Datacenters have a public image problem, industry confesses to The Reg The current craze for AI has helped drive a wave of datacenter building, but the industry has run into opposition from local communities in many areas, something it is understandably keen to address. [People are] saying, 'How can we decide to put datacenters everywhere? We don't want them.' But the same people use that for, …
Or it's because we've had ample examples of science giving us some miracle thing only to discover there's a hefty cost. Take Chloro-Fluro-Carbon as an example: Miracle refrigerant. Just a shame about the environmental impact when it leaked.
Or look at all the wind farms in Scotland: Great, but the grid can't cope with all that power when they're running to capacity so sites get paid not to produce power, at public expense, and now we've got fights over 'wind theft' to contend with, too!
So is it really a surprise that people look at the latest thing and wonder: Do we really need all this now, or should we go a little slower and only build to what we need now plus a little extra to handle growth for, say, the next five to ten years? Only I know darn well the argument then is 'but we don't know what that'll be like and we need to go big in case'... but the people don't see that: They see another carbuncle being erected, making a mess and disrupting lives: Too much change, too quickly, and with a view that it is just another vanity project or a wagon for people to jump on only to be an utter waste of time and effort, leaving another abandoned eye sore to blot the landscape.
Plus: Do they really need to be built on greenbelt? Sure, it's cheaper, but shouldn't they be forced to build on brown belt and without the subsidies the government inevitably offers as incentives? It's not like that ever gets abused, is it? No examples of, say, companies building cars and pre-registering them just to make it look like sales are up so they can claim subsidies (if you've seen those reports) or the mountain of e-bikes in China from start ups that go as far as ordering the bikes, getting the government grant, to just abandon everything (except the money) to rot...
The real function of these projects isn’t infrastructure - it’s optics.
Politicians don’t subsidise datacentres because they’re needed. They do it because it lets them point at a field full of air conditioning units and say: "Look, investment. Progress. Jobs.”
Doesn’t matter that the jobs are minimal, pay is shite, the power’s gone, and the land’s ruined. What matters is the staged photo op, the press release, and a new bullet point on a re-election campaign.
These datacentres aren’t being built for the public - they’re being built on the public. The taxpayer funds them, the community lives with the fallout, and the profits vanish offshore.
And importantly the 1000s of jobs that they allegedly create are only during construction, not operation.
Just like these huge distribution parks, the jobs are based on some arbitrary government figure that that huge developers like Panattoni have managed to get accepted. It is something like 1 job per 100 square meters. Utter bollocks, that is a 10m x 10m square.
"the jobs are based on some arbitrary government figure that that huge developers like Panattoni have managed to get accepted"
I think there's less science than that. Process is as follows:
Government consent the planning of an appealed DC in a contested locations against the will of local planners;
Government consent the planning application, because all business investment is a direct contribution to GDP and the Canutean quest for "growth";
Government civil servants ask the developer how many jobs will be created;
Developer thinks "well, about 6, but that's not the required answer";
Developer works out how many unique individuals will visit the site at least once for at least 2 seconds and says to the civil servants "1,400 jobs, we've checked";
Civil servants tell the Secretary of State's Private Office that the jobs figures are bullshit;
A Special Adviser to the SoS tells civil servants to fuck off, and writes a press release themself, starting with the 1,400 figure, and applying a 3.1 multiplier because those 1,400 people must spend their salaries and create more jobs, right?
And so a new DC creates over 4,300 jobs - it's a fact.
Exactly. They're relying on the public being ignorant and assuming that "Ooh, big shiny tech thing means lots of big shiny tech jobs for people working there". Missing the fact that most of the high-quality, high-tech jobs they're likely to generate will be remote with surprisingly few local roles.
I've said before that your average data centre needs two in-person employees- a security guard and a dog. The dog being there to bite the guard if he attempts to fiddle with any of the computers. (*)
This is obviously a joke, but only a partial exaggeration. Data centres offer few benefits to the local areas in which they're placed, relative to the significant disruption and demands they place on the infrastructure.
It's easy to talk about NIMBYism, but why the fuck *should* they be expected to want a massive, ugly and disruptive monstrosity in their back garden when they're getting little in return and the people reaping most of the benefits are several hundred miles away?
And when it comes to data centres built specifically to serve big business' greedy, self-serving attempt to win the AI race with endless self-aggrandising promises, it's not clear that even *those* "benefits" exist.
Regular data centres are a shitty deal for local communities and AI data centres are worse.
