The reason is because permitting and building a natgas generator is the easiest among the energy production methods in the US. Datacenters need to be close-ish to Internet Exchanges to be cost competitive when lighting up network capacity. Solar cells are expensive (Chinese tariffs or domestic production) and permitting is tough. Nuclear is still a permitting and cost nightmare. Wind requires a lot of land. Hydroelectric is considered an environmental dead end after the ecological effects of the Hoover Dam. Geothermal is still unproven. Transmission lines moving power between generation and consumption is a permitting nightmare.
In that world, natural gas just makes the most sense. The US hasn't build generation capacity in any meaningful way in decades. We've deindustrialized over time so it's been relatively okay, until a new form of industry (datacenters) starts putting pressure on the whole thing.
> Solar cells are expensive (Chinese tariffs or domestic production) and permitting is tough.
The tariffs I understand (even if they really don't make sense in this particular case) but the permitting I do not. Do you have more information or links?
> Wind requires a lot of land.
In rural but populated areas wind is generally installed on someone else's grazing area for a small fee. In truly unpopulated areas (ie desert) access to land isn't usually an issue since there's approximately zero demand for it.
That said I do agree with your general theme that our grid is underinvested and the management and policy surrounding it are a mess.
> The tariffs I understand (even if they really don't make sense in this particular case) but the permitting I do not. Do you have more information or links?
Take a look at [1]. The current admin felt that NEPA reviews were taking too long for utility grade energy projects and put in a cap for NEPA review length that does not apply to wind and solar. The article goes into how long utility scale solar projects can take to go through NEPA.
> In rural but populated areas wind is generally installed on someone else's grazing area for a small fee. In truly unpopulated areas (ie desert) access to land isn't usually an issue since there's approximately zero demand for it.
The challenge then is bringing the power to the datacenter, which often involves transmission lines, which goes back to permitting.
Building a facility that uses megawatts of energy in an old farm field in the country, and having long lead times to get it installed, isn't really indicative of "deindustrialization" is it? Also, I don't think building datacenters outside of Columbus are being driven by closeness to an IX. I don't recall seeing Columbus as being significant on any US backbone map. More likely they just want to be close to each other. Someone must have started that ball rolling.
> I don't recall seeing Columbus as being significant on any US backbone map. More likely they just want to be close to each other. Someone must have started that ball rolling.
AWS us-east-2 (2016) and GCP us-east5 (2022) are both in Ohio. Not 100% sure they're close to an IX but my guess is there's existing infra to route onto.
> Building a facility that uses megawatts of energy in an old farm field in the country, and having long lead times to get it installed, isn't really indicative of "deindustrialization" is it?
Sorry I think my message might have gotten a bit conflated. I meant, in the offshoring that happened in the US in the late '90s-early '00s, the US ended up losing industrial demand. Obviously consumer demand increased in the meantime but we've been living on a mostly stagnant energy supply for a long time.
Your language is ambiguous — your horror is in reference to natural gas turbine generators (used at these installations) and not gasoline generators (like in a home context)?
Why the horror? I'd prefer the gas remain in the ground, but given the gassy production of US shale oil, I guess I'd rather it be used for this than just flared. I am frustrated that pollutant emissions aren't being policed, and also that the sudden turbine demand plus supply chain issues mean using aeroderivative turbines that are quite a bit less efficient than more complex combined cycle turbines.
There is currently 2x us electricity production in solar and batteries stuck in permit hell due to the US requiring they pay for grid upgrades before connection in a first in first out line that has grown in length and costs.
We could have cheap and available renewables, but we instead destroy them in bureaucratic hell that nobody cares about.
> due to the US requiring they pay for grid upgrades before connection
Is that not perfectly reasonable? Someone doing half the job and dumping the rest on everyone else seems like exactly the sort of thing a regulator exists to prevent.
Reading between the lines, it sounds like the issue is that solar would be located somewhere remote, the backhaul to get that electricity where it needs to be requires significant upgrades, and that takes time. Which is unfortunate and indicates historic mismanagement of said infrastructure but nonetheless the present day policy of "fix the problem first" seems perfectly reasonable.
> I guess I'd rather it be used for this than just flared
I doubt this is really reducing the rates of flaring and leaky wells. Its just additional demand.
