China says 'world's first' offshore wind-powered underwater data center has entered full operation, houses 2,000 servers — 24 megawatt subsea AI facility uses ocean water for passive cooling and offshore wind for power
Posted by sr_local@reddit | hardware | View on Reddit | 155 comments
UltimateSlayer3001@reddit
We’re almost finished polluting and destroying the land and air, now it’s time to move towards the sea. What a joke.
toybuilder@reddit
Ocean datacenter makes more sense than space datacenter IMO.
boringestnickname@reddit
Does it really, though?
Do we want to mess with the ecology of the ocean even more than we're already doing?
mechkbfan@reddit
Last I read that the energy we add via these methods is a rounding error compared to carbon & how much energy the sun pumps at is every day
nomad5926@reddit
Yea but the amount of energy from the sun and carbon is in equilibrium with how the aquatic ecosystems operate. You really want to tip it too much to one side?
mechkbfan@reddit
As per my other comment, if you built a data enter that was 1km x 1km x 1km, that's 0.00000007215007215007215% of the ocean.
Will it likely mess with some schools of fish? Yeah, probably. Will it have any actual long term impact? No
But really there's like 1000x things worse than this in scheme of things
I'm not saying it's good but it's kind of not really worth putting any effort towards generating outcry
dopadelic@reddit
It's not any the size, it's about the amount of heat energy going in. Just 1C rise in temp can have huge ecological effects
Jewnadian@reddit
No, this is about you not understanding math. To bring it down to the basics, the ocean is fucking huge. Unfathomably huge. The ENTIRE human output for the whole world raise the ocean temp by 1.5C in 100 years. That's every single coal powerplant, train, steel mill, office building and so on that has ever been built or operated. That's what it takes to move the ocean temp. This is a single large building. That's it.
dopadelic@reddit
I don't disagree with you, but I'm just saying that size of the data centers is irrelevant. The correct way to look at it is the amount of energy required to raise the ocean temps and to compare that with the amount of energy consumed by the data centers.
Jewnadian@reddit
Yes, and the math doesn't work there either. It's the same math actually. It takes a TW to heat a km3 by one degree. Take a wild guess how many km3 there are in the ocean?
Sopel97@reddit
To increase temperature of 1km^3 of water by 1C you need roughly 1TWh of energy.
Roxalon_Prime@reddit
Most of excessive heat is already actually swallowed by Ocean and those numbers are massive, it is extremely unlikely for any your data center to have a major impact, but I didn't do the math, sorry
nomad5926@reddit
100% will have a long term impact. And yes there are worse things. Definitely not a huge deal in the short term, long term if more of this haopens it could be a huge problem.
LaM3a@reddit
On the global scale, for sure, but what about locally?
mechkbfan@reddit
Guess relative to tides & depth.
Putting it in relatively stationary harbour thats a few metres deep? Yeah, it'll probably warm things up a little around it. I don't think they'd be that stupid though.
10m deep with tides? Once again, doubt anything meaningful for exception of school of fish that might find it discomforting going near it
Like we have 1,386,000,000 cubic kilometers of water or so on planet.
Even if you built a data enter that was 1km x 1km x 1km, that's 0.00000007215007215007215% of the ocean.
I'm not pro-AI, I couldnt care if it dies tomorrow and never existed, but the heating impact this will have isn't really worth thinking about.
Mrseedr@reddit
Yes there's no way this could possibly come back to haunt us.
raydialseeker@reddit
Significantly better than having it on land. You pick between the better of 2 evils. Saying that "both are evil" without looking at the scale or impact has historically proven to be catastrophic
ea_man@reddit
Data center proposed outside Paris, France, could heat greenhouse and fish farm
Homes plan shelved in favor of data center, residents rejoice (yes, really)
https://www.datacenterdynamics.com/en/news/data-center-proposed-outside-paris-france-could-heat-greenhouse-and-fish-farm/
DrAnklePumps@reddit
There's another option of not building the datacenters at all because AI doesn't have widespread value.
kenyard@reddit
Local beach gonna be lovely and warm
anuanuanu@reddit
Ecosystem disrupted for every lifeform other than humans.
boringestnickname@reddit
A vast majority of oceanic life is coastal.
