A power shortage could short-circuit Nvidia's rise | Hacker News Reader

A power shortage could short-circuit Nvidia's rise

economist.com

37 points1vuio0pswjnm757 comments

One really interesting strategy the US could pursue here would be to heavily tariff solar[1] and just randomly attack wind projects[2]. Just completely self-own itself on the two cheapest energy sources.

It wouldn't make any sense, but it would be provocative, really drive engagement.

[1] https://seia.org/news/solar-tariff-impacts/

[2] https://www.npr.org/2025/08/31/nx-s1-5522943/trump-offshore-...

Maybe the tariff could encourage local manufacturing of solar? I have no idea, but I suppose that our local manufacturing could be getting killed by economies of scale abroad.

Or I am overthinking it and solar is something that (D) politicians support so the (R) president tautologically must oppose it. Therefore we must not have nice things

> Maybe the tariff could encourage local manufacturing of solar?

That's what tariffs do ... if you leave them in place for extended periods of time.

The problem is that nobody is going to bet their business on what the tariffs will be tomorrow when it could be 10x or zero.

Businesses are just going to stop and hold their breath until Trump goes away.

We have tried locally, with huge government backing, but that failed. But since it was an Obama initiative, a new attempt will never be tried.

Farmers make more money from wind turbines on their land than their crop. https://ambrook.com/offrange/farm-finance/there-will-be-wind

And that is stable money, works without rain, which crops don't.

My guess is that the US is sliding into the pattern of ‘corrupt, resource cursed petrostate’ and all the dysfunction that comes with that. What will happen is whatever maximizes fossil fuel revenues, and lots of shiny distractions to prevent this from being scrutinized.

I worked for a company making GPU clouds. The biggest problem we had for deployments was not getting GPUs -- we had plenty of those sitting in warehouses. The biggest issue was finding data center space with sufficient power and cooling. There was plenty of square footage, just not enough power for it all.

They're now building gigawatt datacenters to handle all the GPUs.

The big question is were to build them. There are only a few places with cheap and plentiful power. One of those is Quebec (but it's not that big and there is a lot of regulation). Another is Texas (except their grid isn't very stable). And the last is China. And you can't build a datacenter in China unless you're Chinese.

It'll be interesting to see how this pans out. Maybe the current admin (which is big on deregulation) will make it easier to build power plants, especially nuclear ones.

Edit: I wrote this comment four days ago. I couldn't figure out why I was suddenly getting a bunch of replies to it. Apparently when HN does a second chance, they just reset the time on all the comments. Odd, but I guess it makes sense knowing what I know about how the sorting is calculated. It's probably the easiest way.

Making it easier to build nuclear power stations would be extremely useful. Let's hope nothing of value is lost in that process.

Places further north are great contenders because of the free cooling. Also, many of them have cheap electricity from hydro or even geothermal, like Iceland.

I bet on natural gas powerplants will start being built together with data centers.

Ironically xAI is already doing this.

Nuclear power plants take something like a decade to build (after permitting)

It makes more sense to go for PV plus batteries that can be installed in a matter of weeks

how much of that 10 years is the physical limit, and how much is social / cultural / organizational / political overhead?

From memory, when the brits got out of their own way they could build a nuke plant every 6-8 years.

Does that matter much when you want to have a datacenter as quickly as possible? It’s not like those things will change quickly.

yes -- knowing the physical limit can motivate moving closer to it.

> after permitting

Load-bearing parenthetical!

> Nuclear power plants take something like a decade to build

The most-recently completed fission power station on this planet needed 23 years under construction and it is still in testing. A recent American one took 15 years.

When the brits were at it (With insourced nuclear engineering and low regulatory overhead) they could crank them out every 6 - 8 years.

The brits let all that technical capability wither and could not do it again right now.

But if someone was willing its still theoretically possible. Just takes total alignment between government and private.