The biggest problem I've seen is they tend to build these somewhat close to residential areas with generation on-site. Often these power generation centers aren't right next to residential areas due to both air and noise pollution. But governments are often seeming to turn a blind eye.
Yes, the noise pollution is insane. Benn Jordan's YT video "Datacenters Behaving Like Acoustic Weapons" is an insightful, scary 30 min video covering the datacenter infrasound noise, and the nasty things infrasound does to people: https://www.youtube.com/watch?v=_bP80DEAbuo
It's not supposed to be permanent, but it also allows them to not waste time waiting on physical locations to be built. Given how highly competitive this all is, I'm not surprised at all.
Elon's was definitely illegal/unpermitted gas turbines as countless news stories that came out can attest but do you have any support to claim Meta's turbines are?
It might be a little more expensive for them, but it's cheaper when you factor in the costs of pollution which they aren't paying for and which they're forcing us to pay for through increased diseases and global warming.
I can't run a frozen lasagna factory from my house. It's illegal. I don't have political approval.
If the people do not approve of data centers, they don't get built. Simple as that. Businesses do not have an inherent right to exist. Businesses are granted their existence and places of operation by the state and local municipalities that license them.
You're right. Of course political approval tends to involve a few more people than a couple dissenting commenters on a ratioed HN comment thread, depending on the regime in question. Pretty sure the entities in question had business licenses as well.
How about being transparent when we ask the residents what they want. If this hypothetical scenario you fear comes to fruition, maybe instead of back door deals/misinformation/straight up lying for hype, we publicly ask the people who have to suffer the consequences of our political desires.
Meta's first five buildings took between two and three years to build, but Williams is almost done building out 200 MW (additional) off-grid power plants in a year, and to match that they're putting their equipment in tents. That raises questions for me:
* Did they expect the next five buildings to also take between two and three years to build if done in the same manner? I'd hope it'd be significantly faster the second time because they've perfected the design, found good local contractors and suppliers, etc.
* How much of the time was the actual structure vs. all the stuff inside they still have to do with the tents?
* How long are they expecting to keep this? Are they anticipating extra problems like leaking roofs?
* What are the "off-grid power plants"? Is this basically a whole bunch of diesel or natural gas generators? [edit: oh, yes, "The site is also powered by 200 megawatts of modular gas turbines". I wonder if they're trucking in the fuel too.] If so, yuck.
I would guess the real problem is contractors are bottle-necked in good times, but not stupid enough to expand - knowing that bad times will come and they have to pay for all the expansion. (humans can be laid off, but you still need to make the payment on the bulldozer)
That makes sense. Although I have to pick on your example a bit: wouldn't they still need a concrete foundation for all that weight and thus still need bulldozers? Still unsure how much of the work they're actually avoiding.
It's funny because at first I thought it made sense but the more I think about your question the more I'm skeptical. You still need the rack. Power distribution. Data lines. Cooling lines and whatever those connect back to. All the stuff that would have been embedded under the floor has to go somewhere (labor) and still has to be assembled (more labor). So they're saving time on what, the cement and steel shell?
Or is this a building permitting issue where for some reason the bureaucracy surrounding a permanent structure is expected to drag on for years but somehow they got the tents permitted rapidly?
> Or is this a building permitting issue where for some reason the bureaucracy surrounding a permanent structure is expected to drag on for years but somehow they got the tents permitted rapidly?
Good point; some permit loophole might make sense.
It occurs to me this also could still turn out to be a giant failure: these may all still be unpowered, empty tents. They might end up taking two to three years to turn on, might never get a critical permit at all, etc. I'm vaguely recalling some story from Google's past. They had an experimental datacenter (`pq` maybe?) built out of shipping containers. There was some way they had hoped this would be cheaper that (iirc) didn't work out at all because the local fire marshal declared each shipping container to be a full structure and thus an unexpected set of regulations applied. and/or each may also have been required to have an emergency power-off button for the entire facility, which were hit by accident more than one might hope. They never built a second datacenter with that design.
Also remembering that for a long while Google's Dalles, Oregon site had building 1, building 3, and an empty concrete slab between them called building 2. I suppose Meta could have done something similar and had the slabs ready to go long ago.