(*) Yes, I did indeed rip this off an older joke about aeroplanes.
Amazing how the datacenter execs gathered to whinge about NIMBYism without once acknowledging the obvious: people don’t want datacenters because they don’t benefit from them.
These hyperscale bit barns don’t host community projects or serve local needs - they’re energy-guzzling, water-hungry fortresses for Big Tech to hoard compute, automate layoffs, and extract value from everyone else's data.
The irony is rich: billionaires want sympathy because their profit engines are unpopular, while those same facilities push up electricity prices, suck public infrastructure dry, and offer nothing in return except faster surveillance and monetised dopamine loops.
You’re not building "public utilities." You’re building private infrastructure for private gain, propped up by public subsidies and political lobbying.
The public gets it just fine. That’s exactly why they don’t want you in their backyards.
These hyperscale bit barns don’t host community projects or serve local needs - they’re energy-guzzling, water-hungry fortresses for Big Tech to hoard compute, automate layoffs, and extract value from everyone else's data.
That's the crux of it. DC's are important. Our modern world is built on digital services. Even our phones are mostly VOIP now. But this gold rush for 150kW racks to power AI workloads is not actually benefitting anyone. Sure, we see reports that tech vacancies are up, with people hiring AI-"specialists". But how much of that is down to actual business and how much is it execs stuffing some money into "AI" because of FOMO. Sure, I'd rather see the cash go out in salaries than as a stock-buyback or dividend. But it's not actually generating value or product. And really, these massive AI datacentres don't benefit locals (or anyone really) the way a telco datacentres does (which is legit utility infrastructure).
But it's all part of the act - muddy the waters that "Datacentres are essential infrastructure". When the truth is that Some DC are essential infrastructure. Some are luxuries, and we should be more discriminating in which luxuries we indulge in.
You know what would be an amazing concept? A community data centre. A room with some mesh-fronted lockers run by a non-profit CIC who has some IPs and a 1-10Gb backhaul. For a network provider like CityFibre, the traffic would be mostly upload and therefore free (since their customer traffic is going to be heavily download). So if you were reasonably on-network geographically, then they should be able to charge a fairly sensible monthly connection fee. Basically a step up from a well-organised hackspace.
You get a power socket and ethernet port in your locker and that's that. You can drop in a couple of RPis, an HP Microserver, whatever. £20/mo, done. All to the benefit of the local community for local workloads.
If the power goes off, you lose connectivity for a while. But does that actually matter? At my first proper job (late 2000s), our public website and customer portal was run on a Dell tower server in the office hanging off an SDSL line. Which part of our infrastructure represented a single point of failure? All of it. The server, the crappy 5-port 100Mb unmanaged switch, the dreadful BT router... I had most faith in the cabling - because I crimped that Cat5 myself.
That wouldn't be tenable today given the growth of that company. But for lots of firms or enthusiasts... it'd probably be fine? How much are we paying for 5x9s hosting when really "available most of the time we're awake" is sufficient. (See also: Low Tech Magazine). There's a fair case to be made that modern DCs have excellent PUE values and are more efficient. It's possibly also the case that distributing our compute to smaller on-prem deployments can be cheaper and offer better experiences to local (LAN) users.
Not everything needs to be K8s in AWS with 99.999% reliability, because at the end of the day - most of those big boys go to sleep occasionally as well when they drop a region from their routing table or fudge their BGP.
For a lot of people, what I've described is enough. They don't need to pay £nnn/mo for a quarter or half rack. But as it stands, if you're not at that point, then you're directed to spin up some VMs with a hyperscaler. I suppose that concept is undercut by the fact most domestic lines these days have at least 20Mb upload, which is fine for most "self-hosting" situations, provided you can get a static IP.
You could have a nonprofit datacenter, but you've basically just reinvented the colo. We have private colos already and they work fine. In many cases, there's not a lot that a nonprofit one would do that a private one wouldn't also do. The only possibility, cutting costs by eliminating normal utilities like UPS or inter-user security so they could cut prices, could easily be done by private ones but it's probably good that it hasn't.
You could have a nonprofit datacenter, but you've basically just reinvented the colo. We have private colos already and they work fine. In many cases, there's not a lot that a nonprofit one would do that a private one wouldn't also do.
