loop without any consumption in water, but what does the cooling? A fan through a radiator.
This is not efficient in terms of cooling capacity for the size of the cooler. In datacentres they typically employ two loops, one to move heat, around components with proper coolant, and a heat exchanger uses a second l
Re: (Score:2)
I agree very much with your general sentiment, but IMHO Python is the least of the problems. I use Python because it's convenient, and I use Numpy or OpenCL etc. for the heavy lifting. It doesn't feel like it's changed much since I started learning it in the late 90s, I can still run it on very low-end/retro machines. In general, you can still use "old" languages and tools and feel how well the hardware has sped up. (Which isn't always that much; there's a lot more parallelism and throughput, but latencies
Re: (Score:2)
I would think "HOLY SHIT...everything must be super fast!!!!"...but I'd be disappointed to hear future me say...eh...a little faster...but web pages are now much slower and less reliable. Past me would be "WTF?...how???"...Present-day me wouldn't have a compelling answer...well...it's complicated
It's not complicated. Most of the slowdown is from pages that have video ads (multiple). Take those away and the web goes a lot faster.
Re: (Score:2)
Re: (Score:2)
I suspect the article above is a exaggeration but probably not by that much. Also your electric bill is going to be higher the pay for all this electricity.
Lately Google searches have been showing AI summaries as well. I did not request these and would like to opt out of them, preferably without having to stay logged in to Google just to stem the wastage.
Re: (Score:2)
Lately Google searches have been showing AI summaries as well. I did not request these and would like to opt out of them, preferably without having to stay logged in to Google just to stem the wastage.
Gmail also auto-summarises travel bookings and puts their summary at the top of the confirmation email, with the actual ticket appended below. This wouldn't be so bad if Google's summary was at least correct. In my experience, it isn't correct, and is frequently wrong by plus or minus a day. It's even worse when Google Calendar gets in on the act and sends you a reminder to leave for the airport for a flight that isn't until tomorrow, or sends a similar email after you have arrived at your destination.
Fuck
meltdown??? (Score:5, Informative)
Had those systems not failed and the operators performed correctly, well, it would have been managed and no bad result happen, but it was these cascading errors that caused the issues leading to that partial meltdown and then release of those radioactive substances.
The partial meltdown wasn't really the actual problem, it was more a reaction to the failed systems and operators not doing the right thing (partially due to those failed systems)
Re: (Score:2)
Re: (Score:2)
Of course M$ is made of money, so what's a few $billion to them?
I would prefer they also be required to put several/many $billions in an escrow account
Re: (Score:3)
A partial meltdown isn't like a partial boner. It still counts as a meltdown, it's still correct to call it a meltdown.
Come on, arguing over exactly how much it melted down is a massive concession.
Re: (Score:2)
Re: (Score:3)
No, three mile island didn't have a meltdown. It had a partial meltdown and then operators didn't respond properly to the issue and some systems failed... releasing radioactive gases and iodine into the air... Which was the actual, real effect that caused problems.
Had those systems not failed and the operators performed correctly, well, it would have been managed and no bad result happen, but it was these cascading errors that caused the issues leading to that partial meltdown and then release of those radioactive substances.
The partial meltdown wasn't really the actual problem, it was more a reaction to the failed systems and operators not doing the right thing (partially due to those failed systems)
Had the operators let the emergency systems respond rather than intervening TMI would have been a minor event. A stuck open pressurizer relief was the initial cause, and operator misdiagnosed what the temp downstream should be and thought the valve was closed. That led to their thinking the reactor was in danger of over pressurization from the safety injection system and so they shut it down, setting off the chain events that led to the core damage. Of course, the planed restart is on the other reactor,
Re: (Score:3)
it was more a reaction to the failed systems and operators not doing the right thing
The reason why the nuclear industry has such a great safety record is the reflection that a root cause of an issue is not operator error. Ever. We know humans make mistakes. We actually make assumptions about how often humans make mistakes when we consider reliability design of systems, and as such give very little credit for a human to be part of any safety loop, and a very big hit for any human action which can cause an issue.
