https://www.theregister.com/2025/05/07/google_signs_another_...
> "Elementl didn't respond to questions by press time. Its public materials offer little clarity on its actual operations—aside from broad claims about providing "turn-key project development, financing and ownership solutions customized to meet our customers' needs while mitigating risks and maximizing benefit."
> "The nuclear developer, founded in 2022, presents itself as a facilitator of advanced reactor projects. But it has not built any reactors to date and describes itself as a "technology-agnostic nuclear power developer and independent power producer," signaling it does not back any specific reactor design."
> "This approach aligns with the background of Elementl's CEO and chairman, Christopher Colbert, who previously served as CFO, COO, and chief strategy officer at NuScale Power."
Holy corporate jargon batman! I love seeing example of phrases like this out in the wild. Stating this implies that minimizing risks and maximizing benefit is not a need of most customers? IMO, it's better not to say stuff like that at all. It's basically a meaningless phrase, it adds no information to the sentence. In fact, I'd go so far as to say it's generally a sign that they are doing the opposite of whatever the phrase means.
It’s not, at least for nuclear power. In Europe, for example, the debate is entirely emotional. So saying they’re working for a rational customer is sort of meaningful, even if corporate speakified.
I believe this should have meaning. It would mean risk mitigation is a primary objective of the company. And not every company decides to consider risk mitigation as a primary objective.
The problem is that risk mitigation is a long term objective. Who has time for that?
I can image that's the stuff kids would say when asked why is the candy bowl suddenly empty: "Well, you see, we were was just meeting our needs while mitigating risk and maximizing benefit".
Honestly, I'd rather them explicitly commit to minimizing risks than say, "We're going to address the needs of our customers, and that probably includes minimizing risks, at least in most cases, right? Product will let us know when they've done the research."
It's better that they say these things than that they don't say them. The real problem is not that they say them, but that we can't be confident they'll live up to them.
NuScale got far enough to get approval to build a test reactor at the Idaho Reactor Testing Station, which is in Outer Nowhere for good reasons. But they never got enough funding to build it.
The trouble with most of these small modular reactor schemes is that their big pitch is mostly "we don't need a big, strong, containment vessel because ... reasons."
There's no inherent problem in building a small nuclear reactor. Here's one from 1957, near Oakland, CA.[1] It's safety if something goes badly wrong that's a problem.
History:
- Chernobyl - meltdown and fire, no containment vessel, major disaster.
- Fukushima - meltdown, too-small containment vessel, large disaster.
- Three Mile Island - meltdown, big strong containment vessel, plant lost but no disaster.
Alternative reactor history:
- Fort St. Vrain - high temperature gas-cooled, subject to helium plumbing leaks in radioactive zone, shut down and plant converted to natural gas.
- AVR reactor, Germany - pebble bed reactor, had pebble jam, had to be shut down, extremely difficult to decommission.
- Sodium reactors - prone to fires.[3]
- Molten salt reactors [4] - require an attached chemical plant that reprocesses radioactive molten salt.
Most of the problems of nuclear reactors in practice involve plumbing. Everything in the radioactive zone has to last half a century or so without maintenance. That's possible with distilled water as the working fluid, but everything else tried has not worked well.
[1] https://www.youtube.com/watch?v=A1O8xAB_FDI
[2] https://en.wikipedia.org/wiki/AVR_reactor
[3] https://www.nuclearsafety.gc.ca/eng/resources/research/techn...
Distilled water is pretty corrosive at high temperatures, isn't it? I'm no engineer but I've read that the water-chemistry management of nuclear reactors is a highly finicky topic.
Here's a crazy fact I can't get out of my head: the PWR types of reactors rely on lithium hydroxide in their nuclear water pipes, as a critical corrosion inhibitor. But the US can't make this (meaning, the isotopically enriched lithium of the correct flavor for nuclear reactors); it imports 100% of this key ingredient from foreign countries— currently, exclusively, China and Russia. Our top geopolitical adversaries could kneecap most of our nuclear power fleet, if they wanted, because of the difficult engineering minutae of "water is corrosive".
