Datacenters face rising thirst as Europe dries up Water scarcity is rising up the agenda as one of the major concerns for datacenters in Europe following an unusually hot and dry summer, marked by intense heatwaves in southern parts of the continent. While much attention has focused on the soaring energy consumption of bit barns due to factors such as higher-density …
A simple fix would be to build them by the sea, which has extra costs for longer links, but that isn't a big hairy deal.
The most significant business centre that's a long way from the see is probably Frankfurt, and even that's "only" 350km as the crow flies. An extra 7.4ms of latency might be significant for a tiny number of financial services gamblers doing HFT, for the other 99% of DC loads it'd make zero difference. And that's just about the worst case in Europe. Berlin's of the order of 140km from the coast, Amsterdam about 10km, Paris 110km, Lyon about 180km, Milan 100km, Munich 200km, Madrid 220km, etc etc.
Water shortages are likely to affect electricity generation before cooling – this has already been a problem in southern France for many years. For data centres, other forms of cooling are possible and water used for cooling can, in theory at least, be kept in a closed loop. As for using coastal water – that's not really possible in much of the eastern North Sea and southern Baltic and comes with its own problems.
Clearly, anyone who can develop efficient waterless cooling, or data centres that require less of it, is going to have a commercial advantage.
@El Reg, stick with one set of units: cubic metres, gallons and litres all in the same article is unnecessarily confusing. For water use the international standard would be cubic metres.
This is what I was wondering. Although there will be some loss due to evaporation, how much water is actually consumed?
I'm sure that they can't be taking mains water in, using it once, then sending it out as waste water.
I know that many data centres use closed loop water circuits, and put gas quenchers and anti-cavitation additives into the water, so it is not correct that that water should be poured down the drain, but I do know that the secondary coolers often use evaporative cooling to cool the primary circuits.
I remember being told that IBM's own datacentres at Warwick and North Harbour used to use large cooling lakes around the buildings as heat sinks when they were built in the 1960s, but I'm guessing that the largest datacentres now probably produce more heat than can be dissipated using that method.
Given water prices in Germany, using large volumes of mains water for cooling would quickly become prohibitively expensive. And, with regard to the example of Frankfurt from the original post, much of Hessen is already considered arid with water conservation plans in place most of the year with priority given to drinking water, agriculture and power stations.
Water cooled heat exchangers should be sufficient in most cases – and these can be powered for most of the year by small solar installations.
"@El Reg, stick with one set of units: cubic metres, gallons and litres all in the same article is unnecessarily confusing. For water use the international standard would be cubic metres"
Indeed, but the El Reg standard for volume would be the grapefruit and the Olympic-sized swimming pool.
https://www.theregister.com/Design/page/reg-standards-converter.html
"As for using coastal water – that's not really possible in much of the eastern North Sea and southern Baltic and comes with its own problems."
Why not? Google have a DC in Finland that's sea water cooled (admittedly more northerly), but various power plants all around the North Sea and southern Baltic have used sea water cooling or still do. Higher sea temperatures may require more conservative design than hitherto, that's still feasible.
Why not? Various reasons but mainly because the North German plain – which was seabed extends shallowly into the sea. This means delicate, actually increasingly delicate ecosystems, that are already suffering due to the faster rise in temperatures. The area is also relatively densely populated and you extensive agriculture You also have to put pipes further out to sea and add more filters, which drive up cost significantly.
Yes, I know about Google's plant in Helsinki and a few "floating" data centres in the Baltic might be possible. But not the North Sea which has much higher tides and fiercer storms.
A simple fix would be to build them by the sea, which has extra costs for longer links, but that isn't a big hairy deal.
Stick them in Scotland, lots of cold water & air and a surplus of electricity from all the wind farms which would not then have to be given constraint payments to not generate. They could ship the data to/fro via longer fibres, which as you say is rarely an issue. Such fibres could be easily buried, so a better option than lots of pylons to get the electricity to where the data is.
That bit of common sense will be ignored.
The UK Bit Barn operators seem to want them within a Uber ride of the City of London and adjacent to the M25.
