Startup Uses SpaceX Tech to Cool Data Centers With Less Power and No Water (yahoo.com) 56
California-based Karman Industries "says it has developed a cooling system that uses SpaceX rocket engine technology to rein in the environmental impact of data centers," reports the Los Angeles Times, "chilling them with less space, less power and no water."
Karman has developed a cooling system similar to the heat pumps in the average home, except its pumps use liquid carbon dioxide as refrigerant, which is circulated using rocket engine technology rather than fans. The company's efficient pumps can reduce the space required for data center cooling equipment by 80%.
Over the years, data centers have used fans and air conditioning to blow cold air on the chips. Bigger facilities pass cold liquid through tubes near the chips to absorb the heat. This hot liquid is sent outside to a cooling yard, where sprawling networks of pipes use as much water as a city of 50,000 people to remove the heat. A 50 megawatt data center also uses enough electricity to power a mid-sized city... Cooling systems account for up to 40% of a data center's power consumption and an average midsized data center consumes more than 35,000 gallons of water per day...
U.S. data centers will consume about 8% of all electricity in the country by 2030, according to the International Energy Agency... The cooling systems are projected to use up to 33 billion gallons of water by 2028 per year... To serve this seemingly insatiable market, Karman has developed a rotating compressor that spins at 30,000 revolutions per minute — nearly 10 times faster than traditional compressors — to move heat...
About a third of Karman's 23-person team came from SpaceX or Rocket Lab, and they co-opted technologies from aerospace engineering and electric vehicles to design the mechanics for the high-speed motors. The system uses a special type of carbon dioxide under high pressure to transfer heat from the data center to the outside air. Depending on the conditions, it can do the same amount of cooling using less than half the energy. Karman's heat pump can either reject heat to air, or route it into extra cooling, or even power generation.
The company "recently raised $20 million," according to the article, "and expects to start building its first compressors in Long Beach later this year...."
Over the years, data centers have used fans and air conditioning to blow cold air on the chips. Bigger facilities pass cold liquid through tubes near the chips to absorb the heat. This hot liquid is sent outside to a cooling yard, where sprawling networks of pipes use as much water as a city of 50,000 people to remove the heat. A 50 megawatt data center also uses enough electricity to power a mid-sized city... Cooling systems account for up to 40% of a data center's power consumption and an average midsized data center consumes more than 35,000 gallons of water per day...
U.S. data centers will consume about 8% of all electricity in the country by 2030, according to the International Energy Agency... The cooling systems are projected to use up to 33 billion gallons of water by 2028 per year... To serve this seemingly insatiable market, Karman has developed a rotating compressor that spins at 30,000 revolutions per minute — nearly 10 times faster than traditional compressors — to move heat...
About a third of Karman's 23-person team came from SpaceX or Rocket Lab, and they co-opted technologies from aerospace engineering and electric vehicles to design the mechanics for the high-speed motors. The system uses a special type of carbon dioxide under high pressure to transfer heat from the data center to the outside air. Depending on the conditions, it can do the same amount of cooling using less than half the energy. Karman's heat pump can either reject heat to air, or route it into extra cooling, or even power generation.
The company "recently raised $20 million," according to the article, "and expects to start building its first compressors in Long Beach later this year...."
Predictions-schmedictions... (Score:4, Interesting)
U.S. data centers will consume about 8% of all electricity in the country by 2030, according to the International Energy Agency...
Is the bubble going to last that long? I have some doubts.
Re: Predictions-schmedictions... (Score:2)
Lord, you went off the rails in your second sentence!
Why can't you stick to the topic, cooling data centers? This has nothing to do with vibe coding, the stock market, or the Fed...
A special kind of carbon dioxide (Score:3)
Re: (Score:1)
Re: A special kind of carbon dioxide (Score:5, Informative)
I guess it could be supercritical, which is just over 1000 psi, so not extremely high pressure but you wouldn't want to be near a leak.
That doesn't make a lot of sense thermodynamically, though, because it's a phase change that does the bulk of the energy transfer in a heat pump cycle. So you would need to cross that boundary line into a two-phase region.
You cannot exceed carnot efficiency, no matter what. There are no tricks. Draw your box around the system doing the work (physics work) and it doesn't matter what magic they claim, it cannot exceed carnot efficiency.
Re: (Score:2)
"carnot efficiency"? If R290(propane) can get a COP of 4x at 70% effectivity, what are you selling?
Re: (Score:2)
If you don't know what carnot efficiency is, and the second law of thermodynamics, the question is what are YOU selling?