True story. https://www.gao.gov/products/gao-13-716
It's worth examining why they never went forward with builds. In 2023 their cost estimates for power went from a manageable $55/MWh to a barely-managable $93/MWh. And that was before all the additional cost increases that are typical for first projects.
They were unable to paint a story that was financially compelling.
Nuclear's problems are not TMI, it's Summer, and other failed builds. The government will insure catastrophic damages. It will not insure against construction cost overruns, and those may not kill people but they kill companies dead.
Probably not the greatest placement in hindsight.
Probably overselling the "large" there... at least on the scale of global power production.
The startups have been bad too, with some disingenuously starting regulatory processes and then not even responding to questions or attempting to follow through.
South Koreas is the most developed nation that has had success building, and even they send people to jail for construction fraud.
There are undoubtedly many honest and earnest people trying to build new nuclear. But it's hard to tell who until after billions have been sunk and misallocated.
But really I cannot emphasize enough how strict and overbearing they are.
"Oh that 12V backup battery pack needs to be replaced? Better get the same one from the same manufacturer"
"They aren't in business anymore but we have this 12V battery the fits perfectly, same specs"
"Nope, not certified with that system. You can start recertification that will cost ~$40M if you like"
"...."
There is so much ass covering and not wanting to take responsibility that the market is basically in paralysis.
I think it's less an issue of anything to do with nuclear in particular, and more that we're just living in an absolute golden age of charlatans. It's like the 1980's all over again except instead of fraud being doable because of a lack of information, fraud is doable because everyone for whatever reason you'd like to describe is thoroughly committed to pretending it's the 1980's.
Yes, even in South Korea: https://en.wikipedia.org/wiki/South_Korean_nuclear_scandal
That's why :)
Russia is also fairly successful at building reactors. Although, somehow their orders pipeline has been getting shorter and shorter (wonder why...).
Missing vowels + no plan + a leadership team stacked with MBAs, investment bankers, and FAMILY MEMBERS?!?= bullshit private equity "play".
Let me predict what is going to happen: a team of connected "playaz" are going to get real companies with cash and the government to give them money to shake up the nuclear market.
Then the leadership team is going to hire a token staff of scientists, engineers, and public policy folks.
They are going to have a groundbreaking ceremony for a facility (not a reactor, like an R&D facility or "rapid innovation incubator" or something) that is highly publicized and subsidized by state and local business development grants and credits but will either never be finished or never fully staffed.
Nothing will happen for four years until they either fade out of existence or declare defeat due to "regulatory and market conditions" with nothing to show but some powerpoints and press releases.
The hundreds of millions of dollars that flowed into the organization is never spoken of or seen again.
Then a couple of years later they'll register another .io domain with missing vowels and start all over again.
Maybe this is related to the talk about splitting Google that's going around these days?
For anyone curious about what happened with that: https://web.archive.org/web/20181117031510/https://blog.veri...
> Our clinical work on the glucose-sensing lens demonstrated that there was insufficient consistency in our measurements of the correlation between tear glucose and blood glucose concentrations to support the requirements of a medical device. In part, this was associated with the challenges of obtaining reliable tear glucose readings in the complex on-eye environment. For example, we found that interference from biomolecules in tears resulted in challenges in obtaining accurate glucose readings from the small quantities of glucose in the tear film. In addition, our clinical studies have demonstrated challenges in achieving the steady state conditions necessary for reliable tear glucose readings.
https://www.chooseenergy.com/electricity-rates-by-state/
Why doesn't the state encourage more capacity to bring costs down? (to encourage electrification/EVs, etc.) Is it because they are phasing out natural gas? Is it to encourage roof top solar? Or trying to reduce consumption by having high prices? Or environmental permitting? "Lobbying" by entrenched incumbents? Or maybe the high price is due to taxes and not the price of generation?
https://www.ewg.org/news-insights/statement/2025/02/pge-reco...