I'm not thinking about the heat at all. I'm thinking about displacement problems, and activity at the seabed near the coasts (with the added general underwater activity to build and maintain.)
deep_chungus@reddit
not to mention all the other horrible shit we constantly pump into the ocean
avengers93@reddit
Want to know how much the Sun heats the oceans every day, and how much heat would be output by data centres compared to that?
AstralElement@reddit
How would a space data center disperse that massive quantity of heat?
toybuilder@reddit
You glow the heat away.
AstralElement@reddit
It’s incredibly less efficient by a very huge magnitude. It’s barely manageable on spacecraft today, let alone a data center.
toybuilder@reddit
I totally agree with you.
It's literally the only way to get rid of heat in space.
iwrestledarockonce@reddit
How does one cool and maintain a data center in space?
boringestnickname@reddit
They can both be bad ideas.
I'm not saying data centers in space somehow makes sense, on some absolute scale, because I question the relative sense they make to ocean data centers.
I'm just saying, maybe we shouldn't be messing much more with the system that hosts 80% of the animal biomass on the planet, not to mention pretty much all the oxygen production. Especially since the coastal areas are where most life exists.
We're already pretty nervous about sudden collapses.
If I'm going to be forced into becoming a hunter-gatherer in some post-apocalyptic world, I'd prefer it if I could at least catch some fish and have breathable air.
DehydratedButTired@reddit
Anyone who has been on the ocean for long enough knows both suck and just have different problems. Salt water is corrosive and alive. There are microbes they break down metal, oil, diesel and gasoline. Salt corrodes. Seawater is great at penetrating everything. I can’t imagine what the maintenance costs for replacing the cooling and pump hardware will be but probably still cheaper than shooting computers into space.
tkeser@reddit
it depends on the connection, and security. space gets connected via satellites (that's why musk is pushing for it)
edparadox@reddit
No.
It simply is a stupid idea.
kuddlesworth9419@reddit
How would you even cool anything in space? Like you can't dump all that heat anywhere at all considering heat doesn't radiate in a vacuum very well.
VenditatioDelendaEst@reddit
Heat radiates in a vacuum exactly as well as a non-vaccum.
I'm having trouble finding the calculation in my shell history so I can't give you exact numbers easily, but it works out that the required radiator area is comparable (within a factor of 2 either way, I don't recall which way) to the area of the solar panels that capture the power.
These4Walls128@reddit
Uh huh. And how much radiator space we talking? Do the math. It's a regarded move to use space for this.
VenditatioDelendaEst@reddit
I just told you???
These4Walls128@reddit
Those both require a massive amount of rocket launches, it's preposterous to suggest this is a good use of resources. It's just Musk trying to keep his businesses afloat.
VenditatioDelendaEst@reddit
Your reply to my reply to this post was caught in some censor. If it was a convincing argument that anyone whose money isn't on the line should be mad about, feel free to host it elsewhere and link. If it was a convincing argument that you have reason to believe people whose money is on the line haven't heard before, feel requested to host it elsewhere and link.
(But to be clear, my money is not on the line either.)
VenditatioDelendaEst@reddit
That is between Musk, his investors, and the laws of physics. Your kind, godwilling, get no input.
These4Walls128@reddit
[ Removed by Reddit ]
BatteryPoweredFriend@reddit
Radiation is so, so much worse than convection for cooling.
VenditatioDelendaEst@reddit
Indeed, but it's still not as bad as solar cells for power generation, so you should not be especially worried about the cooling.
Roxalon_Prime@reddit
In my opinion the most stupid idea in all of it, that the vacuum of space is actually an insulator, so I don't quite understand what they're thinking, but maybe I'm missing something
VenditatioDelendaEst@reddit
Saltwater chemical environment is also very hostile for computing equipment. However, engineering.
tkeser@reddit
it's not if you're spacex and your data center is already serving your customers via space
1eejit@reddit
Undersea cables are pretty much solved
mpgd@reddit
Have you seen space cables? No, wait until musk invents it.
N1NJ4W4RR10R_@reddit
Who needs cables when you have LASERS
ICC-u@reddit
That's what Russia wants us to think
Whirblewind@reddit
Sounds like you don't know anything about the physics involved. No need to tell us that.
toybuilder@reddit
Lot of people seem to agree with my opinion, though? If you would like to offer information to show me why I am wrong, I am happy to learn more!
