Hacker News Reader

Comment by kulahan | Hacker News Reader

With AI on the horizon and each server farm using as much energy as a medium-sized city, I have no idea how they hope to meet demand otherwise, unless the plan is just some equivalent to "drill baby drill".

View on HN

It’s simple, Germany isn’t going to be participating in the next industrial revolution. It will be the US vs. China. You can already see it happening with their car industry as they struggle to keep up with new technology.

Germany doesn't need to participate in the next. They need to participate in something though. They are too small to do everything alone. Even the US depends on a lot of other countries to make things work.

If we’re looking at the car and energy industries, I think China has already won.

Could you expand more on your car point? I thought BMW and Benz were doing great at the moment. I dunno much about Audi or VW, but Mini also seems to be doing well (which I thought was British, but one of their models has literally the same engine as my last bimmer, so I guess they were sold at some point?).

Heh, it was bought by BMW in ‘94.

The German car companies are struggling intensely against Chinese competition, everywhere outside of the US, and especially in China. The Chinese electric cars sell for 3 times less than the German ones in EU. The Chinese also invested heavily in e tech. The Germans? Not so much.

Ah yeah, now that you mention it I’ve heard BYD is doing well in most markets (plus others, I assume). Even ignoring current tariffs, I’m not sure an overtly Chinese car would catch on in the US, but I’m pretty sure I’ve been wrong before.

I believe that if allowed in, they would be competitive and that is why they are not allowed. And because of security claims.

Sure, talk to your grid operators about that! :)

It would take a long time to build new reactors, so not sure that would help.

Germany could also do more wind, solar, tidal, geothermal (fossil fuels aside).

It is going to take a long time and a lot of resources no matter what so maybe we should be building effective longterm solutions like nuclear instead of stopgap solar and batteries

Why would, e.g., solar and chemical or physical storage be a stopgap? Why spend 20 years of building a fission reactor these days (other than for research, medical, or defense purposes) which also make awful targets in a conflict? Maybe just wait till fusion reactors are there.

Because the reactor will still run 20 years after that while the solar and storage will need to be replaced by then

Reactors need ongoing maintenance, repairs, replacement.

Why would fusion reactors magically appear when the entire field of nuclear energy production is, in this scenario, essentially dead??

Not sure why pursuing fusion needs building fission reactors for energy production.

Because nuclear engineers, plant operators, radioactive mining facilities, and other types of workers that will be needed across both, need to be employed from today until fusion reactors are made.

Not even “instead”. We need all of the above: nuclear for base loads, solar for peak loads, batteries for surplus capture.

This right here. It's not one or the other, its a diverse combination of all of them that makes for the best results.

Base load is a concept of the past, grids around the world are being redesigned to be flexible to reap zero-production-costs renewable energy. Nuclear (which is impossible to run economically as a flexible asset) simply does not fit into that new world anymore.

Damn, so we’re left with nothing, because nuclear is by far the most viable moving forward.

It'd be way easier to build a few nuclear plants than it would be to build an equivalent constant energy source from solar+wind and batteries. The nuclear plants would also consume far less land area.

You need solar and batteries for peak loads, not just solar

In many places in the world, peak load does not occur during daylight hours, especially during winter

And yes, further north the days are longer but the solar capture efficiency is also much lower

True. I'm biased by living in a place where the peak load does happen during daylight hours (because that's when you need to run the A/C) and where heating usually happens via gas. Electric heating would indeed shift that dynamic (though municipal water heating would shift it the other way).

I'm not sure how tidal and geothermal fare in Germany

It seems that some geothermal works have caused mini-earthquakes and soil shifts in Germany and the Netherlands

My baseline expectation is some opposition to any new energy infrastructure.

I was under the impression tidal was mostly tapped out because any half-decent location has already been turned into a power plant.

Not with a tech that needs 15 years to be build

If AI server farm operators conclude that nuclear is the way to go, they should be free to do so, yes. If they manage to fulfill all regulatory requirements. (Which means it'll be at least $2 per kWh, yay.)

There's a new kind of "drill baby drill" which we should be embracing: geothermal energy. There's a lot of advancements in that space and it is a perfect base load generation source.

Geothermal is, imo, the only true competitor to nuclear. It's great at providing cheap, consistent, clean energy. Nuclear is really only needed for baseload generation, like when demand massively spikes.