FSCK the lot of them and make them provide their own cooling water supply and not use our drinking water. There is plenty of 'dirty water' only a few metres down from the surface in the London Area thanks to the rise in the water table. They could use that but... it would need treating which costs money and they want us to pay for their clean water.
Actually, it's a bit more than a few metres down - this is the Chalk aquifer, which has been rising since the 1960's, due to the decline of industry in the London area.
The Chalk is typically around the 30-80m below ground (varies quiet a lot, depending on where you are in the London Basin, and can be shallower) but that's still readily pumpable.
The Environment Agency have a Chalk Aquifer management strategy that is controlling the rise in the aquifer
(See here: https://www.gov.uk/government/publications/london-basin-chalk-aquifer-annual-status-report)
Also, it's not really 'dirty' - The Environment Agency tend to have a distinct sense of humour failure with people that pollute the Chalk aquifer. If you are only using the water for cooling, it shouldn't need treating. It's quite routine for Ground Source Heat Pump systems to use open-loop design (extract from the aquifer upstream and re-inject downstream) that operate in the Chalk aquifer.
The big problem with water drawn from a chalk aquifer is that it contains a lot of dissolved calcium carbonate (chalk) which likes to deposit itself in the pipes.
When I lived in Basingstoke, the water was so full of the stuff that it would block pipes in a matter of a few years. We had to have the supply to our washing machine replaced more than once while we lived there, and it didn't do the machine itself, kettles or the hot water tank much good.
If you don't do something to treat the water, you will have problems.
Maybe it's time for large water consumers to be treated like large gas an 'leccy consumers? Give then a deal whetre they are the first to be shut off during a shortage. Of course, a water shortage can last months rather than the usual few hours or days of gas or leccy shortages, so it might hit them a bit harder. maybe only apply it to DCs that run AI models? :-)
"Stick them in Scotland, lots of cold water & air"
Yes, but the article is taking a European perspective, so there's the slight matter of a the need for links from mainland Europe to Scotland (and possibly redundant links). It's a thousand km from Paris to Edinburgh, 1,700km Berlin to Edinburgh, that's both expensive, and an addition to latency that could start to become a problem for some applications. Given all the data sovereignty concerns, why would the German or French government or companies want to be reliant upon very long data links through perfidious Albion?
"and a surplus of electricity from all the wind farms"
What "surplus"? Grid operators are investing billions in at least three major new interconnectors to England, in addition to the growing interconnections to other European nations. The days of constraint and "surplus" wind power are numbered.
You'd have to desalinate the water before use, as the minerals + salt will either corrode or clog the cooling systems long-term. The cooling systems can be designed to compensate but I do not think that has yet to be done to electronics cooling, as the cooling circuits are very small diameter. It would certainly wreak havoc with typical copper or aluminium cooling plates.
No you wouldn't need to desalinate it. Nuclear power stations are often sited on the coast to use the sea for cooling, and the technology of dumping heat into sea water without your plant corroding far enough to melt down has been well understood for many decades. The main element is that you don't use sea water for any primary cooling circuits (and probably not for secondary either).
There's already DC applications where sea water is the heat sink.
I know / knew you'd mention nuclear reactors, but their large-diameter cooling systems have the capacity to deal with low levels of corrosion. A 2mm cooper coolant pipe on a heat block?
Far less so.
They already purify water for data center use, for example
https://www.saltworkstech.com/brochures/data_center_cooling_water_reduction_and_recovery.pdf
You can use saltwater in the secondary loop but nuclear reactors spend millions on demineralization plants
https://duckduckgo.com/?t=ffab&q=demineralize+nuclear+reactor&ia=web
The primary of a server farm would still need fresh water for the small block & pump channels; if the secondary cooling loop is evaporative you'll still, as noted in the PDF, need to treat that large quantity of water in order to avoid massive mineralization of the systems.
So sure, it can work - but, like I said, you're adding big cost factors onto the development and rollout of the project, versus a far less complex filter system when you 'just' hook up to (already pre-filtered, fresh water) municipal systems. Things don't happen in a vacuum - if it was so easy to place data centers on shorelines, wouldn't they be doing so already??
"'just' hook up to (already pre-filtered, fresh water) municipal systems. Things don't happen in a vacuum - if it was so easy to place data centers on shorelines, wouldn't they be doing so already??"