Well, I've defined a particular flavour of colo that doesn't really exist (Tier 0.5-1 if you will). Back in the wild-west days there were "data centers" operating on a wing and a prayer with non-redundant power supplies and transit (although some would tell you they had - that was perhaps "aspirational"). But if you wanted cheap and were willing to accept that risk, then fine. Now it's actually kinda hard to find colo that isn't Tier 3-4, with a list of certifications as long as your arm. And that's an important part of the ecosystem, but it's also excessive for a lot of people. It adds cost to business and eliminates risk that some users would be happy to accept for non-critical workloads.
The only possibility, cutting costs by eliminating normal utilities like UPS or inter-user security so they could cut prices, could easily be done by private ones but it's probably good that it hasn't.
Security is important of course. N+N/N+1 power, N+1 generator and diverse fibre - even a UPS - actually isn't necessarily a requirement for lots of people. People spend a lot of time on it, and then their office broadband goes down (usually single-homed, single point-of-failure) and the highly available, five-nines cloud service isn't accessible! Of course in that case, you're wanting on-prem anyway.
There are some interesting community projects out there. NYCMesh have done an incredible job building an infrastructure commons across the city. Their primary interest is connectivity, and they encourage self-hosting. What I've described is kind of the colo equivalent - run well, to a good standard, but taking a pragmatic view to risk and the diminishing returns of additional redundancy, etc. Not "bad" but saying "This is what you get. It's less, but it's cheaper and more accesisble". But perhaps also obsolete these days given higher upload speeds from residential connections that allow more self-hosting (unless you want to run SMTP, etc in which case you need a tame ISP who will open those ports).
The TL;DR is that for some people, you don't need a hundred-million pound custom campus with all the utilities. A converted office unit with good physical access controls would do the job, but because the market all feel the need to accredit themselves up the wazoo to be taken seriously, everyone is stuck with higher costs. In fairness, there are now some "edge" datacentres emerging who claim "Tier 3 alignment" rather than certification, which seems to show a move back to that.
Your description may work for AI datacenters, but your more expansive phrasing makes that the one argument the people from the article actually countered. Before AI, a lot of those datacenters did, in fact, provide services to the general public. Not a public service as in something without a profit motive, but people who like using the internet were often getting a lot of what they chose to download out of one of those buildings.
That's a pretty good reason not to subsidize them; the private companies will be able to afford them otherwise. It's a good reason not to fight about where they go; they're not actually creating waves of jobs. It's a good reason to make them handle their own additional power and water needs by paying for the increased infrastructure, which they often already do. But as with a lot of other private infrastructure, the average person benefits from its existence more than they know, and any statement as broad as "people don’t want datacenters because they don’t benefit from them" is wrong about any person who uses the internet.
Sure - people benefit indirectly from general internet infrastructure. But that doesn’t justify the current wave of hyperscale AI-focused datacentres being dropped onto local communities with government incentives and minimal public input.
There’s a meaningful difference between a CDN edge node helping you stream a YouTube video, and a water-cooled, multi-gigawatt AI compute farm being built to train proprietary models no one asked for - models that will then be sold back to the public, often to replace their jobs or monetise their behaviour.
And yes, companies can afford these facilities without subsidy - which is precisely why the political photo-ops and taxpayer incentives ring so hollow. If these were just mundane bits of infrastructure quietly doing their job, they wouldn’t need special economic zones and reclassification as "critical national infrastructure."
The average person did benefit more clearly when datacentres powered a more decentralised, open web. That era is fading fast. What we’re looking at now is something else entirely - high-impact, low-return infrastructure built for profit concentration, not public good.
While there are politicians who don't get it, a lot of people will try to see if they can get someone else to pay for their new building with bad promises. We should reject those applications if we get a chance. Keep in mind that many of these arguments aren't about government-funding, or that would be the first, most convincing, and likely only argument. They're about not wanting the buildings there at all, even when they're privately funded. There are some valid reasons not to want them there, but as long as the people making those arguments stick in the "I don't benefit from datacenters" argument, they're hurting their own point because they benefit from many, and this might be one of them. It's not only the CDN nodes that send static content, either. Plenty of DCs that have more processing capacity are important for many people, and if it's not you, it is your neighbors. I may not use TikTok, but a lot of people do and they want their videos chopped up and filtered and distributed, and that takes servers. If my neighbors switched the script, they could name things I use servers for that they don't. The argument of benefit is never going to work well. Those who would prefer for the DCs not to be built would do better to focus on the harms they cause rather than the benefits they think they're not getting.
How much of that is due to the advertising industrial complex and now AI? If it weren't for those two things, there'd be a lot less data about everyone in data centres.
And...
