Three Mile Island nuclear plant restart in Microsoft AI power deal

> You can absolutely build solar+wind+storage systems that deliver 24/7/365 energy with many nines SLAs, on the real earth with real statistics on weather.

Once you start adding nines, nuclear starts to be become attractive again. Hence these deals.

The cost effectiveness plummets after one point.

> We know the probability that the sun stops shining and the wind stops blowing for N days, we can calculate it from historical data.

Exactly. And everyone is putting their heads in the sand over this calculation.

Seasonal energy storage is simply not a problem we have even begun to solve yet. We know the numbers, we're just not talking about them.

Plus if it's not sunny/windy in one part of the country, it may well be very sunny or windy in other parts of the country. We have a whole freaking energy grid!

This isn't aimed at you, but more at the people dismissing the utility of solar and wind power.

It’s literally just that having a solar+wind grid that can survive 3 days of bad weather costs 6x as much as a grid that can survive normal day to day.

> I don’t understand why wind solar is subject to absolutes devoid of probability - “what if the sun and wind stop simultaneously for 2 months?”

Something close to that happens almost every year in Europe. Last year there was nearly 3 weeks in winter where there was very little wind across the continent (<20% capacity production on all wind farms), and obviously little to no solar (high latitudes).

Nuclear isn’t a replacement for solar and wind. It’s additive.

Build all of it.

You make very legitimate point which I appreciate; I agree than in aggregate weather forecasting should be fairly reliable as regards sun and wind.

Question though... do we understand the tail risk from volcanos on solar outputs, either natural-occurring eruptions or maliciously-triggered ones?

>“what if the sun and wind stop simultaneously for 2 months?”

Anything alive after that wouldn't be human and wouldn't be worried about electricity.

The problem is that you need to plan for the absolute worst case, once in a century, scenario and size your batteries accordingly.

Same reason why almost no one goes 100% solar at home deep in the south/north because you'd need 20x your needed capacity in batteries to account for that once in a decade bleak winter week

Somehow you never hear these people say "What if the nuclear fuel supply chain becomes compromised for an arbitrarily extended period."

Nuclear plants often operate at 90%+ capacity factors, while solar can range from 10-25% depending on location. So you might need 4-9 times more solar capacity to match total energy output. So depending on the region, you might need 4-9 gigawatts of solar capacity with a higher amount of battery storage to match 1 GW nuclear output. This is because you need to account for winter vs summer, for bad weather etc. Maybe it will take a lot more space and cost more to build than restarting a nuclear plant for 1.6 billion.

> We know the probability that the sun stops shining and the wind stops blowing for N days, we can calculate it from historical data

We absolutely do not have long-term climate models with that sort of accuracy.

> what if the sun and wind stop simultaneously for 2 months

Forget about two months, the cost to battery store just for two or three days is enormous. At the same time, it does not happen often enough to justify the cost over just using natural gas as a backup, so that is how it is done.

We'd be screwed if a Mount Tambora black swan event happened again.

Many documented cases of non zero marginal cost energy systems failing. And in response we have learnt to eliminate, minimise, and manage risk better in those systems.

But apparently, in zero marginal cost energy it’s impossible. The only energy systems where you can have reliability have rent seeking minerals companies selling you fuel.

That's not how this works though, energy abundance means industrial abundance, more energy availability (for cheap) would simply mean that demand will scale to meet the output.

The only time it won't scale to meet output is if the price stays high, but if the nuclear energy is already on the grid and has elastic pricing, the rest will take care of itself.

Easily solved via sane energy policy. Of course politics will likely make this impossible in the US.

In a rational environment you run your nuclear at 100% 24x7. The cost of the fuel is not material running at 10% vs. 100%. Letting parasitic intermittent power sources screw this up is simply financial engineering by largely bad actors. At least currently.

Then you use intermittent power sources to provide your peak loads during the day, and any excess you ideally start putting into storage - whatever that may be. If you have effectively free marginal power during certain peak times, I'm positive industry will find a way to turn that into money.