While the see (and are charged for) water as if it's an almost free and unlimited resource, why indeed would they look at alternatives? Of course, the whole point of the article is that across much of Europe, water is becoming a less cheap, less abundant resource, so "just hooking up to the municipal systems" is likely to become less and less viable if people have to choose between having a drink from the tap or getting fast answers about $whatever from ChatGPT.
This is where the huge scale of these facilities simply has to take all factors into account.
You cannot build something on the assumption that power and water is magically available and bluntly if you using water to cool and it is in short supply just throwing money to pay a higher price is not the solution.
The reality is that the huge expansions of these data centres and the increasing demands on resource are not sustainable. Unfortunately what actually happens is these very well funded corporations just pay to make the problem go away. That impacts all of us as water is required for some much more than keeping servers and storage cool whilst it runs worthless AI, Social Media or storage of cat photos.
"That impacts all of us as water is required for some much more than keeping servers and storage cool whilst it runs worthless AI, Social Media or storage of cat photos."
What about "adult content"? Surely that is the finest achievement of the internet. If a few farmers have to swap from maize to olive groves so I can indulge my baser instincts then that's fine by me, I like olives.
What i see here is a market challenge.
We have a scarce resource that has three use cases:
1, Where we use it to drink and keep ourselves clean
2, Where we use it for irrigation so we can feed ourselves
3, Where we can use it for industrial uses, hosting cat videos and other things
I suspect that we have an inverse incentive.
That the users in use case 1 should only be charged the lowest possible rate - its low on the hierarchy of needs after all.
That the users in use case 2 are also in need of lower rates to avoid inflating the price of 2, which is also low on the hierarchy of needs.
The users in use case 3 are by default the highest in the hierarchy of needs and consequently should pay the most. However, they also use vast quantities and therefore demand low prices or subsidies to 'bring jobs to the area'.
These lower prices or subsidies are paid for either by 1 or 2, either in their water prices or in terms of higher taxation to pay for 3.
So we end up in a perverse situation where the least valuable use case gets a preferential rate vs those of the higher value.
One option is to have a 'left over' model. We calculate the amount of water that's required according to the hierarchy of needs at each level and then what's left is passed up to the next level. If there's none left for cat videos, that's decision made. Of course the cat video folks can go talk to the water management companies and ask for more water to be created somehow, but this should have no impact on the current application of other use cases.
Yes, we'd need to have some sort of hierarchy of importance of cat videos vs farming, but that might be a useful exercise in public education. Most people are concerned with current, small, but noisy problems, rather than realising how lucky they are and how good things are.
They don’t need to consume water at all. Dry cooling just needs bigger cooling towers.
Examples of Dry-Cooled Power Stations:
Kendal Power Station (South Africa): Uses an indirect dry-cooled system.
Matimba Power Station (South Africa): An Eskom giant using direct dry-cooling technology, significantly reducing water consumption in a water-scarce area.
Majuba Power Station (South Africa): Employs direct dry cooling.
Kusile Power Station (South Africa): Another large station that utilizes direct dry cooling.
Medupi Power Station (South Africa): Also uses direct dry-cooling technology.
I suspect power stations operate with much higher flow and return temperatures than ICs, high return temperatures are not so good if you're trying to dump a few megawatts of DC heat.
That doesn't mean air cooling isn't possible - it works just fine for most PCs - but for the volumes of heat at a DC and the necessarily low return temperature moving to air cooling has its own problems. As with all such engineering challenges, they're technically resolvable, but the cost of doing so is usually the problem.
The AI bubble will deflate soon, and there will be surplus data centre capacity for everyone for a few years whilst the industry crawls away and licks its wounds for a while, until they think up another scam.
Using less software, holding less data (it is a risk not an asset), using simpler software, holding your own data on your own systems, and using hybrid systems (photocopies of staff passports in a locked cupboard) are all good solutions for more secure systems, that will require fewer data centres.
More data, more computing and more processing increases costs without adding anything to your revenue. They are scamming you when they claim that Big Data or AI will make you richer. Get your heads out of your arses and do your sums.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