"But AI can also decrease greenhouse gas emissions or lower overall emissions at a company level"
Bullshit.
The companies selling phones and OSs want to pretend that this "AI agent" is working for you, the customer. So, they play down the fact that it is in fact a giant Shodan-like program residing in their private datacenter. They want you to give their thing permission to buy things online, run programs on your desktop, make calls on your behalf...everything! They definitely don't want you thinking about how its training data and prompts might be tainted by nefarious (or good) intent, about what might happen if some hacker removes the ethical constraints from it, how a government or corporation might use it to shape your worldview, and a thousand darker uses that sound hyperbolic but if they occurred to me just now writing this sentence, you know much nastier people have been thinking about them for months.
You think the people who don't understand the technology are afraid? Try talking to people who do.
Yes, both of those things use datacenters too. However, the servers we're getting the article and forums from are almost certainly in a datacenter. So are the ones that provide any other website we've visited today. If we eliminated both advertising and AI, we'd still need a bunch of datacenters, and the only reason our need for them would decrease is if the loss of advertising killed off most websites. You already know this, I'm sure. So much of what we do involves sending data through the internet, and almost all of that traffic goes to some datacenter. It might be a private one, only operated for one company. It might be a colo. It might be a cloud one where lots of people rent space. But it's almost always one of those. That is not going to change.
Classic IQ vs EQ problem.
How is it news that everyone wants the nice bits of something without having to suffer the consequences of that nice thing?
I want decent roads, but I don't want to pay the tax to make them decent.
I want a stable electricity supply, but I don't want any power stations or distribution network disturbing my view.
We live in a world where the gap in understanding between production and consumption is so vast that I'm bewildered that supposedly clever people are surprised.
This isn't about people whining that “they want the benefits without the consequences.” It’s about the fact that the consequences are local, and the benefits are global - and mostly captured by a handful of trillion-dollar companies.
Your roads analogy would work if these datacentres served the public the way roads do. But they don’t. They serve hedge funds, adtech, and job-eating AI.
Communities aren’t confused freeloaders. They’re rightly pissed off at being asked to sacrifice land, water, and grid capacity for infrastructure that makes companies like Google richer while giving them nothing back but higher bills and surveillance.
So no, this isn’t an IQ vs EQ problem. It’s about exploitation. The public sees it clearly: they get the noise, the water shortages, and the power drain; tech giants get the profits, the subsidies, and the AI arms race. No amount of smug analogies is going to cover that up.
Yes exactly. The DC-builder mouthpiece says "But people don't understand that these bitbarns power their TikTok videos" when actually, the world would be a much better place without TikTok, so what useful purpose do these energy-glugging barns serve again?
"People think they can be built anywhere but they can't" - They CAN be built anywhere, except when they are serving the high-frequency traders, who are the reason the stock market goes to hell after every fart that comes out of Trump's arse. They cream off the top and we all know it's a zero-sum game, so the loser is your pension fund. What useful purpose do these bitbarns serve again?
"What useful purpose do these bitbarns serve again?"
Well, as a pornoisseur, I can assure you these bitbarns and their AI models aren't producing any advances whatsoever in the world of manly self-enlightenment.
<fx: checks for the thousandth time that the "post anonymously" box is ticked before pressing "Submit">
You are completely right.
When I wrote that I questioned the use of the term nice since it's subjective.
However, I very, very much doubt that people are really complaining about the water usage, etc.
The primary concern that I've seen reported from the reports in the US and UK is that these buildings are huge and disruptive to the locals who live there. They're often pushed through against the wishes of the local people because the economic benefit to the region is greater than the desire to maintain the current local look and feel.
The secondary things that then come up are water usage etc, once they lose the argument about maintaining the local feel of the proposed location.
Correct for traditional datacenters, incorrect for AI ones.
For both, the buildings are huge and disruptive, consume vast amounts of power, and often vast amounts of water. These are quite legitimate complaints!
For traditional datacenters, the benefits are mostly the profits for the companies building them (so stop giving them government subsidies!), but do have some benefit for locals, region, country, and world.
For AI datacenters, they consume MUCH more water and power, while providing no apparent benefit to anybody other than the companies. So the locals are absolutely justified in being against them!
Show a picture of a data center with typical uses. That rack contains your isp email,that rack contains Netflix movies,that rack contains BBC radio/iplayer. Realistically it will be more boring such as company xyz data. Arguably government databases are important ( I couldn't sell that sorry). If people can relate to what it contains they maybe more tolerant.