The claim of reducing energy usage by "80%" must be vetted against the limits of thermodynamics. That's a 5x reduction in work, so a 5x increase in your COP. Doesn't pass the sniff test.
using water in a closed loop system is irrelevant (Score:4, Interesting)
It doesn't matter if you use water to cool in a closed loop system. The problem is only if you're using fresh water and just heating it up and dumping it and pulling in more fresh water.
So in this case, it's completely irrelevant.
Re: (Score:2)
Re: (Score:3)
The radiators get awful big at MWs of power and only 25 or so degrees C relative to ambient. So you spray water on them and they work a lot better.
Re: (Score:2)
Or rather you just spray the hot water into the air, I forgot how they worked.
Re:using water in a closed loop system is irreleva (Score:5, Interesting)
At the top of the tower (hot side) are sprinklers, at the bottom is a pool (cold side).
I.e., it's a closed loop, with biocide and lost water injection into the loop.
Re: (Score:2)
Closed loop minus what evaporates.
Re: (Score:2)
Like... mostly closed.
Re:using water in a closed loop system is irreleva (Score:5, Informative)
Actually it's a combination of both. Most cooling systems have a closed loop and an evaporative one with heat exchangers involved moving heat from one to another. The reason being you want to carefully control the chemistry of the liquid required to get heat from a concentrated area (such as the fins on a waterblock on your CPUs), but don't want the expense of blasting that expensive liquid into the air.
Re: (Score:2)
Re: (Score:2)
they use the heat of evaporation of water to cool the hot refrigerant and release that to the atmosphere, because it's more efficient than just air cooling.
it's somewhat like setting up a small water sprayer on your AC condenser outside. (they actually sell kits... and iirc they're cost effective, though I've never tried one).
Re: (Score:3)
We have cooling towers that directly evaporatively cool the loop (and replace lost water, inject biocides, etc)
Re: (Score:3)
Directly? I find that hard to believe. Are you sure you're not running two loops with a heat-exchanger between? That would be how nearly every other datacentre in the world works regardless if the actual cooling element is liquid or air via a chiller. My mind would be blown if the actual liquid hitting your servers isn't running in a closed loop as you would expose yourself to a world of hurt with water chemistry having to have that continuously replaced and bio controlled. It's hard enough to prevent growt
Re:using water in a closed loop system is irreleva (Score:5, Interesting)
Directly? I find that hard to believe.
Alright.
Are you sure you're not running two loops with a heat-exchanger between?
Yes.
That would be how nearly every other datacentre in the world works regardless if the actual cooling element is liquid or air via a chiller.
Incorrect. Though many very large systems do for obvious reasons. The larger the loop, the more work you need, and not to mentioned a generally larger pumping requirement. There's also the limitations that you had better hope you never have high humidity, or freezing temps. Not a problem in our area.
My mind would be blown if the actual liquid hitting your servers isn't running in a closed loop
You mean near our servers, I imagine... I was definitely not trying to imply that we're pumping this shit into water blocks on our CPUs.
as you would expose yourself to a world of hurt with water chemistry having to have that continuously replaced and bio controlled.
We had problems at first. Once we got the chemical and biological problems sorted, it hasn't been a problem in years. The maintenance regime is strict, though.
It's hard enough to prevent growth on the open loop which is far less sensitive to corrosion and biological growth.
I wouldn't say any part of the system is more or less sensitive to the corrosion and biological growth. Biocides and water conditioning are a must.
Re: (Score:2)
You mean near our servers, I imagine... I was definitely not trying to imply that we're pumping this shit into water blocks on our CPUs.
Gotchya. You are already talking about an outter loop where the closed loop internally would be an air circulation. I thought you were talking about a liquid cooled servers which most definitely have a closed loop exchanging heat with the chilling system.
I wouldn't say any part of the system is more or less sensitive to the corrosion and biological growth. Biocides and water conditioning are a must.
You misunderstand, when I was talking about a closed liquid inner loop (increasingly common in dense datacentres) there's orders of magnitude differences in sensitivity. It's much easier to block a channel that is 0.3mm wide than a tube that is 1" which lea
Re: (Score:2)
Gotchya. You are already talking about an outter loop where the closed loop internally would be an air circulation. I thought you were talking about a liquid cooled servers which most definitely have a closed loop exchanging heat with the chilling system.
You could call the air circulation an inner closed loop if you wanted, but that would be very nonstandard nomenclature.
You misunderstand, when I was talking about a closed liquid inner loop (increasingly common in dense datacentres) there's orders of magnitude differences in sensitivity. It's much easier to block a channel that is 0.3mm wide than a tube that is 1" which leads to vastly different water quality requirements for closed liquid loops vs the external chiller loops. You simply don't liquid cool the chips directly with an open loop due to this reason.