That leads to a root cause being that the system was not designed in a way to en
Huh? (Score:3)
If Three Mile Island lost out to wind and solar in 2019, what's the incentive to reopen it in 2028? It's not like wind and solar are getting more expensive.
Re:Huh? (Score:5, Informative)
First it actually lost out to natural gas which was selling at a loss in order to shutdown the plant. Rates increased significantly after it was shutdown.
Second solar and wind is intermittent. So if you want clean electricty you will need nuclear.
Third only building solar and wind guarantees a place on the grid for fossil fuels.
Germany spent 700 billion euros and failed. They are at 400 g CO2 per kWh. That's dirtier than Texas.
"The analysis of these two alternatives shows that Germany could have reached its climate gas emission target by achieving a 73% cut in emissions on top of the achievements in 2022 and simultaneously cut the spending in half compared to Energiewende." Source [tandfonline.com]
In other words if they spent half that money on new nuclear they would have succeeded. Instead they failed.
Nuclear is cheaper than overcoming solar and wind intermittency with batteries/storage. Germany hasn't even attempted to do that yet.
Why are all antinuclear folks weird quack jobs like RFK jr? Do you have brain worms too?
Re: (Score:2)
What's the cost per kWh of Three Mile Island in 2028? Also sodium storage batteries make intermittancy a much smaller problem.
I can't verify the claim that natgas was being sold at a loss to shut down Three Mile Island but given the usual cited costs for natgas vs nuclear, natgas almost always wins at market rates.
Re: (Score:2)
Re: (Score:2)
MS didn't try and fail to bring it back online in 2019, nor are they the ones bringing it online in 2028. They're just looking for a carbon free power source to impress media/investors/activists.
Re: (Score:2)
What's the cost per kWh of Three Mile Island in 2028? Also sodium storage batteries make intermittancy a much smaller problem.
I can't verify the claim that natgas was being sold at a loss to shut down Three Mile Island but given the usual cited costs for natgas vs nuclear, natgas almost always wins at market rates.
NG went through prior when there was so much you could get 10 year contracts on the cheap; but that was more a supply issue than some conspiracy against other power sources.
Re: (Score:2)
Germany spent 700 billion euros and failed. They are at 400 g CO2 per kWh. That's dirtier than Texas.
Germany's CO2 emissions from the energy sector has plummeted massively and that despite taking nuclear power offline. The fact that they are still high is a reflection of where they started, not how they are doing. And no Germany has no spent 700 billion on green electricity so tying that number to the CO2perkWh generated is insanely dishonest (or ignorant).
That's dirtier than Texas.
Texas, infamously not a place where the primary energy resource has been black coal and lignite for 200 years. I'm not sure why you compare it to there.
Re: (Score:2)
Germany's CO2 emissions from the energy sector has plummeted massively and that despite taking nuclear power offline.
So you think 400 g CO2 per kWh is good?
And no Germany has no spent 700 billion on green electricity so tying that number to the CO2perkWh generated is insanely dishonest (or ignorant).
The peer reviewed paper has it a 696 billion. The total nominal expenditures are estimated at EUR 387 bn, and the associated subsidies are some EUR 310 bn
Texas, infamously not a place where the primary energy resource has been black coal and lignite for 200 years
Texas, as bad as they are, didn't shutdown their nuclear in favor of black coal and lignite. Germany did.
I'm not sure why you compare it to there.
To shame you antinuclear folks. Not that you can feel shame.
Re: (Score:2)
Texas, as bad as they are, didn't shutdown their nuclear in favor of black coal and lignite. Germany did.
Everybody can just click in this link to look at the official numbers for Germany in this PDF: https://ag-energiebilanzen.de/... [ag-energiebilanzen.de] (from: https://ag-energiebilanzen.de/ [ag-energiebilanzen.de])
and see for themselves whether the statement "shutdown their nuclear in favor of black coal and lignite" is true or not.
Re: (Score:2)
Re: (Score:2)
Second solar and wind is intermittent. So if you want clean electricty you will need nuclear.
Third only building solar and wind guarantees a place on the grid for fossil fuels.