Show a picture of a data center with typical uses. That rack contains your isp email,that rack contains Netflix movies,that rack contains BBC radio/iplayer. Realistically it will be more boring such as company xyz data. Arguably government databases are important ( I couldn't sell that sorry). If people can relate to what it contains they maybe more tolerant.
The problem being that they can't do that - and they know it. The bulk of these new DCs are going up for AI because old DCs can't support the per-rack power/cooling demands. If they were doing something genuinely useful like "here's your email", then yes, people would be tolerant. But they're being asked to watch greenfield sites get paved into GW-scale bitbarns (plus gensets & ancillary infrastructure), and to give up their water supplies for evaporative cooling (and somewhere to dump the toxic brine), all in pursuit of a chatbot that lies sporadically.
And for (the island of) Ireland in particular this is a touchy subject. Loch Neagh is still in the grip of total ecosystem collapse. Turns out the 2023 algal outbreak, collapse of the eel fisheries and subsequent wetland-bird population crisis wasn't just the "bad year" that politicians were praying for (so they didn't have to fix it). It's as bad as it was. Ireland's largest freshwater body is fecked, and likely to remain fecked for the foreseeable future. In fairness, this collapse is not due to datacentres, but has rather focussed attention on the (mis-)allocation of (finite) natural resources and the economic effects of their collapse.
"But AI can also decrease greenhouse gas emissions or lower overall emissions at a company level,"
Excise me, what?
It goes on to say racks went from 10kw, to 150kw and soon to 600kw, each.
How exactly is increasing the power used year after years, (non linearly?) reducing green house emissions.
"But AI can also decrease greenhouse gas emissions or lower overall emissions at a company level"
Feck off, you either know that's utter bollocks or you've had a pickaxe lobotomy.
Each of those 600kW racks would have to perfectly replace about a thousand people to be net zero.
600kW is 5.3 GWh per year. Double it for the cooling, so 10GWh per year. With UK power mix, that's about 5,300,000kg of CO2 equivalent.
UK mean per person CO2 is 5000kg (2022 figures). Almost exactly a thousandth.
Composting the CEO of the hyperscaler would offset it too, but you only get to do that once and there's definitely more than one rack in there.
But AI can also decrease greenhouse gas emissions or lower overall emissions at a company level,
I am very disappointed the register allowed such an obviously unjustifiable statement go unchallenged.
What we're saying is that, for sure, AI will be demanding more capacity or more power supply, for sure, but still, we have quite a lot of power availability in the different markets.
Only by hoovering up all the capacity added by every new renewal scheme and the the capacity freed by the shutting down of major industrial users due to the higher costs of the former.
<......"Most people are f**king scared of AI, like we're feeding a monster"....>
I am not sure that scared is the right word, but many people feel the way they do about supposed AI because they can already see that the bit that is being thrust at the population at large <is> a monster.
Walsh is right about the public image problem though - and it I think it is he and his ilk who are scared that the public/population at large might start to see through the con that he and his associates are perpetrating with pushing stuff at them as Artificial Intelligence, when it is nothing of the sort. Not only that, but it is misleading or just plain incorrect in its replies far too often for it to be considered useful for anything.
No one that I know of is “f**king scared of AI”. A lot of us are fed up with the inaccuracies it’s spewing out that far too many people are stating as facts, and having it rammed down our throats. See also the article about how it’s becoming harder to buy a laptop / PC without AI capabilities. Yet again, it’s marketing droids coming up with solutions to problems that don’t exist.
"people are clearly aware of how essential water is, but NIMBYism means they still don't want a reservoir built where they live, as anyone in the south of England can attest."
We have long had privatized water, main modus operandi of these companies seems to be to rip off the (captive market) consumers as much as possible whilst spending as little as possible (hence lots of leaky pipes, sewage discharge into rivers / sea as saves costs, especially with paltry fines so much cheaper to keep leaking & polluting).
Building new reservoirs & infrastructure in general is expensive, time consuming & only really benefits customers (it is a bottom line hit) so zero incentive for water companies to do it. There will be a few half hearted planning applications so water companies can say "look, we are trying" but you will see a distinct lack of dogged pursuit of these reservoir building applications (if all else fails you can get compulsory purchase orders for critical infrastructure such as reservoirs, which are arguably a lot more important than HS2 which pissed off lots of people with many compulsory purchase orders. However with current huge (unfixed) leakage on water networks no compulsory purchase orders likely to be granted until effort made to fix existing leaky system)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