Framed-Photo@reddit
Water cooling is great of course, but that's often done with chemically treated water to avoid corrosion even in small PC builds, and even THEN you can still get build up of sediment and corrosion if you're not careful.
With salty ocean water? All I can say is good luck lol. I don't know how they're doing this, but there's a reason why this hasn't been done in the past.
Intelligent_Top_328@reddit
For now. We need so much that we will run out of room. We need more datacenters. They are the future.
FuturePastNow@reddit
oh yeah cut out the middle-men and dump that heat directly into the ocean
auto_named@reddit
Space datacenters make a lot more sense actually. If they start building them in the ocean, then eventually the ocean will just be full of them. There’s a lot more room in space.
Hybrii-D@reddit
In fact, no.
ShrimpCrackers@reddit
Yes. That said Ocean datacenter sometimes doesn't make more sense than land one. The reasoning is datacenters need human maintenance and hate water and humidity.
MarcoVinicius@reddit
Sure, just cook the oceans and pollute them even more. The space ones sounded like sci-fi but this sounds like a disaster.
stikves@reddit
Probably not...
The pressure differential itself will make a huge difference. Space vs land is 1 atm, Deep sea vs land is easily 10s of atm.
Power in space is essentially free, cooling as well (just look at ISS). Deep sea? More problematic
Maintenance. This is one area sea has benefits. But... these modules can be considered "disposable" (Look at the size of that "data center"). It is going to cost SpaceX about 10 million to launch one, and that is peanuts in the total cost.
The reason people are going for these crazy ideas, and not just using barren land on earth... is our unrealistic pushback for growth. But, I would expect China would also be in the "space datacenter" race if they actually had the lift capacity to do so.
(They can't, their toy space station is not even fully operational yet)
WealthyMarmot@reddit
uh is this some creative writing exercise set in a bizarro universe? Power in space is one of the primary limiting factors for damn near everything we want to do up there, and cooling is a nightmare - vacuum is an incredible insulator.
Orolol@reddit
Haha, no, not at all.
Exist50@reddit
You wouldn't be putting these in "deep sea". You're talking more like 4-5 atm for the places these would be deployed in.
You're joking, right? Both power and cooling are immense problems, even for the ISS, which demands a fraction of what a datacenter would.
NoHetro@reddit
Yeah i feel like way more places make better sense before we jump to space, like why not antartica or greenland?
wondersnickers@reddit
I think the primary goal with this is to be out of any jurisdiction, isn't it?
WTFAnimations@reddit
China out here using the sea as a huge heatsink, meanwhile American data center companies still insist on draining millions of gallons of freshwater despite closed-loop cooling now being fully possible.
luvsads@reddit
American companies experimented with this a decade ago.
Kavani18@reddit
But that goes against the “China really good and America really bad” narrative that Reddit loves to push. We only glaze authoritarian regimes if they are Eastern.
R-ten-K@reddit
LMAO.
Kavani18@reddit
It is pretty funny, isn’t it? My favorite is when people say the Chinese government are the “adults in the room” when in reality they throw a fit every time someone says anything that hurts Pooh’s feefees
R-ten-K@reddit
I was laughing at you, not with you.
Kavani18@reddit
I know, dunce. That’s why I continued saying what I said. Anyway I don’t see what was so funny about what I said. It’s true. In every political sub is non stop glazing China. But let’s ignore China’s terrible record on human rights, genocide, government corruption, speech rights, and fairness. Let’s pretend that China is perfect fairytale land like Redditors love to do
R-ten-K@reddit
What can I say? Some of us find extreme projection and zero self awareness hilarious. So please, do continue. LOL
Kavani18@reddit
Projection? That a new word you learned? Buzz words don’t have meaning when you don’t know how to use them. I’m not sure exactly how I’m projecting as I never once said anything about how my country conducts itself at this moment. Please do pick up a dictionary. Have a nice day
Odd_Cauliflower_8004@reddit
because thermal pollution is not a thing... 24megawatts.. dumped in the ocean in a very concentrated area.. absolutely going to have no effect oon the marine flora and fauna of the area, i'm sure..
R-ten-K@reddit
24MW is a tiny amount given the heat capacity of the ocean.
Even if we were to submerge all the data centers in the world, it wouldn't raise the temperature of the ocean in any significant way.