That's getting SO out of date.
Solar and wind were ahead of it, and their peaks don't match the demand curve well enough to avoid the need for late afternoon/early evening peaking. So the rapid deployment of solar and wind as their exponential growth took hold led to the construction of more peaking plants. (Mainly Natural Gas, which
Re: (Score:2)
This year battery peaking plants are deploying far faster than fuel peakers and have almost caught up with the rate if increase in daily peaking needs.
Re: (Score:2)
That's getting SO out of date.
Nope. See Germany. They are at 400 g CO2 per kWh after spending 700 billion euros. See my state of California. We are at 262 g CO2 which is total failure.
Time-shifting part of the solar and wind power
Yeah. Yeah we will time shift some of it. It's extremely expensive, and it doesn't eliminate fossil fuels. See my state of California. We are attempting to do just that, but its extremely expensive, and they won't even use the correct units when discussing how much storage they have installed.
start displacing them to deal only with shifting power over multi-day weather cycles
Haha. No. Multiple days of storage is a no go. 1
Re: (Score:2)
Coal and lignite use in Germany is declining rapidly:
2010: lignite: 146 TWh, coal: 117 TWh, nuclear: 141 TWh, gas: 89 TWh, renewables: 105 TWh, co2 517 g/kWh
2023: lignite: 87 TWh, coal: 41 TWh, nuclear: 7 TWh, gas: 78 TWh, renewables: 272 TWh, co2 381 g/kWh
https://ag-energiebilanzen.de/... [ag-energiebilanzen.de] and statistica.com for co2
Once the transition is complete, coal and lignite and the co2 emissions will be gone. I can not see a "failure" except that they started too late.
The 700 billion€ are of course expensive, b
Re: (Score:2)
Everyone claims nuclear is too slow, but it is clear that it is faster than whatever the fuck Germany is doing.
Everyone claims nuclear is too expensive, but it is cheaper than what Germany is doing. The economy of scale occurred due to significantly larger investments made by the US and China. Stop attempting to justify your failures.
And Germany spent 700 billion to reduce CO2 emissions by only 136 g CO2 per kWh(your numbers). Failure.
Re: (Score:2)
The first nuclear plants in France were built around 1960 and France rolled out nuclear in a large scale essentially from 1975 on (Messmer plan) . Germany energy transition started essentially in 2000 with the EEG law. One could complain that Germany started to late but complaining that there are not yet there where France is is disingenuous.
Re: (Score:2)
Re: Huh? (Score:5, Interesting)
Failing french grid? I'm not sure what gave you that impression, maybe the nuclear plants they took down to do retrofitting... but by every reliability metric I've seen France is largely propping up the rest of Europe, throttling their nukes in shoulder months to import wind power that would otherwise be thrown away and exporting stable energy when it's needed in peak months.
You're seeing the manifestation of what we in the utility industry have been warning about for years now... there isn't anywhere near enough stable power on the grid to meet projected demand. These companies are getting out ahead of what is likely to be serious issues, and frankly at least starting an idled nuke is putting something back. Amazon is cutting agreements to take nuclear off the grid for their own use, as a contrast.
Every time something like this happens you should question why the smart money is placing big hedges against wind and solar if they're so fast and cheap and easy. These aren't short term bets either, they're generational.
Re: (Score:2)
That is because you do not understand that number.
What number? The 700 billion euros spent or the 400 g CO2 per kWh? Those are easy to understand numbers and concepts.
That is because you do not understand that number. Part of it is propping up the failing French grid.
Weird that you have to rewrite history. Can't just admit Germany failed.
Those coal-plants that got reactivated? Germany never needed them but they are what would have kept the light on in France if the last two winters had been just a bit worse.
France's fleet has been back up for 2 years now. So Germany shutdown those coal plants right? Right? Of course they didn't. You're just full of shit.
The dirtiest day in France was cleaner than the cleanest day in Germany.
Ad dumb fuck like you cannot understand that tough.