Ya, I wouldn't even imagine trying to directly cool electronics components like that, lol
Re: (Score:2)
Yeah it's plumbing meets tech. Quite interesting stuff. Here's an example from Amazon https://www.aboutamazon.com/ne... [aboutamazon.com]
I've seen a small datacentre do this too, but to bad effect. They were so focused on water cooling they forgot about the things that weren't water cooled. The datacentre was a sauna after startup, and they had a high rate of SSD failure before they realised you still need an AC unit in side because the water doesn't touch all heat generating components.
Re: using water in a closed loop system is irrelev (Score:2)
Many, many cooling systems use direct evaporation. Especially, but not only, in drier environments.
There are maintenance concerns, but the operating and initial capital costs are lower for those systems, so they are attractive on that level.
Re: (Score:2)
I was under the impression the GP was talking about liquid cooled datacentres, not the liquid loop of the chiller. Yes those lines in the outer chiller loop are usually evaporative cooling or open loops via some natural heatsink (stream). But in situations where datacentres are directly liquid cooled (i.e. coolant is pumped through CPU / GPU blocks in high density environments) those are universally closed loop systems with a heat-exchanger to the evaporative loop / chiller loop.
Re: using water in a closed loop system is irrele (Score:2)
I haven't designed for data centers specifically, but in every other cooling application that exists direct liquid exchange is not uncommon.
Re: (Score:2)
It doesn't matter if you use water to cool in a closed loop system.
The closed part of the closed loop system only moves heat from one area to another. There's a whole different part of the system involved in actually moving that heat out to the environment, that system is open. The water consumption being discussed is usually due to evaporative cooling - which is still the cheapest way to run a cooling system unless you have a specific environment that allows for some other alternative (e.g. dumping heat into a river).
Re: using water in a closed loop system is irrelev (Score:2)
Exactly.
Bigger facilities pass cold liquid through tubes near the chips to absorb the heat. This hot liquid is sent outside to a cooling yard, where sprawling networks of pipes use as much water as a city of 50,000 people to remove the heat.
And from the cooling yard the water goes where? Oh, yeah, back into the cooling system. The water is not "used," it is "re-used" - yes, some evaporates, but not 50,000 people worth of water (BTW, how much water does "one people" use?)
Re: (Score:2)
TFA doesn't explain anything about the system and seems to have a lot of misunderstanding. For example: 30,000 RPM is not unusual for centrifugal chi
Re: (Score:3)
How does that work? (Score:3)
The system uses a special type of carbon dioxide...
As far as I can tell, the only way CO2 can be "special" is if it's in a liquid state. (Normally, it transitions between solid and gas with no intermediate liquid state, but it does become liquid at about a little over 5 bars). But liquid CO2 is commonly used for preserving food and in other industries, so I'm confused as to what might be special about the stuff they plan to use for cooling data centres.
Re:How does that work? (Score:5, Funny)
But liquid CO2 is commonly used for preserving food and in other industries, so I'm confused as to what might be special about the stuff they plan to use for cooling data centres.
What's special is they're using a high-tech emulsion - liquid CO2 blended with equal parts Elonium and Baloneum.
Re: (Score:2)
I saw that in a motion picture once - Return of the Jedi.
'Carbonite' was used as a liquid solution for cryostasis.
Re: (Score:2)
Some gasses work better than others for given temperatures in heat-pumps. CO2 is best suited to very high temperatures, like increasing the temp from say 50C coming from the GPU to say 90C.
It's sometimes used for domestic hot water heat pumps, mainly in Japan because they do love their baths and hotter water means it can use a smaller tank to fill the tub.
Re: (Score:2)
I think it's just an article written by a generalist so I think he means to say "special refrigerant"
as in different from what's used in normal HVAC and is probably unfamiliar that, as you say, it's existing technology
e.g. https://www.scmfrigo.com/en/bl... [scmfrigo.com]
the only new thing mentioned is maybe using some snake faster compressors... but it's not mentioned of those are more efficient somehow.
Re: (Score:2)
Yeah, but the quoted 30,000 RPM is not that high of a speed for centrifugal compressors using standard refrigerants.
TFA doesn't really say much, but I'm guessing that it's some system that runs CO2 at high pressures and temperatures to cool the CPUs directly, and reject the heat to the outside air without a mechanical refrigeration cycle to raise the heat rejection temperature, since you're already running the CPUs at above-ambient t
Re: How does that work? (Score:2)
None of it is new. They just want to say "we are former spacex guys using spacex technology, give us money."