If demand spikes nuclear power plants aren’t fast enough

They are when the power spikes for the day, in a typically predictable fashion. I’m not sure of anything more available that isn’t really dirty.

> baseload generation, like when demand massively spikes

That is unlike any definition of baseload generation I have ever heard.

Yeah, advanced geothermal is very interesting. They're taking fracking techniques and using them to get to hot rocks, which opens up geothermal to a much, much wider set of locations. Interested parties say it could provide everything we need beyond wind/solar, and seems much simpler than building out nuclear plants.

Check out:

https://www.volts.wtf/p/catching-up-with-enhanced-geothermal

And leveraging fusion research to rethink how drilling is done: https://www.youtube.com/watch?v=g8sjdOjNxIE

You limit data center power demand until the AI bubble pops.

Peak Bubble - https://news.ycombinator.com/item?id=45218790 - September 2025

US Data center projects blocked or delayed amid local opposition - https://news.ycombinator.com/item?id=44097350 - May 2025

Cool, your country fell way behind every other developed nation in this and you've missed out on a huge industry. In the end, your citizens will still use the products, they'll just probably end up having to pay more for the same functionality.

Other countries can shoulder the cost of the hand waving grift. If it turns out they succeed, lift their models and weights. Eat some potential IP liability for not incurring economic damage ("inefficient capital allocation") chasing magic. Be first, be smarter, or cheat ("you can just do things"). DeepSeek showed a bit of this (model training efficiency), as Apple does slow walking their gen AI. Why incur material economic risk to be first? There will be no moat.

https://hbr.org/2001/10/first-mover-disadvantage

> Wall Street Journal columnist Christopher Mims shared another chart, saying: “The 'magnificent 7' spent more than $100 billion on data centers and the like in the past three months alone.” Man, are they optimistic. Mims linked to an article by Paul Kedrosky, who offers another perspective [1] on the AI bubble, as a percentage of GDP. Kedrosky, in turn, quoted Chinese President Xi Jinping, who warned of overinvesting in AI-focused datacenters. When Xi Jinping and Wall Street traders are on the same page, you know it’s bad. [2]

[1] https://paulkedrosky.com/honey-ai-capex-ate-the-economy/

[2] https://open.substack.com/pub/thealgorithmicbridge/p/im-an-a...

They can’t even use the products as a result of their obsession with government regulation. For example, Apple released a universal translator, literally right out of Star Trek, but the EU won’t be getting it either.

https://news.ycombinator.com/item?id=45215548

https://news.ycombinator.com/item?id=45217477

Given how fast compute needs replacing, it's not much of a fall behind.

Citizens will indeed use them anyway, but there's already free models that are OK and which only need 8x current normal device RAM. Bubble bursts tomorrow? Currently-SOTA models on budget phones by the end of the decade.

The already use it and are not impressed.

AI wears out quickly if you have special demands.

A country is not forced to have AI farms running in it. Building giant powerplant for the AI tech (possible) bubble not seems wise.

The plant will take 5 - 10 years to build, who knows what demands AI will have at that point.

SO let some countries that want to spent enormous amounts of their energy on AI do so, adn the rest can connect to those.

> who knows what demands AI will have at that point

This is true for any investment pretty much.

Well, not really. Investing in heating homes or powering light bulbs is, outside of extremely extreme situations, always a good investment, because people will always want to do that.

AI is also just super young, has apparently zero mote, requires insane amounts of hardware that basically becomes useless after a couple of years, and has promised, over and over, the AI revolution is just around the corner multiple times without ever delivering.

This is shortsighted. China routinely experiences large overcapacity in their electricity grid just to deal with the unknown unknowns of outages and other new demands. Suppose that the AI bubble burst and AI energy use is negligible, the extra capacity could be used for something else: retire your traditional coal fired furnaces for steel making and replacing them with electric arc furnaces; produce more aluminum; build more EV chargers.

The wait until after the AI bubble and buy the cheap surplus of energy.

AI is useful but nit as useful as the AI companied claim it to be and the ROI isn’t as great neither.

I willing to wager that the AI bubble will burst before you could even begin to build power plants for them.

I'm sure the bubble will burst. However we have already found a few uses for AI and those uses will continue after the burst (if they are economical)

Reply on news.ycombinator.com