PR staff will talk about the insurance liability and mandated action to improve infrastructure (wildfires keep starting on power lines and then burning down cities) but it's hard to look away from the record profits
I'm by no means excusing pge, they were pretty clearly negligent and failed to meet their obligations. But it's a state-backed operation, which pretty much always means less punishment for failure to operate effectively.
Where did the money go? Paying for wildfire damage.
There's not really any way around that. Capital expenditures are profit.
The costs come from the wildfires and a derelict grid requiring large infrastructure upgrades.
It is a very silly cycle which could be ended by either removing green energy goals so they could improve infrastructure, and to not hold electric companies directly liable for all damage from a fire.
It’s PG&E and regulatory capture. Santa Clara County is off PG&E and has normal energy tariffs.
https://en.wikipedia.org/wiki/2000%E2%80%932001_California_e...
California is raising rates to build out infrastructure for electrification and mitigation of the dangers that now exist due to climate change.
This is a good thing, but will be fruitless unless the US NRC modernizes in parallel with the industry to actually approve a new reactor in less than geologic time.
https://www.powermag.com/vogtle-3-reaches-initial-criticalit...
The identical AP1000 reactors under construction at VC Summer in South Carolina also suffered enormous cost overruns and delays, again not caused by the NRC or lawsuits. The construction problems were so severe at the VC Summer project that the project halted after spending over $9 billion, it led to the largest business failure in the history of South Carolina, and a couple of company executives went to prison for securities fraud:
I guess VOD is new, but does that really demand that amount of compute?
Plot twist: the computer's last act at the end of The Last Question was just an LLM's hallucination.
What if hackers/terrorist attack the power plants?
What if the operating companies values profit over security?
What if an earthquake or Tsunami hits nuclear power plant?
> What if hackers/terrorist attack the power plants?
Wind has down sides like moving parts and requiring giant concrete poors. Birds strikes, noise as well as ground vibration are also issues.
Many birds die as a result of human activity. In the US, the leading cause of these deaths is cats [1]. Cats cause four times more bird deaths than the next anthropogenic cause of death, flying into windows.
Cats cause ~1000x more bird deaths than collisions with wind turbines.
[1] https://www.statista.com/chart/15195/wind-turbines-are-not-k...
Unless you're vegetarian, or vegan, how so?
> Ontario set to begin construction of Canada's first mini nuclear power plant
So the starting stated price is only 20% cheaper than that train wreck. Will love to see how high this number gets given it's a first of its kind.
What's slightly different is the financial risk profile. Failing on a 1x 300WM $5B project is slightly easier than a $18B 1GW project.
My personal hypothesis is that nuclear decisions are made almost entirely along financial lines, instead of the safety concerns that dominate most debate about nuclear.
This is simply false. At this point, its falsity has been sufficiently well demonstrated and communicated that you should have known it was false. If you are not deliberately lying, it's only because you steered yourself away from learning the truth.
If we just sustain human life and pleasure then yeah renewables are probably fine. If we want to pursue highly energy intensive applications and then further if we want to pursue those applications with mobility then we need nuclear.
At this point, that’s sufficiently well known that you should have known it. If you’re not deliberately lying, it’s only because you steered yourself away from learning the truth.
I am not fully detesting nuclear, but I do disagree it a cure to the environment crisis as Solar is plenty and free; as are Wind and Water too.
The risks of what if; and that now we live in such a volatile world. How are you going to convince me it's safe?
How do I know a drone won't strike it in the next war? Some sponsored hack?
Stuxnet was an organised hack that was created to aid destruction to nuclear hardware.
Chernobyl is still unsafe and that's many years ago and was recently damaged again by a drone.