OK, but people who want these actually have to build them in the countries that exist. This isn't Civilization. You cannot switch to a civic that gets more hammers at the beginning of your next turn.

Sure but the argument against nuclear is always the same. Its hard right now.

The problem is that if you dont start correcting that hardness now, the next time you think you might need a nuclear industry, its still hard.

You need to train, grant experience and give money to professionals so they can exist to build stuff. It took the UK ~ 15 years to get started, then they were cranking them out like no ones business.

Think of it this way, in 15 years you can have something you might need, or you can guarantee you wont have it whether you need it or not.

Thing is its the same for all large infrastructure products. But only in nuke do we have people actively trying to prevent the industry from being created and maintained so they can use the unreadyness as an excuse.

How do you get the expertise back on a dime?

You have to tack on a few slow builds that create talent and local knowledge for that, or poaching people from China Operation Paperclip style maybe.

It's why the Vogtle plant is on track to lead to the worst possible outcome: yes it was very expensive. So the best thing to do is to start building another one right away while all those lessons and skills are still current.

Or we could just use solar and batteries that can be installed by a combination of literally any basic carpenters, electricians, and truckers.

South Korea built 13 nuclear reactors in recent decades, with an average construction period of 56 months...

Apparently Japan is the fastest builder (46 months).

the 10 year+ issue is a western problem.

Ok but telling someone they can have GPUs online in a mere 5 years if they build as fast as SK is still going to be a very painful pill. How do we get a DC in a year?

  How do we get a DC in a year?

Migrate an existing plant? https://www.datacenterdynamics.com/en/news/elon-musks-xai-bu...

> The company has apparently been thinking outside the box to meet its power needs, with Musk stating a couple of weeks ago that it intends to buy a power plant from abroad and import it into the US to provide energy for its data centers.

That did not go well for the Japanese

Skill issue!

The Texas grid is stable now that they added batteries BTW.

> One of those is Quebec (but it's not that big)

Sort of a baffling statement. Quebec is gigantic. It would be a top-20 nation, by extent, if it were a nation.

Quebec has a lot of empty land. But it does not have a lot of buildable land near plentiful power, super fast internet, and highly skilled technical workers, which are all things you need to build a datacenter.

There is tons of big power producing electric dams in the middle of nowhere (north of Québec province). It’s also cold most of the year. As far as I know, datacenter doesn’t mean tons of technical workers (except during the building phase). Bringing fiber would not be impossible.

> Maybe the current admin (which is big on deregulation) will make it easier to build power plants, especially nuclear ones.

They aren’t big on deregulation at all. They’re big on selective regulation. They’re also big on killing any power project that isn’t oil or coal.

https://apnews.com/article/trump-offshore-wind-renewable-ene...

We are in for a painful lesson on why China’s investment in renewables wasn’t just good for their ecology.

There are hydrogen pipelines in texas. There is at least one DC provider using that for power instead of the grid

In m

In my part of PA there are 3 in the process of going in nearby. I think the largest of the 3 is "only" 828 megawatts though. One of the others is supposed to be 300MW, and I'm not sure about the 3rd. There is another group talking about 3 more campuses with a combined power budget of 1.3GW about 55 miles from here. But then while we don't have cheap land, we do have nuclear and hydroelectric in the area, so I guess the makes it attractive.

Curious what you think about Oregon.

Has a lot of hydroelectric and the nights get super-cold, so you could open the roof for free ventilation

I think Texas (ERCOT) is a terrible option these days. Meta recently made a fantastic choice by picking Louisiana for their new monster. The MISO grid tends to be cheaper and less volatile than ERCOT.

I got down voted when I said China is likely to win the AI race as they are also targeting the other big cost of computing power/energy on another thread. Today solar + BESS is cheaper than coal where as costs for both keep decreasing each year.

Remove restrictions on solar import from China. 62 GW may sound a large number, but China added 277GW solar in just 2024. They have the surplus capacity and hence cheapest price.