I still have hope sanity returns to the energy discussion, but it likely won't happen in the US during my lifetime. The cost of solar and wind is entirely politics - the storage cost is literally never considered when reading articles on the subject. The hidden costs are likely 10x or so what the marginal cost per kwh everyone loves to spew. Lots of folks getting massively rich off this disinformation so there is huge inertia behind ignoring it - even from very smart people that should absolutely know better after a few hours of reading on the subject. Just look at many of the posts here at HN.

The environmental costs of methane (natural gas) are simply not being considered. The methane leaks into the environment are underestimated by at least 10x if not much, much, more. Pivoting from nuclear and to natural gas has been an utter environmental disaster.

And thank-you for making the point that utilities should be socialized and not a profit center.

EVERYBODY deserves clean water, electricity, housing and food.

There is nothing inherent or natural about that scenario, it is all due to the failure to properly design electricity markets.

See also: the 2001 California electricity crisis.

https://en.wikipedia.org/wiki/2000–2001_California_electrici...

See also: Enron

https://en.wikipedia.org/wiki/2000–2001_California_electrici...

If we have designed our markets to price cheap, reliable electricity out of the market and instead prefer expensive unreliable electricity, we've really f*ed up.

The problem you describe seems to be one targeted by the TerraPower project in Wyoming. It plans to operate at 100% capacity at all times (345 MWe), but makes itself more akin to other renewables by incorporating energy storage into its design. It is supposed to be able to increase capacity to 150% (500 MWe), allowing it to respond to energy scarcity. But it can also respond to energy abundance by storing excess production in molten salt storage tanks.

> When you get plentiful and nearly free solar energy at certain times of the year, the utilization of your nuclear power plant will drop

In the short run, sure. In the long run, cheap power builds its own demand. A country that commits to a certain amount of nuclear baseload, even if run at a loss in the short term, is injecting a very real industrial subsidy into its system. (The way to ensure you don’t get a dog is to subsidise long-term loans for private borrowers. They still need to make a profit someday. But you reduce the time value of money for them.)

The investment doesn’t make sense for a non-nuclear power. But if you’re already producing nuclear waste at scale for your military, it’s a little silly to pretend you’re safer without a civilian reactor in the middle of a desert while all manner of subs and ships patrol your coast.

It's worth remembering that the anti-nuclear crowd was always backed by big oil, and even in the age where solar and wind are being pushed, big oil benefits as the unmet demand for base load power, in the absence of nuclear, means that they get the business as there is no alternative.

They pulled the green crowd along for the ride hook, line, and sinker.

To be fair, the unsafe designs are all 50 years old because of FUD from incidents that are also generally pretty old. I’m not sure it’s comparable to modern designs - this isn’t a “new iPhone every year” deal.

> Nuclear is not safe

Nothing is 100.0% safe - but as far as I can find nuclear comes incredibly close[0] even when including Chernobyl and Fukushima.

[0]: https://ourworldindata.org/safest-sources-of-energy

Spain also has a hugely different energy market than the United States. Spain is around 5% of the size of the US, being closer to a state in our context. Different states have different climates, making them more aligned with different energy sources. It doesn't make sense to compare the US and Spain in this context.

> We are at 20% nuclear,

Doesn't this prove the parent comment's point, that the best mix of energy is renewables supplemented by a base of nuclear power?

> Get down your "education" horse and solve your own problems first, asuming you are from the USA

This feels unnecessary. Spain has done an amazing job building an energy base, let's talk about how we can export to different countries and climates instead of putting people down.

> However, nuclear power plants cost about the same amount of money per hour whether they're running or not, so are poorly suited as a solar complement.

I think this is why they call it baseload. Yes it costs the same, used or unused... but it's always there, ready to be used. This is its strength, not a defect.

> If you really want 100% reliability out of a primarily-solar grid

But no reasonable person wants that. They want "100% reliability, and who cares where it comes from". If you're environmentally minded, you can tack on a "with no carbon"... but even that isn't the same thing as a "primarily solar grid".