WoW. I was thinking about building a new home recently, but now I'm thinking about pitching a membrane-tensioned tax haven, cuz last time I checked, temporary structures are exempt from property taxes. This is like the ultimate real estate hack.
I'm not sure why they even call them tents. Those are called fabric buildings. Fabric buildings have a lifespan of over 20 years[1] and this is limited by the PVC membranes, the actual steel structure can last a lot longer.
The whole thing is like a video game: your construction and power are your limiting factors. We need to CONSTRUCT ADDITIONAL PYLONS and so on. It's interesting that this is such a limiter on the ground that SpaceX is pursuing AI satellites in space. Truly an incredible time to be alive.
> this is such a limiter on the ground that SpaceX is pursuing AI satellites in space
What is your source for this claim? It sounds like conjecture to me.
Neither power nor construction is easier in space than it is on Earth. Higher power means higher heat dissipation, which means higher-cost satellites or downsized AI hardware. Construction is monumentally more expensive when you have to ship GPUs and their associated infrastructure away from the planet where they're manufactured.
SpaceX's orbital compute will not compete against any ground-based AI capacity. It will most likely be used for edge processing of Starlink sensor data, either as an SAR solution, multiband jamming apparatus, or any other SDR applications used at the orbital scale. There is no other justification that I am aware of that necessitates space-based AI inference. The commercial space-based AI line is a glaringly obvious coverup.
Elon has mentioned this in interviews and the S-2 for SpaceX calls out permitting challenges as a reason for moving to space. Elon being Elon, I don't really believe him but this is what he claims at least.
I hear they have real loose permitting regulations on Mars, and I think Elon was supposed to have a colony up and running about a decade ago, so that should all work out for him.
Microsoft had a neat trick here in the Netherlands: instead of opening a new site they decided to make a their existing site higher by adding a few floors.
Ofcourse that only works once the Dutch borg adapts!
So AI infrastructure buildout is starting to feel a lot like emergency industrial mobilization...
Also, building rapid temp shells plus nearby gas turbines paints a very different picture than the one conveyed by the "clean-energy" PR around hyperscale data centers.
The real sickness is the years spent chasing stock buybacks and low taxes while infrastructure and industrial capacity languished. What an absurd situation where arrays of "temporary" on-site gas turbines are the only viable way to build new infrastructure like this.
Better than most businesses. Most data center campuses I've seen, regardless of construction, were behind a substantial permitter fence and had some on-premise guard force. Like razor blades, they would be better to intercept GPUs or other equipment when in transit though there have been decades of rumors that such goods are usually done by trucking companies with ties to organized crime.
As always, there are some very erudite comments here on HN, which is why I like the site so much. My erudite comment, or rather question, is this: why aren't these people using AI to solve all these problems? Surely it would be a good test of The Product and maybe it would give s[ck]eptics some food for thought?
Why not ask someone that can do astrology to do your stock picks and sports bets for you?
You know and I know that The Product is at best nowt more than 'astrology'. The Product does do search engine things though, and it could be scaled down to fit in a phone or even a watch, to be good enough for 'the pub quiz' or for writing a gormless email.
As for the article, META does very little for the vastness of the corporation. They have gazillions of developers yet Facebook and Instagram are as boring as ever, Threads and the Metaverse are just lame and what else do they do, apart from serve ads?
Microsoft had trialed datacenters in tents nearly 20 years ago. I remember hearing about their trials at some talks back in the day. Crazy to look back on the dates here, felt like it wasn't that long ago.
They also tested out sinking data center modules in shallow water to reduce the difficulties surrounding both cooling and land use. I wonder if anyone has revisited that idea in light of present market conditions?
These datacenters have been under construction since at least June of 2025. You can see 1 building was already up in 2025 and the land was just farmland back in 9/13/2024. So this construction has been over 1 year in the making. Does this mean its construction slop? For reference colossus xAI datacenter was up and running in less than 7 months. I couldn't tell you at what capacity but this doesn't seem like quite the same story.
Edit
For a little more context
xAI colossus 2 looks to be an empty warehouse on 3/10/2025.
By 12/2025 they had already either filled the warehouse or they couldn't use the space because they appear to built multiple structures outside for the datacenter.