Anecdotally, I live near the Palo Verde nuclear powerplant in Arizona, we receive all of our electricity through a combination of solar (clouds are very rare here) and nuclear. These 2 factors mean energy is abundant in the state, and necessary in the summer for survival; air conditioning is a necessity due to the extreme temperatures in the summer.
The Palo Verde plant was commissioned in the 1980, and provides more power than any other reactor in the US. Since its not located near a body of water, it uses treated wastewater for cooling. It is a Pressurized water reactor design similar to the ones used on Naval vessels, a much safer design than the one used in Chernobyl, and none of which have ever experienced a meltdown or critical failure. Overall, I've never experienced any anxiety regarding the reactor not too far from where I live, it is the least of my concerns.
I believe the future will need to be a combination of renewables, to put all our eggs in one basket in foolish. Smaller and safer self contained nuclear reactors (like the ones used on Submarines) seem very promising for data centers. AI is on the rise, for better or worse, and it's power demands are constantly growing.
The Gen 4 designs, which they also have, are physics safe: literally drop bombs on them and they still won't fail (bombing a nuclear plant in general is an over stated risk for other reasons too). They're building those now too.
But anyway, if anybody (other than the government, which gave up long ago) can pay the upfront costs of nuclear, it is the big tech companies like Google.
> […] Google has set 2030 goals to reach net zero emissions across its operations and value chain, […]
Man, I remember when 2030 seemed like the future. But now it seems downright aggressive. Good luck Google.
If we (the West) had built out nuclear to satisfy our electricity needs, implementing new nuclear power tech as it improved, we could have electricity subscriptions like we have mobile / home internet planes.
You’d just pay for amps, say 50 amp, 150 amp, 300 amp, all you can consume.
But instead we have expensive electricity (at least here in Australia), where your mind is constantly loaded wit being aware of your energy consumption.
https://www.sciencedirect.com/science/article/abs/pii/S03014...
You don’t get ”free electricity” with absolutely massive handouts to the nuclear industry.
Instead renewables and storage are delivering on the ”too cheap to meter” promise.
But that doesn't inform us on what the optimal policy decision is in the current year of 2025 given 2025 prices and time-to-build of the various options.
In Australia renewables have the perfect confluence of multiple factors:
- low seasonable variability of insolation in the north
- high wind speeds in the south
- land availability for solar
- high statistical diversification of renewables due to size
- higher than normal costs of nuclear due to first-of-a-kind costs dominating the total build-out costs due to the small energy needs of the country, and higher labor costs
The CSIRO studied this for Australia and released a report about it. Even when you factor in storage and transmission costs, renewables are significantly cheaper than nuclear.
What exactly are you talking about? It does not sound like it describes the way nuclear power, uranium mining, and nuclear waste storage works.
The real long-term effect of past nuclear accidents is a subject of debate, and the potential worst case a concern for all.
> the only power generation that has a completely closed fuel life cycle
Not at industrial-scale.
> and believe in a technology we don’t have.
https://ourworldindata.org/grapher/electricity-fossil-renewa...
All these nuclear announcements are smoke screens to cover construction of large amounts of gas fired capacity. Anyone expecting dramatic near term increases in electricity demand will need to go with gas (or renewables, but tariffs make that less competitive); nuclear, especially new designs, cannot be rolled out quickly.
Excellent safety, if you ignore Chernobyl, Fukushima, Three Mile Island, the Tokaimura accidents, the Church Rock spill, the beaches near Dounreay, and dozens more.
Nuclear power rarely kills anyone, but when (not if) things go wrong, it tends to create a massive mess which costs billions to clean up - if a cleanup is even possible at all. It is the only power source which has made entire cities impossible to live in.
I personally don't believe this is necessarily a dealbreaker with modern nuclear plants in rich countries, but if you want to convince people of its safety you probably shouldn't be mentioning its historical record.
> a completely closed fuel life cycle
Only if you completely ignore the huge amount of pollution and waste generated by mining, reprocessing, and disposal.