What you say makes sense. But what if we need to generate 50 times more electricity than we do now? At some point solar and wind energy will start being very expensive because we'll run out of cheap land. You might think 50 is absurd, but consider this: right now only about 30% of our energy consumption (and therefore emissions) is electricity. If we want to cut emissions to zero, we need to transition most of our energy use to electricity. The big culprits are transportation, steel, concrete and fertilizers. That's a factor of 3. Then we want to increase everyone's standards of living to match the current standard of living of the average American or European person. That's another factor of 3. Then we want to increase our (clean) energy use. Each SpaceX Starship launch puts about 5000 tons of CO2-equivalent emissions in the atmosphere. If, as Elon wants, we end up having hundreds or thousands of launches per day, and we want these launches to be net zero, then we need to find a way to manufacture hundreds of millions of tons of synthetic fuel each year. AI will also consume a lot of electricity. We will also need to build cities, if we are to lift billions of people out of their current rural life (in Africa, South America, India, etc). Everybody talks about solar as if the goal is to replace the current generation capacity. But the goal is to do that, and then more, much, much more. I think you need nuclear for that.

There are huge downsides though because moving power generation further and further from where it’s used opens you up to more disruption from terrorist attacks on the grid (and cyber attacks) unless you also do less centralized production which is going to be more expensive.

That doesn't help after sunset on the west coast when residential demand peaks and there is little renewable power being generated anywhere in North America. In theory batteries or other storage could close that gap but realistically it's going to take decades to scale that up.

> > Nuclear fission will play an important role in the future of grid energy.

> No.

> An energy mix is the answer.

Maybe that “No.” was unnecessary.

> One of the parts that is mixed in, can be nuclear. Probably should be. But always a small part of the mix. Nuclear is slow: slow to power up, slow to adjust to demand or supply from others in the mix. And extremely slow to build.

All of those things, to me, mean that it should be a large part of the mix. Nuclear should be the backbone of our energy production, and should be sufficient to supply our baseline needs.

That said, in places where hydro or geothermal are practical, those should be used in preference to nuclear. They're cheaper, more reliable, and just better in almost every way.

Solar, wind, tidal, wave action, and so on should be what we build on top of that baseline. They should be cheaper to build and operate, but far less consistent.

If you want to train a huge LLM, or smelt metals, or do anything that's energy-intensive but not very time-sensitive, you schedule those loads during times where energy production exceeds existing demand.

... and you know when that happens by pricing energy based on availability. Electricity should be cheapest when we have more than we need, and more expensive when the inconsistent sources listed above aren't producing. In other words, market forces are sufficient to make this happen.

In fact, one of the cool things about solar and wind in particular is that they are so aggressively cyclical that it's possible that energy prices could actually go negative - not often, or regularly, but possible. That opens the door to all kinds of use cases that would otherwise never be profitable, and using those types of technologies often leads to efficiency gains that can eventually make them more efficient than the current alternatives.

The discussion of incorporating variability into the power grid is incredibly unsophisticated on HN.

The US (and Europe) are big places with diverse energy demand and weather patterns, and smart renewable energy developers will seek out resources that are decorrelated with existing supplies because the power can be sold at higher prices.

Batteries are also getting cheap enough to profitably arbitrage intraday price variability.

Another thing that gets ignored in these discussions is that in any net zero scenario there is going to be a fair bit of green hydrogen/biofuels/synfuel produced for transport and industrial use. A fraction of that could be diverted to gas peaker plants when monster dunkelflautes manifest themselves.

For the explanation of hysteria: 1) Nuclear disasters are acute, traumatic events that can have drastic effects on health of people and ecosystems for generations 2) Nuclear power sounds like nuclear bombs and nuclear radiation, both of which are scary and resonate with long-held cultural fears of nuclear technology that have existed since the Cold War.

Basically, yes most fears of nuclear power are irrational. But that's part of the problem—even if nuclear power polluted in a manner similar to coal power and there wasn't the risk of acute disaster, I still think Americans would be quite wary of the technology purely due to cultural priors.

I doubt that anything that connects to the name "Three Mile Island" is going to be good for PR.

Battery arrays aren't primarily rated for their storage capacity but for their ability to provide/consume power.

At the moment the goal of battery storage additions to solar plants is to soften their peak power production and broaden the time they provide power to extend to the morning and dinner time.

