Google to back three new nuclear projects • Hacker News Board

Re: Google to back three new nuclear projects

#171

Post by PaulHoule » Thu, May 08, 2025, 4:43 PM UTC

ttfkam wrote: post #149

pfdietz wrote: post #129

Earlier quoted context omitted.

Yes, hydrogen is clearly a much easier technology to make work than fast reactors. Why is this even a question? For example, fast reactors have the issue that in an accident, if fuel melts and rearranges, one can have potentially have a configuration that is prompt supercritical on fast neutrons. This is functionally an atomic bomb. Also, even in a Fallout Future where everything is nuclear powered, hydrogen is still…

"Functionally an atomic bomb"? Why do you speak on topics you obviously know so little about? Where did you get this nonsense? Fast neutron designs aren't without their challenges, but causing an atomic explosion is not on that list. Hydrogen explosions? Possible. Steam explosions? Possible. Atomic explosions? Not even theoretically can you get enough U-235 to clump together to do that without cancelling known basic…

The more detailed simulations have gotten the less bad a meltdown looks in a fast reactors. Usually some of the molten core flows away and no more critical mass. If it goes over critical there can be some energy release but over time it looks less and less and not a problem to contain.

Sodium has its problems (burns in carbon dioxide!) but the chemistry is favorable for a meltdown because the most dangerous fission products are iodine and cesium. The former reacts with the sodium to make a salt that dissolves in the sodium, the second alloys with the sodium. Either way they stay put and don’t go into the environment.

Re: Google to back three new nuclear projects

#172

Post by ViewTrick1002 » Thu, May 08, 2025, 4:43 PM UTC

olalonde wrote: post #162

ViewTrick1002 wrote: post #74

Earlier quoted context omitted.

The average wholesale prices in California is nothing special. The costs come from the wildfires and a derelict grid requiring large infrastructure upgrades.

I can see a derelict grid and wildfires increasing power outages but how does it increase the cost of electricity itself?

Those costs include grid fees?

Grid fees pay for damage caused by wildfires.

Re: Google to back three new nuclear projects

#173

Post by bryanlarsen » Thu, May 08, 2025, 4:45 PM UTC

r0m4n0 wrote: post #118

mmmBacon wrote: post #96

It took about 1GW to train Chat-GPT4. If you look at the locations in the United States (>70% of all AI is in the US), there are only ~63 geographic regions you could put a 1GW data center. As AI models are growing at ~5x per year, it seems like the infrastructure is no in place to keep the AI models growing at that rate. As companies like Google, Meta, and others look to nuclear power (it has the highest up time of…

What is the alternative though? I think it’s fair to question a decision but if people put their foot down when they don’t see the answer as good or clear enough then you end up with the status quo. This is the same thing that happened with housing (and building projects in general) in many larger cities. If all the housing projects are squashed for some decent alternative reason, you end up with the alternate realit…

The world added 600GW of solar last year, and is adding at a 1TW annualized rate. We do not have the capacity to add any other power source at that rate.

Re: Google to back three new nuclear projects

#174

Post by cryptonector » Thu, May 08, 2025, 4:46 PM UTC

Aloisius wrote: post #104

ruined wrote: post #80

Earlier quoted context omitted.

california electric rates are so high because the state board keeps raising them 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

Spending more money on infrastructure means profits will increase. There's not really any way around that. Capital expenditures are profit.

Well, no, capital expenditures can create future profit. Emphasis on _can_ and _future_.

Re: Google to back three new nuclear projects

#175

Post by freshpots » Thu, May 08, 2025, 4:46 PM UTC

jeffbee wrote: post #111

mmmBacon wrote: post #96

It took about 1GW to train Chat-GPT4. If you look at the locations in the United States (>70% of all AI is in the US), there are only ~63 geographic regions you could put a 1GW data center. As AI models are growing at ~5x per year, it seems like the infrastructure is no in place to keep the AI models growing at that rate. As companies like Google, Meta, and others look to nuclear power (it has the highest up time of…

Training GPT-4 used (claimed) 62GW-h over 100 days, for an average of 26MW. Rest of your comment follows from this error. 26MW is a fraction of the primary power consumed by a single passenger aircraft, by the way. It is an absolutely trivial energy input.

It is not trivial at all, it's the same energy used as about 9,000 homes and roughly 50-100 L/s of water wasted to evaporative cooling of said hardware.