China installed 3 gigawatts of solar power every day in May: equivalent to building one coal-fired power plant every 8 hours. They are so far ahead of the US on renewables now that even if Trump had not sold out the future of the country to the fossil fuel industry, the US would have been hard-pressed to catch up.

This is honestly one of the stupidest things about this sort of policy: solar panels last 20 years and China can't take them away once you have them.

If you're doing something much more valuable with the power, then buying a lot of PV from China makes sense. If you think the panels are being unfairly subsidized then buying a lot of PV from China is effectively having the Chinese government pay you to have cheap power.

There's an enormous difference between being dependent on short term consumable resources,.and acquiring multidecadel productive assets.

The US at all points seems to not understand it's relationship with China at all.

Most people are economically illiterate and don't understand what subsidies actually imply.

They literally only "hurt" you if you have the local industry to harm to begin with. Otherwise, if someone else is paying the subsidy, it gives you a good for cheaper than you could have had otherwise.

The current admin just turns this up to 11 with their ideologically driven nonsense.

My hunch: we’ll see three things happen in parallel

- AI backend providers vertically integrating into energy production (like xAI’s gas plants, or Meta’s local generation experiments),

- renewed interest in genuinely efficient computing paradigms (e.g. reversible/approximate computing, analog accelerators),

- a political battle over whether AI workloads deserve priority access to power vs. EVs, homes, or manufacturing, alongside an increase in energy prices.

You need cheap, reliable power + political/regulatory willingness + cooling. That’s a very short list of geographies. And even then, power buildout timelines (whether nuclear, gas, or grid-scale solar+batteries) move at "utility speed", which is decades, not quarters. That doesn’t match the cadence of GPU product launches.

Energy efficient computing is a very exciting field. I hope it will get more attention driven by these economic constraints.

As a short teaser: Landauers principle suggests that the energy required to erase one bit off information is bounded from below by k_BTln(2). This could lead us down a path towards reversible computing, to avoid energy costs for deleting information.

https://en.wikipedia.org/wiki/Landauer%27s_principle

https://en.wikipedia.org/wiki/Reversible_computing

The work started around 10-15 years ago and is now largely done. Many people confuse large absolute numbers like 1 kW with inefficiency but today's GPUs/TPUs are close to the practical efficiency limit with today's 3 nm technology.

I think it's mostly because laypeople think that all heat is wasted heat. Most people are pretty surprised to learn that there's a fundamental energy cost to flip a bit and that there even is a lower limit to the amount of heat generated by a bit flip.

I think the problem is that you can't get the chips hot enough to drive an efficient industrial process unless you are using GaN or something.

The carnot cycle on waste heat is bad

They might be close to the limit for what they do (I don’t know), but are they close to the limit of what we need them for?

Why Nvidia is desperate to get back to China: https://youtu.be/DMVAqABLkxk?si=FsyHMGmGEVPUGAot

The timing mismatch is crucial - data centers can be built in 12-18 months, but new power generation takes 5-10 years minimum. We're essentially trying to scale AI demand faster than energy infrastructure can physically respond. This creates interesting arbitrage opportunities in power-rich but compute-poor regions.

Question: Is piping data a long distance more cost effective than energy?

I would have thought so.

If so building data centres near hydro or geothermal plants (I'm from New Zealand where we have a lot of both) would make sense

Yes, but I think all of that cheap power has been used up already in the US.

Ah, finally an acknowledgement that the melting 12VHPWR connectors short-circuiting Nvidia's top-of-the-line hardware, the RTX 4090 and 5090, may finally have economic ramifications. Heh.

What is the point of all this? Why are we using so much power for hallucinated google search?

Because, for the time being, people are buying it.

Well .... investors are funding it. Take those away and nowhere near enough people want the price on the tin.

Even if that were true, that wouldn't be a useful justification. People buy drugs because they are addictive, it's not a rational decision about costs and benefits. Externalities are almost not at all considered in the contract here.

Reply on news.ycombinator.com