Again, I personally don't believe this has to be a dealbreaker, but the waste generated by the nuclear industry is still an unsolved problem. We've been operating nuclear reactors for 80 years now, but permanent waste disposal and reactor decommissioning is still in its infancy. The current state-of-the-art is essentially "let it rot in place and hope nothing goes wrong while we figure out a way to deal with it". I think it can be solved, but unless we've done so you probably shouldn't make it part of your argument.
> If we (the West) had built out nuclear to satisfy our electricity needs
We did. France hit 80% nuclear, for example. 9% of global power is supplied by nuclear plants. There are over 400 plants currently operational, and 700 have been decommissioned. We aren't on "baby's first nuclear reactor" anymore.
> implementing new nuclear power tech as it improved
We did. It made the plants too expensive to be commercially viable.
> You’d just pay for amps, say 50 amp, 150 amp, 300 amp, all you can consume.
Not a chance. Although fuel would indeed be quite cheap, power still isn't going to be free: someone has to pay off the massive construction loans.
Consumer power consumption is also a lot more flexible than something like internet. People don't suddenly start to consume a lot more data when their internet gets faster - a single person is still only going to watch one Netflix stream at a time, and that'll work just as fine on a 100Mbps connection as on a 8Gbps one. And all the equipment is already prepared for the faster connection, so it's not like they are saving any money by keeping it slow.
But if your power is free, why bother with gas heating? Why go for a heat pump when resistive heating has cheaper equipment? Why bother isolating your home? Why shut off your lights when you leave your home? Making electricity free means we'll be using a lot more of it, which means having to build significantly more expensive nuclear power plants.
If this was an option, countries with abundant hydro would be providing free power. And they aren't.
> But instead we have expensive electricity
Taking all costs into account, nuclear is currently the most expensive form of generating electricity. While building additional nuclear could get us (mostly) off fossil fuel, it is definitely not going to make your power bill any cheaper. Nuclear power is only viable with hefty subsidies - which in practice means turning off dirt-cheap solar and wind to run expensive nuclear plants.
Solar power is great but intermittence is the main issue with it. If you look at 30 year historical weather data, many highly populated regions have two week periods with almost complete cloud cover. Storage and intercontinental power transmission are usually listed as the solutions to this, but the costs of these solutions are rarely included.
Solar plus storage is included in all the major levelized cost reports, like from the NREL.
The problem is, the weather dependency makes it harder the more you add, because you will have too much when the weather is optimal and next to nothing when it isn't.
[1] https://www.reuters.com/business/energy/warming-rivers-threa...
Replacing Vogtle with renewables TWh for TWh and then building $63/kWh [3] storage with the money leftover leads to enough storage to supply the equivalent to Vogtles two new reactors for 10 days.
That is how utterly truly insanely expensive new build western nuclear power is.
[1]: https://www.ess-news.com/2025/01/23/chinas-new-energy-storag...
[2]: https://www.eia.gov/todayinenergy/detail.php?id=64586
[3]: https://www.ess-news.com/2025/01/15/chinas-cgn-new-energy-an...
We’ve just about hit peak coal.
https://www.theguardian.com/business/2024/dec/18/coal-use-to...
Not sure that counts as "just about hit peak coal".
Assuming the projected 2025-2030 installation speed is realistic and flattens out - bit "if", but not completely unrealistic - that means we'd be looking at 75% renewables in 2040 and 90% renewables in 2045.
Nuclear reactors take 15 to 20 years to build, and it'd take an additional year or 5-10 to scale up construction capacity. If we go all-out on nuclear now, that means significant nuclear power starts coming online in 20-25 years - so 2045-2050. At that point there is no more renewables gap left to bridge. There might be a small niche left for it if there is going to be essentially zero innovation in storage and short-term peaker plants, but who's going to bet billions on that?
Nuclear would've been nice if we built massive amounts of it 30 years ago, but we didn't. But starting a large-scale nuclear rollout in 2025? It just doesn't make sense.