StickiStickman@reddit
Yes, this but unironicly. It won't have any noticeable effect unless you're immediately next to it.
Odd_Cauliflower_8004@reddit
What about sealife?
Qweasdy@reddit
24 megawatts is less than half of the heat output of just a single large ship.
UmaThurmish@reddit
you can easily spot someone that isn't critically thinking
the 24 megawatts of heat is going to happen regardless of where, however, if you don't cool it in the ocean, you have to spend even more energy to get rid of that heat using fans, or pumps with liquid cooling
using the ocean as a heatsink reduces how much energy you will overall spend
in the end obviously I think these datacenter projects are a waste of resources and will just make climate change worse and worse.
hackenclaw@reddit
depends where you put it, if you put it deeper part of the sea, it might acually spawn life due to the heat.
Check out what Volcanic created under sea.
Reporting4Booty@reddit
Not really. Geothermal vents are usually found at the deep sea floor and the reason they are useful is because they provide a lot of minerals for chemosynthetic microbes to feed on, which become the basis of an ecosystem.
For ocean water closer to the surface, the bottleneck is phytoplankton, i.e. sunlight. If anything, excess heat at the surface is detrimental because it interferes with natural convection, preventing the flow of nutrients. Maybe if they put it closer to 200m depth it wouldn't be as bad or could actually stimulate nutrient flow, but I'm not a scientist.
YSoMadTov@reddit
Don’t have to worry about local marine and fauna eco system if you had already destroyed it long ago throught unchecked trawler fishing and pollution. *tap head*
Seriously thought the seas around China are so thoroughly depleted that Chinese government sponsored pirate….err fishing fleet have to violate the eez of other countries to fish there, they’ve been found as far as South America.
EeveelutionsFucker@reddit
Have you know how cold is deep ocean is. We are not talking about 2 meter high water.
anders_hansson@reddit
Swedish nuclear reactors use seawater cooling, and it definitely has an effect on the local ecosystem.
EeveelutionsFucker@reddit
Yeah but dude they are also not stupid. Im pretty sure they have engineers or geologist that survey the area first that has the least impact. Rather than consuming millions of tons water with current datacenter
1eejit@reddit
Yes but capitalism doesn't care about the environment more than governments force it to. Engineers might be able to solve the problem, but only if the suits budget them to do so.
Seref15@reddit
Solvable problem though maybe not as a submerged datacenter.
US nuclear reactor Turkey Point in Florida uses an open-air canal system as a closed-loop heat sink. Water comes out ~10F hotter than it went in, then it slowly travels through miles of canals to cool. They only open it to the sea when they need to raise the water level.
jin85@reddit
Found the, “but at what cost?” comment
Simp_Simpsaton@reddit
And building them in areas prone to drought xd
ea_man@reddit
There was a guy the other day that wanted to build a datacenter the size of 2x Manhattan powered by gas! In the desert! He's gonna power + cool it by gas!
here: https://www.theguardian.com/us-news/2026/may/13/utah-approves-datacenter-backlash
LightShadow@reddit
Live here and nobody is thrilled. Gov. Cox on TV claiming the data center will provide more water than it uses; make it make sense.
ea_man@reddit
That's Jensen motto: the more you spend the more you save!
SmileyBMM@reddit
That actually could work if it's using nonpotable water and distilling it...
ireadoldpost@reddit
AI generated images of water.
SmileyBMM@reddit
https://en.wikipedia.org/wiki/Project_Natick
Marty_McFuckinFly@reddit
Shut up China bot
PhaedrusNS2@reddit
How does an under water data center make sense? Wouldn't that be far more expensive than a traditional data center?
ju2au@reddit
The biggest costs in a data centre are energy and water.
By placing it in a position that is very windy, you get free energy from wind turbines.
By submerging it in water, you can free cooling from all the water around it. Obviously, not via direct contact but the data centre probably has a close loop water cooling system where the heated water is cooled in a heat exchange that has contact with the outside water.
StickiStickman@reddit
That's just not true. It's the hardware by a wide margin.
T-MoseWestside@reddit
Upfront, yes. But over it's entire operational time?
