Cloud lobby warns EU: Clamp down on water rules and we'll evaporate The Cloud Infrastructure Service Providers in Europe (CISPE) trade body has put forward recommendations for the EU's Water Resilience Strategy, perhaps mindful that datacenters are perceived as hugely wasteful of precious water resources. My take on this is that maybe they are not moving as fast as the EU would want, given …
I don't particularly like cloud providers as an industry. However, one one thing I can agree with them - creating new water rules specifically for datacenters feels a lot like soundbite-driven policy. Datacenters use water, but so do a gazillion other industries, and some of them use a lot more water than datacenters, and I don't believe all of them are already maximally efficient. Enforcing water efficiency is probably a smart move, but it's really not clear why only datacenters need new regulation, if not just because they tend to get in the news a lot more. But that's not a good basis for sound policy.
What's different is that DCs new and potentially very large consumer of water with limited return to the system. Evaporative cooling takes water out of the local and regional water cycle. If I use water to flush a toilet, that water goes to be cleaned and released back into the river, with potential for further reuse downstream, likewise most industrial process water. Industries that historically used evaporative cooling (say power gen or steel making) would always source their own supplies rather than hoping to freeload on the public potable water supply.
very large consumer of water with limited return to the system
Its multiple problems wrapped in one. If they use the potable water supply then the city has to increase its treatment capacity, and its delivery capacity. Generally that means existing users will end up with higher bills to pay for the added treatment infrastructure, and if upping the delivery capacity means tearing up streets to put in larger pipes that's an inconvenience for everyone in addition to paying more for water.
Increasing capacity also means getting more water into the system. If it comes from a river or reservoir they need to take more, which affects downstream users (which in the EU will likely include other countries) or drilling more wells which may mean drawing down aquifers which is effectively robbing water from future generations.
Even if their location allows them to use grey water (i.e. untreated water from storm sewers or "lightly treated" industrial sources) that may avoid the treatment/delivery capacity issues but still leaves less flow or lower reservoir capacity for other users. Even if that's not an issue today, or only an issue every few decades during a drought, that may change during the lifetime of the datacenter.
The problem with sound policy is that it has to come from sound minds which, in the normal course of events, are not involved in this kind of decision making.
Water is getting more scarce, power is getting more expensive and land is at a premium. At some point we need to factor in people as well as businesses in how we make use of those resources.
We have to start somewhere and it's harder to withdraw existing permissions than it is to restrict new ones. If the lights-out, employment-free, water-guzzling, grid-draining AI megashed goes elsewhere then I struggle to see the downside.
An interesting thing that DCs might focus policy maker's attention on, is that generally consumers pay to buy treated water, but then pay about the same again* for the water company to take back and treat the same (assumed) volume of soiled water. Obviously in water constrained locations, there's actually a lot of value still in the soiled water, but the pricing structure doesn't reflect that because almost all users of potable water do return about the same volume as taken. The DC anomaly could be easily fixed by making an additional charge for non-returned water.
* In my part of the world
I think for many it’s the sheer waster of energy and water to produces reels of Ho’s with Big Arses and bad Llama eyes.
If the heat and water could be reutilised for agriculture Greenhouses/agriponics, Old Folks homes, council municipal pool,my Dad’s house etc … that would negate much of the negativity.
Hot/Warm Water just down the drain and out to sea is no use to anyone.
No, don't treat them like everyone else, treat them as the parasitic freeloaders that they are. Whether it's planning rules, energy connections, energy generation, hyperscalers and wannabees demand preferential treatment for a service that has yet to demonstrate much real value to society.
Agree, going through their list of recommendations demands, they can all be met at no upfront cost to tax payers by simply banning data centres /cloud operators from connecting to external water supplies. Also grade the water used for cooling as toxic waste so it can’t be simply poured into water causes.
Agree, going through their list of recommendations demands, they can all be met at no upfront cost to tax payers by simply banning data centres /cloud operators from connecting to external water supplies.
Surely this would be a simple problem for an 'AI' to solve? Hasn't anyone asked ChatGPT?
(which could be a fun question to actually try asking it..)
The first is to define a formal EU framework to support Europe-wide industrial water reuse/return schemes to enable the treatment and reuse of municipal wastewater for other uses, including datacenter cooling. This would help to prevent drinking-quality water being used for this purpose.
Not convinced there's any need to do this. Consumers are charged for water consumption and for waste water disposal. Drinking-quality water has often been through many sets of kidneys to recycle it. Datacenters could copy what other large water consumers have done and like NPPs, put their bit-barns on the coast. Which oddly enough is also where the cable landing stations are, although that plumbing often gets a little complicated. If they insist on bulding datacentres where water isn't, and water demand and waste management is already high like around the M25 and Slough.
So sounds like the usual special pleading, especially the suggestion of 'public-private partnerships' which normally just means socialising the costs.
By everyone, do you mean other water-intensive industries? Or everyone else?
In parts of Germany droughts are forcing water restrictions for many residents, but not for businesses such as Tesla's super subsidy slurping site at Grünheide in Brandenburg… Golf courses are also usually magically exempt from any restriction…
Treating them like everyone else means higher power and water bills for consumers. Is that what you want? Is that what you think is fair?
If they need a few hundred megawatts from the grid then HV lines may need to be upgraded or new ones built. New generation capacity may be needed. A new substation near the site is likely required. Normally that happens slowly, as cities expand and more houses are built. Those houses pay for power in the same way you do, so they mostly offset the added cost. A huge industrial scale user is different. They pay less for power than you do (buying in "bulk" as it were) which leaves more of the burden for that added infrastructure on you. Everything I've written above is also true for water.