Will it be built before we have sufficient renewable capacity?
At this point optimising their electricity cost by load balancing their compute to where electricity is cheap, free or negative on a minute by minute basis would be a sizeable cost saving. Savings that would possibly offset the hardware overprovisioning that they would need.
Like harnessing the atom for enormous amounts of 24/7 power per unit volume of fuel and not emitting CO2 while we do it? Yes! Let's do that! And work on making reprocessing more affordable, so we don't even have to mine any more fuel (at least for the next 150 years).
Is avoiding HW underutilization really worth going nuclear? The most expensive energy source of all?
We burned a few decades saying solar and wind are the solution. This set us back greatly in the struggle to reduce greenhouse emissions.
As an system-oriented person, give me a healthy combination of available, battle tested, new and promising solutions, fine-tuning weaknesses with strengths.
Go to the stadium to solve your local team/visiting team issues. You are all falling to Big Fossil antics.
An extra 50 years to solve the problem changes everything.
Yes, but that's not what's concerning the skeptics anymore, especially for solar (thankfully - the cost reductions and efficiency gains have been great). Aside from the well known geographical variance, I think the biggest legitimate concern is intermittence.
Let me try to turn that into a decent question: What variable other than energy output is most useful in order to compare energy sources? For context, all I've seen when it comes to intermittence is flame war with weak arguments thrown from both sides of the debate, i.e. "intermittence is not a problem at all, we just need batteries" to "intermittent sources are worth a fraction of an equivalent baseload source".
Honestly, I've not been convinced of either side, and (if I'm not alone in that sentiment), it may be a problem of education and communication.
We threw absolutely massive handouts at the nuclear industry 20 years ago.
Only look to Vogtle, Virgin C. Summer, Olkiluoto 3, Flamanville 3 and all other projects. Moorside, Oldbury, Wylfa and countless in the US.
Had new built nuclear power delivered on budget and on time nuclear power would definitely have been part of the solution.
Instead Vogtle provides electricity costing 19 cents/kWh. Virgil C. Summer is a $10B hole in the ground and Flamanville 3, which is not finished yet, is 7x over budget and 13 years late on a 5 year construction schedule.
The true underdog from that time, renewables (and storage) deliver energy cheaper than even fossil fuels.
What set us back was and is resistance to action on climate change, led by fossil fuel corporations and US conservatives, which has continued for decades. It's a fundamental policy of the Republican Party. Trump is already taking drastic action in that regard; it was one of his higher priorities. To try to blame someone else is absurd, and probably a talking point from their playbook.
Don't try to position yourself as the majority voice. Nuclear energy is all but clean and I don't understand the odd push in the last few years to go all in on nuclear even though there was a common understanding that it should not be the future.
Today solar & wind are the fastest, cheapest and best method for carbon free electricity, so the fossil industry pushes nuclear as a distraction.
He said no nuclear power plant was ever profitable
The histories of pretty much every green party in the western world and their anti-nuclear activism suggests otherwise.
But, in all seriousness, this could realistically be saving lives if you go with the assumption that Google was going to use this energy either way, and it otherwise would be coming from anything other than solar.
Every other source causes more deaths per energy produced. Coal is by far the worst, but natural gas, and even hydroelectric cost more lives than nuclear.
As companies like Google, Meta, and others look to nuclear power (it has the highest up time of any power source), I'm wondering how localities are going to react. Are people who are local to nuclear plants just going to be OK with these gigantic corporations consuming all this power in their backyard with no benefit to them while they take all the risk and impact of that power generation? I'm also wondering how these companies are going to deal with the excess nuclear waste. Ultimately it won't be Google or Meta dealing with the waste. How do we ensure that all the nuclear waste from AI is dealt with responsibly?
26MW is a fraction of the primary power consumed by a single passenger aircraft, by the way. It is an absolutely trivial energy input.