R-ten-K@reddit
FWIW
The cost breakdown for data centers has stayed remarkably stable for decades: roughly half goes to hardware-related costs, basically: servers, racks, networking, components, etc. The other half goes to infrastructure and operating costs, that basically includes: power, cooling, connectivity, labor, facility expenses, etc.
That said, even small reductions in power or cooling costs can create a meaningful economic advantage for a data center operator, especially at scale. Anything that lowers those recurring costs improves margins if the operator can actually leverage it.
VenditatioDelendaEst@reddit
I do not think this is true. The water stuff is pure propaganda, and previous analyses have found the majority of the cost is the computers. It's a bare majority, so I can imagine energy might become the largest share if you're spending money to save time (with gas turbines).
R-ten-K@reddit
I think you meant a "plurality," 45% is not a majority.
Anyhow. The assumption regarding DC costs has always been that roughly half goes to compute HW costs, and the other half goes to the infrastructure/support costs (electricity, AC, telco, etc). As your study reinforces.
DaBombDiggidy@reddit
Yeah once you throw that crap under the ocean water gets dwarfed by exponential maintenance costs.
We’re going to have abandoned data centers in the ocean in no time.
hackenclaw@reddit
yep a giant anti corrosive "heatsink" that expose to sea water pretty much can dissipaate real quick for free.
Sh1rvallah@reddit
Isn't that what they use to cool the reactors on nuclear vessels
R-ten-K@reddit
I mean, they are trying to answer that very question. Ergo the experimental test they are conducting.
RandoScando@reddit
Short term, yes. Long term, no. The U.S. is currently draining local water supplies in an unsustainable way. It’s also adding a massive draw on the power grid that is similarly unsustainable. This model can’t scale. If you can cool it with the ocean being the heat sink, and can power it with local wind, you’ve solved the biggest problems of scaling.
Now whether or not this ends up actually working as advertised on the box is another matter.
Whirblewind@reddit
Proof? And not just one - show me proof of the plurality of this claim, which you did with "supplies". Thanks in advance.
viperabyss@reddit
Long term is likely yes too. Putting metal pressure vessel in the ocean would cause corrosion, and it’s very difficult to service the server without pulling the entire datacenter up.
anders_hansson@reddit
Depends on the scale/design of each underwater module. You could have a service entrance avove water and enough space in the module for a human to move around.
ICC-u@reddit
The cheapest way would probably to have some level of expected failure rate, rather than trying to repair equipment that's underwater. Perhaps there's a level of redundancy with extra racks and drives that can come online if others go down.
anders_hansson@reddit
Agree. For pure compute farms it's even simpler. If a node within a module goes down you simply lose a fee percent of performance. Once a module's performance drops below a certain level you can bring it up to surface and either reuse the good parts for new modules or replace the bad parts with new parts.
Moscato359@reddit
They simply just need to design it with internal redundancies
And accept if a machine dies you leave it dead
surg3on@reddit
And outdated machines ? They don't exactly have a super long life.
0xe1e10d68@reddit
Replace the whole module
DameLasNalgas@reddit
It probably can be pulled up by design
hackenclaw@reddit
it does not limited to wind power only, they can put a flat giant solar farm or even ocean waves to generate power, and it does not require clearing out forest since it doesnt require land at all.
smokefoot8@reddit
It still makes far more sense than an orbital one like Musk keeps talking about.
UmaThurmish@reddit
nah, it saves a lot of energy, it costs energy to cool things. and heat goes into environment anyway. when you put it in the ocean you aren't spending energy to move energy
Canadian_Border_Czar@reddit
Air is a poor conductor of heat, and it doesnt take much energy to heat up a gram of air 1 degree.
Water on the other end is a good conductor of heat (not the best, but better than air) and has a very a higher heat capacity, and is abundant in a place where we dont need it (at least in the sense of immediate survival)
If you took the same surface area of heat transfer area on a closed loop fluid cooler, used to cool a condensing loop (for a water source heat pump), and stuck it in the ocean.. you'd probably maintain loop temps at or around the temperature of the ocean, and you would only need one pump. No heat pump. I also didnt do the math
So yeah, designing something water proof, rust proof, sway proof, etc. Would be very expensive, but running it would provide significant cost savings.