"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...

Re: Google to back three new nuclear projects

#176

Post by Hilift » Thu, May 08, 2025, 4:47 PM UTC

bpodgursky wrote: post #6

Every compute company knows that power shortage is a looming crisis. They don't have nuclear expertise in-house and are desperately looking for somewhere to put their money that seems to have experience and capability 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.

Ludicrous. You can't build a reactor in the US for less than $10 billion. Combine that with natural gas at prices five times less than Europe and that means that no-one will loan money for a project. If they do, it is usually subsidized by naive taxpayers. Meanwhile a windmill can transported on the Interstate in Kansas unattended and installed in two days.

Re: Google to back three new nuclear projects

#177

Post by cryptonector » Thu, May 08, 2025, 4:49 PM UTC

delusional wrote: post #94

muth02446 wrote: post #65

I view nuclear as a prudent diversification of energy sources: What happens if some supervolcano erupts, and because of the ashes significantly less sunlight reaches the surface of the earth. Presumably, there will also be less wind then.

Even on a dead earth the AI must consume and indescribable amount of power.

Shades of Isaac Asimov's The Last Question.

Plot twist: the computer's last act at the end of The Last Question was just an LLM's hallucination.

Re: Google to back three new nuclear projects

#178

Post by bryanlarsen » Thu, May 08, 2025, 4:50 PM UTC

barbazoo wrote: post #86

Related: https://news.ycombinator.com/item?id=43927371 > Ontario set to begin construction of Canada's first mini nuclear power plant

$20B for 300MW, and that's before the inevitable massive cost overruns. Continuing the Ontario provincial government's history of lighting taxpayer money on fire for electricity.

Re: Google to back three new nuclear projects

#179

Post by croes » Thu, May 08, 2025, 4:50 PM UTC

muth02446 wrote: post #65

I view nuclear as a prudent diversification of energy sources: What happens if some supervolcano erupts, and because of the ashes significantly less sunlight reaches the surface of the earth. Presumably, there will also be less wind then.

Are we playing What-if?

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?

Re: Google to back three new nuclear projects

#180

Post by ViewTrick1002 » Thu, May 08, 2025, 4:52 PM UTC

PaulHoule wrote: post #113

ViewTrick1002 wrote: post #19

Earlier quoted context omitted.

Storage is absolutely exploding. With China adding 74 GW in 2024 [1] alone and for the US it was expected to make up 30% of grid additions [2] before Trump came with his sledgehammer of insanity. 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 i…

That's a reasonable way to think about it, but is 10 days enough? It seems 12-24 hours would be needed to smooth out diurnal variations, but there is also the seasonal variation of 2x-3x in many places which either requires a large investment in overgeneration or huge amounts of storage. There is also this problem https://en.wikipedia.org/wiki/Dunkelflaute which means the storage requirements will be a bit more than…

Throw some gas turbines on it. Low CAPEX high OPEX. Just like we’ve done for the past decades with the previous ”base load and peaking” paradigm.

Those gas turbines will be a minuscule part of the total energy supply.

When it finally becomes the most pressing issue the gas turbines can trivially be fueled by green hydrogen, green hydrogen derivatives, biofuels or biogas from collecting food waste. If they are still needed.

Lets wait and see what aviation and shipping settles on before attempting to solve a future issue today.

Yes we already have a solution for all those industries which require stable power: buy an electricity future.

But somehow we need to treat the grid differently and handout untold trillions to the nuclear industry.

We have research on when we achieve learning effects.

> If you look at the data specifically you're going to find learning but for that there's a several requirements:

> - It has to be the same site

> - It has to be the same constructor

> - It has to be at least two years of of gap between one construction to the next

> - It has to be constant labor laws

> - It has to be a constant regulatory regime

> When you add these five you only get like four or five examples in the world.

From a nuclear energy professor at MIT in a nuclear power industry podcast, giving an overly positive but still sober image regarding the nuclear industry as it exists today.

https://www.youtube.com/watch?v=dDzaSucDg7k

China is not succeeding? They have been averaging 4-5 construction starts per year since 2020 which tracks to a 2-3% nuclear power in their electricity mix.

From their 2011 target of building 300 GW nuclear power in the next 10-20 years they have so far managed to complete 46 GW. But surely those final 254 GW will show up before 2031.