Its just like all the former Google or (now) OpenAI people - put that on your resume and convince people to give you their money.
Re: (Score:3)
"Special" here means supercritical, because apparently the editors at the LA Times don't trust their readers with words of more than four syllables.
Apart from that, I don't know. I've seen press releases disguised as news before but this one really takes the biscuit. I think they're plugging some kind of pump, because engineers are always coming up with ways to use supercritical CO2 for seemingly random things (I read an article about thirty years ago for using it in industrial dishwashers) and the idea of
Re: How does that work? (Score:2)
Can't exceed carnot efficiency, no matter what sleight of hand they use. I doubt the energy reduction touted is even possible.
Re: How does that work? (Score:2)
Re: (Score:2)
Posting here because it's currently the last comment on my screen: Thanks to all for informative and/or funny answers. It's always good to learn and to see things from other perspectives. I hadn't heard about supercritical CO2, and hadn't considered direct-to-air cooling of the coils versus water-cooling them. Also didn't know that it was used in hot-water heat pumps
I also learned about two new elements, Elonium and Baloneum. I'm guessing they're one above the other in the periodic table... ;-)
How is this supposed to help? (Score:2)
When it's hot outside, you could probably get the radiator temp difference to ambient around 2x higher with the heatpump at good COP. If radiator cooling requires too much radiator volume with normal liquid cooling, 2x is unlikely to make the difference.
Cooling towers with evaporation are used because of the far higher power density they allow.
Re: (Score:2)
Cooling towers with evaporation are used because of the far higher power density they allow.
Yup. This is why we swapped out our traditional refrigerant HVACs. Denser, and cheaper. Though part of that reduced cost is the fact that we get too good of a price on the water we use.
Re: (Score:2)
part of that reduced cost is the fact that we get too good of a price on the water we use.
In Loudoun County, VA, where all the important data centers are (us-east-1, et. al), there's a municipal water treatment plant that serves recycled "grey" water to the data centers.
Who uses fans to circulate liquid anything? (Score:3)
"its pumps use liquid carbon dioxide as refrigerant, which is circulated using rocket engine technology rather than fans."
So these guys thought that a common way to circulate liquid CO2 is using fans?!? Or was it just the AI that wrote the startup's business plan, or perhaps just this announcement, its summary?
Re: Who uses fans to circulate liquid anything? (Score:2)
And since when is a turbopump âoeSpaceX technologyâ, as if no one had done it before.
Re: Who uses fans to circulate liquid anything? (Score:2)
Hey, growing up my morning breakfast used NASA Technology to create a healthy powdered morning drink mix that reduced the number of oranges required to fill my juice glass!
Re: (Score:1)
I bet there's some misunderstanding on the reporter/marketer's side, whether deliberate or not, to make this seem cooler (heh) than it is. Conventional refrigeration systems already exist that use carbon dioxide, though they seem to be more focused on walk-in coolers and such rather than air conditioners. The only technical challenge to using carbon dioxide as a refrigerant is that the whole system runs at a much higher pressure than those using more conventional fluorocarbon refrigerants. That means the
Re: (Score:2)
You're assuming this was written by "guys", whereas it reads like a poorly AI generated summary.
Summary convinced me not to read the article (Score:2)
Karman has developed a cooling system similar to the heat pumps in the average home, except its pumps use liquid carbon dioxide as refrigerant, which is circulated using rocket engine technology rather than fans.
Heat pumps don't circulate refrigerant with fans. Get this fucking bullshit the fuck out of here.
Re: (Score:2)
That post had me wondering if my home heat pump uses impellers or just relies on the action of the compressor's pump for coolant circulation.
Based on the cold snap we're having in Virginia, the noise our heat pump makes when reversing to defrost the outdoor unit makes me think it's just using the compressor's pump and not an impeller or "fan."
Thanks for that insight.
Wow.. (Score:2)
"no water" sure, ok, maybe good....
"liquid carbon dioxide" You get how that's worse right?
Maybe there is something valuable here, but it does a pretty bad job of conveying what that would be.
Liquid Carbon Dioxide (Score:1)
What struck me about the article is they use Liquid Carbon Dioxide as the refrigerant. Carbon Dioxide has long been a possibility for a refrigerant. It's cheap, non-toxic and except for the high pressures needed is almost ideal. But, for reasons I don't understand, it's not used much because it has a tendency to turn from liquid to dry ice instead of turning into a gas, thus clogging up the plumbing. I wonder how they have solved this problem?
Heat pumps are air conditioners (Score:2)
In other news, heat pumps are air conditioners.
News at 11.