But it is actually even worse than simply jacking up your power/water bills, because if we keep building huge AI datacenters at the rate we are we risk destabilizing our power grids - and it is a problem quite specific to AI. There is some indication that may be what happened in Spain. Here's an excellent description of the problem (it is a long article, but very informative)
https://semianalysis.com/2025/06/25/ai-training-load-fluctuations-at-gigawatt-scale-risk-of-power-grid-blackout/
Following the links, the water use "factlet" doesn't seem to be any sort of proper academic study, just a link to nearly ten year old web page by the Uptime Institute which appears to be a company touting consultancy to the bit barn industry.
It may or may not be accurate, but I'd hope for better evidence to drive policy than US merketing* guff from 2016.
* It was a typo, and yet it seemed appropriate.
If in areas like this one, which uses area piped hot water and local heat exchangers to handle domestic heating and hot water needs, the current gas-burning heaters could be replaced - or at least preceded by - a suitable sized data centre. If replaced, you'd need a bigger load in winter, of course...
District heating sounds like a good idea but as you say it has a much higher demand in winter. In the summer there's less demand for heat but the datacenter will keep pumping out just as much in July as it did in January. I suppose it is good if it can take a break from needing water a few months of the year, but that's not the best solution.
I think a better solution would be co-locating it near industrial users that need a lot of process heat and operate year round. Cement companies, steel mills, that sort of thing. Yes the datacenter produces low grade heat and those users need something much hotter, but heat can be concentrated with the same technology (scaled up a lot) that a home air conditioner / heat pump uses. There's some efficiency loss of course, but it is "free" so even if you lose 50% of it from datacenter rack to steel mill's furnace that's still a big win.
Solar thermal cooling* would be an option for these sites which can easily afford the necessary capital expenditure and, given the scale, the by-products could be industrially relevant.
* The physics for this are odd but it probably makes more sense for summer cooling than traditional air conditioning.
They can. Not directly but a heat exchanger can transfer heat from circulating clean water or cooling fluid to the seawater. Problem is, that's not really practical unless they're located very close to the ocean, and there will be environmental concerns - spot heating of the ocean near the outlet, fish and other stuff getting sucked into the inlet, and so forth.
If you're talking using the seawater directly in the evaporator that's not going to work, when saltwater evaporates it leaves behind a lot of salt which will clog it up in no time. Plus saltwater is highly corrosive, you'd have to design the whole thing to handle it. Much easier to design half a simple heat exchanger to handle it but then you have to dump the heated saltwater somewhere and unless you have a salt lake nearby that "somewhere" is gonna be the ocean.
The heat generated is a relative 6-8 degrees. Suppose: the fan output of an average server, sometimes idle, sometimes under load.
For a sterling engine you need enough energy to, say, expand something. Steam? Air? You would have to find a way to concentrate that 6 degree difference be be 20, and then another heat exchanger to bring it to 40, another to bring it to 80, another to bring it to 200.... think: multi-stage cryogenic cooling system, in reverse, and more. The cost of gain (in energy) due to loss to inefficiencies and equipment cost vastly outway the cost (to society) of just throwing it away. The harvested energy would not be self-sustaining.
The harvested energy would not be self-sustaining.
It can be, ish. It's something I've been working on for a house design. Basic principle is like this system-
https://www.oekofen.com/en-gb/pellematic-condens_e/
But I've been looking for a higher power solution. For 100kW+ most Stirling engines need working fluid at around 500C and there's FUN! with different fluids, ie some use helium, others nitrogen. Which can also mean FUN! materials in heat exchangers like inconel. And suppliers have a tendency to go bust, like this one-
https://en.wikipedia.org/wiki/Stirling_engine#/media/File:STM_Stirling_Generator_set.jpg
Which kinda started my journey. There's also been some interesting developments like hydroacoustic heat engines that are sorta Stirling engines, but not. And also not currently commercially available. Which is the biggest issue. For them to work well, installations need to be designed around a cogen/CHP solution, which isn't very easy to retrofit into existing building designs. And then issues around building codes, especially on the residential side. Which is something I've been working around and just moving the power/heat generation to the workshop I want. Then it's desiging the HVAC to use 'waste' heat, or recycle that, plus dumping some into a convenient heat store.. aka swimming pool.
And then trying to keep everything as simple as possible & low maintenance. Capital cost is high, but it should be pretty much self-sustaining using a combination of solar, heat exchangers and wood. Which is also an issue around systems like Oekofen that rely on pellets. I have a lot of trees, so want something that can run off a regular log burner and maybe a wood gasifier.
Problem is, that's not really practical unless they're located very close to the ocean, and there will be environmental concerns - spot heating of the ocean near the outlet, fish and other stuff getting sucked into the inlet, and so forth.
They're pretty much solved problems, ie NPPs around coastlines. The spot heating isn't really an issue given thermal capacity of sea/ocean vs the amount of heat dumped. But that can have some interesting effects, like getting big crabs & shrimp from around Sizewell. Just need to watch out for the Deep Ones swimming along from Dunwhich to bask in the warmth.
They just do not want to manage the water they use properly / expensively.
If they are going to "waste" fresh water via cheap & nasty approach of water based cooling where they lose water from their system then that's hugely detrimental to everyone else.
There are alternatives e.g. sited near the sea and use seawater (more corrosion issues, but the knowledge is there on how to deal with it), or use closed systems where you do not waste water (more expensive which is what the data centre companies hater)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