Dramatic_Investing@reddit
Well said. But comparing this to the cost of Data centre in space probable also makes this look cheap. I also didn't do the math though
Canadian_Border_Czar@reddit
Imo the idea of a data center in space is frankly silly at least with our current technology. Vacuum is an excellent insulator. Like if it wasnt so expensive, and hanging a picture wouldnt be a very expensive mistake, everyone's walls would be vacuum instead of fiberglass insulation.
A data center in space youre going to need to build like a geothermal field on the moon, or otherwise find a use for the heat.
StickiStickman@reddit
You're vastly overstating how hard it is to get rid of the energ.
It is in fact pretty easy with a relatively small amount of radiators. We're talking about the same area as the solar panels.
Canadian_Border_Czar@reddit
Jfc. Okay prove me wrong, do the math.
StickiStickman@reddit
Using a heat pump and radiator:
Area = Power emitted / (emissivity x T^4 x 5.67x10^-8)
1kW = 1000 / (0.98 x 473^4 x 5.67*10^-8) = 0.3m²
PoL0@reddit
when you say "expensive" you factor in long term operational costs? not only money but energy and water.
PoL0@reddit
when you say "expensive" you factor in long term operational costs? not only money but energy and water.
RuleIV@reddit
Lots of potential advantages. Whether it's overall economically viable or advantageous over traditional datacenters has yet to be proven. On top of the savings from free cooling, studies have shown substantial hardware reliability benefits from strictly controlled environments (inert gas, low vibration, rock steady temperature, etc)
anders_hansson@reddit
I get a feeling that Chinese policies, unlike western, favor long term profitability over instant returns. In the near future it's almost certainly more expensive, but a couple of decades from now when the technology has matured and scaled it's probably way cheaper compared to using nuclear/gas/... powered AND cooled data centers.
Even if you use wind/solar to power a land based data center you are wasting about 50% of the power on cooling, which you get almost for free in an under water data center.
Marty_McFuckinFly@reddit
Bad China bot
hollow_bridge@reddit
It might not, this isn't the first under-water data center experiment; it's just the first one that's wind powered.
https://en.wikipedia.org/wiki/Project_Natick
TotalWarspammer@reddit
How does a location with unlimited power and cooling make sense for a data centre? Well...
gomurifle@reddit
An American company was trying this idea. What happened?
Glittering_Let2816@reddit
Not just some company...IIRC, it was Google themselves.
jeff3rd@reddit
to add to this, they took it back in 2020 to retrieve data and afterthought of the project although they've confirmed that it's been terminated in 2024, not sure whatever came out of it.
tildenpark@reddit
IIRC, someone tried it in the early 2000s. Salt water for cooling had issues and maintenance was too expensive.
fozziebox@reddit
Microsoft trialed this in Scotland a few years back, nothing came of it though
Appaamma123@reddit
What about rusting issues, wouldn't the maintenance upkeep be very costly
RedditNotFreeSpeech@reddit
ectomobile@reddit
24 megawatt ain’t cutting it.
namotous@reddit
Passive cooling makes sense but it’d be a pain for maintenance
starlightserenade44@reddit
...is it gonma release huge amounts of heat into the ocean...?
VenditatioDelendaEst@reddit
Unless you think moving heat from the wind into the ocean changes the Earth's albedo, this "problem" (such as it even exists, which is doubtful) is entirely local.
bhop_monsterjam@reddit
what better way to heat the oceans up by dunking them in directly?
Exist50@reddit
The biggest source of net heating is the GHG emissions.
SpinDryCycle@reddit
There was a successful feasibility experiment by Microsoft for this type of undersea data centre back in 2020 in the UK. So not quite a world first... link
Dramatic_Investing@reddit
Oh this is super interesting!
TophxSmash@reddit
I think its just the wind powered part thats a world first.
kai_ekael@reddit
No filling way I would work there. Damn you Spielberg!
DameLasNalgas@reddit
Put them all on the moon
Smallmyfunger@reddit
Which side?
DameLasNalgas@reddit
Dark side ofc
AutoModerator@reddit
Hello sr_local! Please double check that this submission is original reporting and is not an unverified rumor or repost that does not rise to the standards of /r/hardware. If this link is reporting on the work of another site/source or is an unverified rumor, please delete this submission. If this warning is in error, please report this comment and we will remove it.
I am a bot, and this action was performed automatically. Please contact the moderators of this subreddit if you have any questions or concerns.