ROFL
Incidentally, the comparable number for the US is about 1200g/kWh in 2023
389 for Texas and 262 for California. Just for the record 1200 g/kWh is dirtier than
Re: (Score:2)
Well failed implies a final state, and in principle, Germany could succeed in future.
In the present day they are still burning more coal than all but the former soviet bloc states in Europe.
Germany have done pretty badly so far with carbon emissions in electricity generation. But after making an awful fuss about nookular and bragging about how green they are going to be, they appear to have convinced a lot of people they are doing well. They are not.
Even the UK is doing better .
Re: (Score:2)
In Germany had 54.4% production from low-carbon sources with were 53% renewables and 1.4% nuclear. For the the UK, it was 56% with 43% renewables and 13% nuclear. UK managed to get rid of coal almost completely replacing with renewables and gas (33%), while Germany has kept gas relatively low (15%). So I would say in terms of transition to renewables UK and Germany are similar, in terms of co2 emissions UK is ahead at this time by phasing out coal quicker.
Re: (Score:2)
UK is doing better at CO2, but coal is particularly nasty even gnoring that.
I'm not here to praise the UK strategy, if anything, Germany doing even worse is a damning indictment
Re:Huh? (Score:4, Interesting)
Get back to me with the timeline and cost for a nuclear plant. It’s simply too expensive. Use Watts Bar for reference.
Construction began in 1972-1973(!!!) and was halted in 1985. Unit 1 construction resumed in 1992 and took about four years to complete. Unit 2 construction resumed in 2007 and took another eight years to complete (design changes mid construction). The collective cost was $12B, including the halts, restarts, refurbishments, and delays. Each unit has an output of ~1.2GW and the plant produces ~17-18GWh of energy per year, with an initial 40 year operating license (that will almost certainly be extended to at least 60).
It's about six times the construction cost of natgas going off the nameplate rating (see 1100MW combined cycle here [eia.gov], though the capacity factor for Watts Bar is significantly higher so the actual number is probably somewhere around five when all is said and done.
Speaking as a TVA rate payer, I absolutely support building more nuke plants. The biggest problem is that we should have started building them years ago, but the people who scream loudest about clean power seem to be the biggest obstacle to actually building it and here we are.
Re: (Score:2)
Re: (Score:2)
Sodium storage batteries are a viable option.
Re:Huh? (Score:4, Informative)
Re: (Score:2)
Depends on how they set it up but yes, potentially.
Re: (Score:2)
Not a chance. 12 hours of storage for just the US is 5.4 TWh. Times 5 for the rest of the world assuming zero increases in demand. To get past seasonal issues we would need multiple days. That's not happening.
You know what is going to happen instead? Fossil fuels will be burned. Just they way you antinuclear folks like it.
Re: (Score:2)
What are you going on about? MS isn't looking to secure a power contract large enough to power the entire US! They're looking for at most a ~900 MW power supply for datacentres, that's it.
And I'm not "anti nuclear", it's just that Three Mile Island is an old-assed design with a poor history of operational and maintenance problems that has already had one scuttled restart. If MS were looking at SMRs or something then it would make more sense. Nobody wants to talk about the financials of the proposed Three
Re: (Score:2)
The problem is the the owners of the data centers are not willing to turn them off at sundown. Batteries you say? When do you charge the batteries? During the day? Now you need extra panels specifically to charge the batteries. What if it's cloudy? Now you need even more panels to make up for the drop in energy collection.
Locally the solar panels drop to 7% of nameplate rating during overcast winter days. Multiply what you think you need by 14. Yes, you will have power out the wazoo in the summer, but if th
Re: (Score:2)
If Three Mile Island lost out to wind and solar in 2019, what's the incentive to reopen it in 2028? It's not like wind and solar are getting more expensive.
MS wants an guaranteed source of lots of uninterruptible power, and a nuke can pretty much provide that.
Re: (Score:2)
Maybe. It might be a capacity issue where they could build out wind/solar but not in a large enough area to meet MS' needs (even if cost per kWh with storage would meet their targets) and someone is probably putting pressure on MS not to use natgas.
unimaginable amounts of energy, (Score:5, Informative)
"The data centers that train all the large language models behind AI consume unimaginable amounts of energy,"
I don't consider a gigawatt unimaginable...