Figures I can find suggest that a 737 uses approximately 7MW to stay aloft.
So a couple things I learned -- I think it's still a notable amount of power, enough to power ~6,000 homes for a year just to train a single model. But also, I learned that planes use a whole lot more power than I thought!
Training a single model is essentially consuming one plane-year's worth of power, or 3-4 flights continuously while it trains. I had no idea planes used so much energy.
But also, I bet most of these companies aren't training one model and calling it done. There's probably 1s or 10s of models being trained per year per company. That's a material amount of energy use. If we could power tens or hundreds of thousands of homes, that isn't 'trivial' energy input.
I think it's useful to put it into context next to other things we take for granted, but I don't think it's fair to diminish it as nothing either.
"The Dongfang Electric Corporation's 26 MW offshore wind turbine is the largest in the world, surpassing previous models like the Mingyang 20 MW turbine. This turbine's larger size and capacity enable it to generate about 100 GWh of electricity annually, potentially powering 55,000 Chinese homes or 9,200 American homes."
Edit: more info here, https://www.bloomberg.com/graphics/2025-ai-impacts-data-cent...
The best estimate I can find is 7.2GWh.
Which would be...7 hours of output from a 1GW powerplant.
In the energy case, we will be more reliant on non nuclear power: coal, fossil fuel, etc. I’m not sure you can scale “clean energy” at the rate we are moving.
Look at what was deployed last year, in GW terms:
https://www.eia.gov/todayinenergy/detail.php?id=64586#:~:tex...
but note that gas produces at a capacity factor of ~50%, and solar at 25%, so scale solar down by half to better compare gas to solar.
Batteries are also here in great force. The average cost of battery-backed solar is cheaper is comparable to gas, and cheaper than new nuclear.
The main barrier to new solar and batteries are grid expansion to ship the electricity places. Putting a datacenter next to a proposed site for building solar + batteries that's waiting for its turn to get connected to the grid would probably be the fastest way to scale, if fiber can go there.
Energy is what matters when training a model.
Please get your units right. In the meantime, down voted.
I don’t think a reply like this is in the spirit of this site.
Actually the amount of power available matters because you are consuming energy in time. If I have a 1MW plant and a battery, I can generate 1GWh in about 3 weeks. This seems a little silly though. A Hyperscale DC campus is ~150MW to 200MW. If you plot the larger ones, they are almost all near power stations with >1GW capacity (not all).
The industry trend is towards building 1GW datacenters. Last I checked these would consume ~8.7TWh (assuming PUE of 1). However, the 8.7TWh while relevant is meaningless unless the power to the DC can be 1GW. Since the plant itself has to generate more than 1GW (the plant has a cap ratio so more than this, plus other demand, etc..) for such a site, then it follows that there are limited number of sites in the US (this is public info see EIA.gov or Wikipedia).
Grok3 is already at 140MW (100 days of training ==> 336GWh) at ~10^26 FLOP. Model FLOP is increasing at ~5x per year so by 2030, we are expecting to be ~10^28 and that would take ~10GW (24PWh). If I am optimistic and say that the efficiency can improve by 1.3x per year, then we still need a very large generating station to meet the demand or we need to distribute among many smaller sites.
You can push the numbers around however you like but the conclusion is the same, the timing may be different.
There's a reason why all the hyperscalers are investing in nuclear, large generating capacity and the highest cap factor of any form of energy.
My 2nd comment still stands, and you left unaddressed (remember the forest?)..
- You get free cooling, but if you use too much you melt the permafrost, which has huge environmental cost.
- Building in remote locations is enormously expensive, especially with the requirements of a nuclear generating station.
- Now you have to run a city for the operators to live in and ship in everything they need (not to mention hardware to the DC.
- Denmark (and so presumably Greenland) has a law against building nuclear generating stations.
Besides, building nuclear power stations with the concept that we accept an accident will happen is crazy. Better to invest in preventing them than mitigating them.