690 MW.

Nuclear plants trip off all the time. It was a big problem in one of Texas' recent cold snaps where people died.

And many of those nuclear down times are for month+ times. Good luck getting enough batteries to get through that!

Offshore wind typically has a capacity factor of ~50%. Add a smaller number of batteries than you would need for nuclear, and it's going to be cost competitive against any new build.

As far as if it's cost competitive with TMI, I'm not so sure. New build nuclear is likely going to be >$190/MWh. Running the numbers being bandied about for the "Pennsylvania GDP increase", which is a really funny way to say "cost of the electricity for 20 years," and you get $115/MWh. Which is really pricy still. I'm not saying that it's a bad idea to get TMI going again, just that it's pricy, and there's likely lots of other constraints that went into deciding to do this.

But the actual investor is Constellation, which has operated the plant for decades. Microsoft is just committing to buy power.

Are you suggesting Constellation will be surprised at the hidden costs?

If only safety was a matter of system design and operational practices, not geographic location.

Because the alternative, fossil fuel generation, is much, much more polluting and deadly.

Coal plants have killed a minimum of 500k people over the past 20 years[1]. It's not an accident in that case, it's known and planned for (or at least easily predicted enough that it should have been).

But when a few hundred people, or really just 0 people[2] die in one place at one time, people lose their minds.

[1]: https://www.theguardian.com/environment/2023/nov/23/coal-pow... [2]: An inter-agency analysis concluded that the accident did not raise radioactivity far enough above background levels to cause even one additional cancer death among the people in the area https://en.wikipedia.org/wiki/Three_Mile_Island_accident

Its crazy how this resulted in 0 deaths, yet it pretty much halted nuclear power production for good in the US.

It's insane because killing people is the modus operandi of fossil fuel power - and I'm not even talking about climate change. People often die on oil fields and in mines, toxic waste from coal plants leaches into our water supply, and ash enters our air and causes asthma and heart problems. Coal alone has killed about 460,000 Americans in the 21st century: https://www.theguardian.com/environment/2023/nov/23/coal-pow...

Nuclear power plants, in contrast, very rarely have issues: https://ourworldindata.org/grapher/death-rates-from-energy-p...

Accidents will happen daily. Those accidents are extremely well controlled. You are providing an excellent example by jumping to that insane NIMBY conclusion. I think MS can successfully bulldozer these extremely weak arguments about catastrophic accidents better than I could. Stay tuned.

Nobody died at TMI.

Nobody died at Fukushima (from the nuclear incident that is, 10000 died because of the Tsunami)

Hundreds of people died at Chernobyl.

Those are all the major accidents at production nuclear power plant that have ever occured. There are no others.

There is just one that was deadly, and it is about as representative of the safety of nuclear power as flying in a 1920s' plane compared to a state of the art Airbus.

What were the effects of the TMI accident, in your opinion?

> How are nimby crowds insane? Accidents may happen. And did happen in the past in that same place: https://en.wikipedia.org/wiki/Three_Mile_Island_accident

That would not be a problem. FAANG controls the OSs and the media. /s

Microsoft is not operating the nuclear plant, so that ideology isn't being applied to the operation of the plant.

Microsoft engineers aren't running the power plant.

> A corporation getting its way, irrespective of the democratic will of the people, or rather, those "insane NIMBYs". Corporatocracy truly is a beautiful thing.

Generally a NIMBY is anti-progress and very selfish, putting themselves ahead of everyone. My dad was called to help resolve TMI during the meltdown and was part of the engineering design group the the Beaver Valley nuclear power station. People were complaining they had cows dying of radiation years before the plant was completed. NIMBYs suck.

the democratic will should be only on state or common goods. On their land persons should be able to do whatever whitlist following safety guidelines. including operating a nuclear facility

But were said people well-informed to begin with? Voting can only work if everyone's on the same page regarding both the policy being implemented and its consequences, thus being able to defend their best interests. If the people have been consuming provably false messaging pushed by a fossil fuel industry lobby and sensationalist media, can they be trusted to make a choice that affects not just themselves their entire city, nation, and planet?