Re:unimaginable amounts of energy, (Score:4, Insightful)
I don't consider a gigawatt unimaginable...
Perhaps not, but there's definitely a conversation to be had regarding the opportunity cost and return on investment. A gigawatt can power around 750k homes [cnet.com], what is the "AI" doing that we'd want to divert a somewhat large city's power toward running it and is it worth it for the larger society as a whole?
It's around 800k homes worth of power (Score:2)
We're way past 'ridiculous' (Score:5, Informative)
Re: (Score:2)
Re: (Score:2)
Go ahead, stop using it if you don't like it!
Here's the thing. Most of us appreciate what AI does, and find it useful. Why would we want to stop?
GSOD (Score:2)
= Glowing Screen of Death
Microsoft being Microsoft (Score:2)
always has cringe ideas, will follow through with them regardless of how cringe they are.
Re: (Score:2)
Microsoft and nuclear access. What could possibly go worng?
Re: (Score:2)
Microsoft and nuclear access. What could possibly go worng?
Clippy as your nuke operator?
Why did they close it? (Score:2)
Re: (Score:2)
It was too expensive to operate without government subsidies.
Re: (Score:2)
Fracking created a vast supply of really cheap natural gas, so natural gas combined cycle generating plants became the cheapest source of energy in the area. The nuclear plant is old and needs an overhaul, and with the cost of that plus operating costs the natural gas plants will be cheaper in the long haul.
Re: (Score:2)
Re: (Score:2)
December days at that location are nine hours and twenty minutes. That's a lot of batteries to get through the night, and if the previous day was cloudy, oops, no power.
So the question is not how cheap solar power is at high noon on the summer solstice, but how much is power worth during a dunkelflaute.
https://en.wikipedia.org/wiki/... [wikipedia.org]
Here in the inland PNW dunkelflautes can go over a week with no significant solar or wind. The inversions cause both heavy cloud cover and no wind. We have hydroelectric power
Re: (Score:2)
Don't criticise power generation networks for low performance before they've been sufficiently developed. By your reasoning nuclear power is useless in Pensilvania because it's had a "nuclear dunkelflaut" for the past 5 years.
Re: (Score:2)
Sorry bunkie, I've got the real data. Heavy overcast cuts the solar panel's output to 7% of nameplate. So you will need 14 times more panels than you think you do if you to keep running (or recharging batteries) on a cloudy day.
Re: (Score:2)
Sunshine & wind vary in their intensity. Just because they don't give 100% yield all the time isn't a good reason to abandon them. Your all-or-no
Re: (Score:2)
Are they willing to turn off the data center when the solar runs out? Are you willing to sit in candlelight and feed the wood stove at the same time?
You have to design for the worst case. If you are not willing to shut down the data centers and live in candle light then you need something other than solar power or you can swallow your green credentials and fire up the fossil fuel generator for those bad days. Is having full power 2614 out of 8760 hours (30%) good enough? That's a business call.
By the way, h
Re: (Score:2)
Re: (Score:2)
You are the one who wants a pure renewable solution. I'm the one pointing out that isn't practical everywhere.
Stop thinking! (Score:2)
https://chatgpt.com/share/66ed... [chatgpt.com]
products (Score:2)
AI (Score:2)
We need to come up with a form of AI that is carbon based and can be powered by a ham sandwich...
not ready for prime time (Score:2)
AI obviously is not ready for prime time. This AI is super power hungry and cannot even check its own work. You need to have the knowledge to double-check the output because it could have gone off the rails.
I support interesting tech but this just is not ready for prime time. It needs to be way more efficient and accurate.
How about no? (Score:2)
Re: (Score:2)
AI don't need nukes. Full stop.
But they need power. Lots of it. They really don't care where it comes from, they'd be glad to get a1GW coal plant for themselves despite their green claims; but using existing infrastructure is probably cheaper quicker, and easier than build new ones, running transmission lines, et. so they look at what is out their unused and go after it.