For comparison again meta already had that datacenter there for a number of years and then over 2 years added those structures. In 9 months it appears Tesla built a datacenter into an existing building and added structures.
Off topic. The internet is so broken when someone still posts to X.com and then a news outlet posts that as an article. Maybe if it wasn’t X.com being used it wouldn’t be so bad. Seems like we can do better.
revealed preference theory is typically applied to individuals or single decision-making units like a household, rather than worldwide aggregate market behavior.
in any case, its not really that complicated in this case.
- the ai companies have a billion active users and billions of dollars in revenue.
- a poll comes back with 30% of respondents saying they are angry about ai.
so, why do the ai companies keep doing ai things despite ~30% of people not liking ai? well, its because they are making billions of dollars in revenue from their billion users. from the ai company's perspective, it would be madness not to keep shoving ai everywhere.
It can be simultaneously true that a large part of the public objects to a thing while the market generates profits from that very thing.
Of course the answer is for governments to step in as, as you point out, the AI-related companies are behaving as rational actors given their incentive structure.
I don't hate the data centers near me. However I do hate the tax incentives they were given. My local school district could really use those millions of dollars per year that someone decided we don't get.
But a billion active users != number of US citizens that take on the burden of AI. So go build your AI on land where your customers are if they like it so much.
Yeah in the states where the majority of people approve. But we already know the data centers aren’t being built in those areas. They’re being enforced on people who for the most part dont approve of their intent.
this does not matter from the business perspective.
microsoft does not care that your company forces you to use their products. google does not care that your school forces you to use their products. TSMC does not care that you are forced to use their products when purchasing ~any electronics. etc.
If a large proportion of people are only using AI because they are being threatened with unemployment if they don't, then there's going to be massive resentment building up
You may think that doesn't matter, but it does. History has shown over and over that you can only keep a lid on massive social resentment for so long before things break
I think good governance would listen to polls over metrics.
A good example of how this works is cocaine.
Capitalism and competition isn't always good governance. It works brilliantly in many places, such as restaurants or commodity goods. It fails completely for medicine or banking. It's in between for tech or education, but it's clearly failing for AI.
>I think good governance would listen to polls over metrics.
hypothetically, you own a widget company. you sell a lot of widgets. every month, you are selling even more widgets. the widgets are flying off the shelves. you keep ramping up production, and the consumers keep on buying.
gallup releases a poll that says "people hate widgets".
>why should a company listen to a gallup poll of ~1,500 people over their own internal metrics?
for the same reason Vladimir Putin should listen to Russian milbloggers rather than his own subordinates, the metrics are being cooked up by people who get promoted for good metrics
The public is against it for hypocritical reasons that they're not afraid to inflict on someone else. They're against the water and electricity use but they also angrily demand more semiconductor production that require building whole new reservoirs in Tainan, irrigation channels from the Han river, and coal plants in Anhui, as well as restarting Fukushima. Then when those companies come build plants in Arizona and New York, suddenly the reality of how the cow is butchered hits them and it stalls in permitting for years.
The public hates ai but also uses ai in mass quantities.
Capitalism abides by your dollars not your voice.
So people can decry ai all they want but if they keep using it, it won't go away.
Even then it's probable that AI is a big enough productivity boost for certain industries that even if no consumers used AI, businesses would still prop AI up enough for it to live on.
Nah, dollars buy war machines. And for the first time in human history, we are on the precipice of projecting substantial ground force without the need for humans.
Arguably, they don't take one in the chest. Luigi Mangione lives in notoriety what he did; most people can't even name the CEO he killed, let alone frame them as a martyr.
I disagree with theories of continual revolution, but it's pretty clear that class warfare still has valuable asymmetric qualities.
The public is forced to use AI at work and outside of work because the corpos are determined in inserting their AI everywhere. Then the people come back home and see that their energy bills have doubled because of AI datacenters. Of course people hate AI.
Yeah but the public is against progress. The public cries for material needs like medicare for all, universal childcare, a jobs program? These are all clearly foreign actors that want to prevent American progress on AI! They must be Chinese agents for all we know, what sort of American wants to provide healthcare for their family over proudly paying higher utility rates to ensure a new batch of tech bros become billionaires?