The representative democracy does attempt to solve this problem; you pick a representative you understand is aligned with your interests on the issues you do understand and trust they and their experts will do well on the rest.

It's not a perfect system, but unless the populace can be given free time and education to consider individual policies based on reliable evidence and not hearsay from biased sources, it's the best we can do.

Often mainly due to political reasons though.

Also you can't really make meaningful conclusions based on that plot due to obvious reasons. It only includes "Number of nuclear reactors shutdown worldwide". The construction peak was ~1980. The oldest commercial plant was opened only 68 years ago and obviously the early plants much less safe and had a shorter lifetime which introduces another bias...

There has been a generation of activists and NIMBYs campaigning and protesting to artificially inflate the cost of nuclear, while having this delicious benefit of now being also able to claim about the cost of nuclear.

The ballooning of costs is not even significantly due to changing safety requirements, but often due to compliance and environmental requirements. Those are political requirements that could be removed in one fell swoop.

Yeah. The only way the sentiment "I don't like that ___ is using energy, even though the energy is completely clean" makes sense is if they've already made up their minds that ___ is bad, and thus any energy usage is bad. The energy usage argument is a red herring; it's actually just a value judgment on AI itself.

> If you're an AI hater, it's frustrating to see what you consider a useless technology growing to take up more and more of our energy mix, eating into climate gains being made from the immense growth of renewable energy. If you're an AI maximalist, on the other hand, the significant energy use projected for AI is a small price to pay for a technology that you think will revolutionize our lives much more than technologies like air conditioning, refrigeration, or the automobile ever did.

> The answer probably lies somewhere in the middle. In the long run, AI's energy use will likely level off at a significant but not grid-melting level that's roughly commensurate with the collective economic value we as a society get from it. Whether the trade-offs inherent in that shift are "worth it" involves a lot of value judgements that go well beyond how much electricity a bunch of servers are using.

Source: https://arstechnica.com/ai/2024/06/is-generative-ai-really-g...

Worse: Clippy running a nuclear plant.

> market-based carbon accounting […] not necessarily worse

They are better for the corps communication and worse for anyone else.

Transporting electricity over distance has a non trivial cost in $ but also ressources and energy, as well as relocating an industry near a clean energy source or optimizing the production units.

Carbon-matching systems are great for entreprise wanting to claim they don’t produce carbon (24/7 carbon neutral) while they do. It does not depict the CO2E one (entreprise) did product de facto.

I think the GP knows that. That's why they said "out of action for half a decade".

2019 was half a decade ago.

And this particular refurbishment is in one of the countries with serious cost and schedule overruns for its nuclear build, and a country and specifically a region that is bad at physical infrastructure.

If using a ton of energy has to happen then I'd prefer it be generated from nuclear, but I don't see the need to supply LLM data centers with a ton of energy as a positive.

Solving the political problems behind nuclear power (getting TMI restarted being one of the more difficult politically but easier technically), and generally bringing the "build more nuclear" to forefront.

(Although to be fair there's been a fair bit of wind and solar subsidized heavily by crypto, as well as flare gas power generation, and a bit of PV solar)

Hah, good one.

Getting Americans to accept the metric system is probably a taller order than getting them to accept nuclear power.

Saying ESG is entirely a scam because some ESG backers travel in airplanes is like saying climate change is a hoax because it was colder than usual in Peoria today.

ESG is a huge thing. Some of it is greenwashing, some is outright scam, some is legitimate pressure to make better environmental decisions.

But insisting that ESG proponents can’t make suboptimal individual choices is silly.

Ah yes. Some people use airplanes so we need to keep burning coal, who cares when the sea levels rise and the water wars begin! The important thing is that some people acted silly by gluing themselves to the road! Can't give in to them, even if it would save the world for our children.

It’s even worse. If you have money and are against deploying AGI once we get there you still want to be the first to achieve it, because it’s the only reasonable chance of stopping others… assuming you can even align it.

You're claiming it would be terrible for the rest of us, without supporting that assumption in any way. It's not a fact, pseudo scifi action movies don't count as facts.