Since 2020, Meta’s hyperscale data center — spanning 50,000 square meters on an industrial estate on the edge of the city — has been pushing warm air generated by its servers into the district heating network under Odense. That heat is then dispersed through 100,000 households hooked up to the system, with Meta providing enough heat to cover roughly 11,000.
Heat pumps are great, but what this guy is saying is wrong. Generating heat in the thing you’re trying to cool won’t help save any money no matter the technology.
Let’s say you were deliberately trying to heat something and cool something else, like a water heater and your home. Then heat pumps are doubly effective. Maybe that’s where the confusion in this comment stems from, but that’s not what’s going on with a data center.
I should have been more specific, the savings would be from compartmentalizing the hot room from the rest of the HVAC system.
If you partition the lab from the rest of the hvac system and install a heat pump for the lab, it would drastically reduce the amount of heated air the HVAC system has to cool and move around the rest of the home.
I suspect the reasoning here is going to not be obvious so some people so I’ll add a little. Heat pumps are more efficient when there’s a larger difference between where they’re getting heat and where they’re putting it. I’m going to call this difference a gradient, because otherwise later I’ll be saying “differences of differences” which gets confusing. The argument here is that moving heat from compartmentalized 90 degrees server room to outside at 95 degrees with a separate system for the house, moving from 75 degrees to 95 degrees, would be more efficient than a single heat pump moving air from the mixed rooms at 80 degrees to 95 outside.
The magnitude of that effect would depend on how nonlinear the relationship between efficiency and the gradient is. I’m not very familiar with that. I assume it’s nonlinear, but whether it’s highly nonlinear from a gradient of 5 to a gradient of 20. From here, it’s quite nonlinear from a gradient of 25 to 40, but from 5 to 20 it’s pretty linear: https://www.ecosia.org/images?q=heat pump effeciecy vs temperature gradient#id=598B80C1EB5A721C392964CB7708512FC496B78F
This also doesn’t consider that these are operated with thermostats. Presumably someone is going to set all of the thermostats to the same temperature, 75 degrees or whatever the preference is. The gradient at which the pumps start will be the same in all cases, and the difference will be in how often the pumps run. There will differences in the average efficiency because of the time difference, but it’s by no means obvious to me that there would be a significant benefit for a typical home. I would want some clear evidence before spending money on this.
Edit: I said they’re more efficient when there’s a larger gradient, but I should have said the efficiency depends on the size and direction of the gradient. When the gradient is positive, it’s less effecient. Overall, the conclusion is the same. It’s dubious that, for this case, using two heat pumps with compartmentalized rooms provides a tangible benefit over a single pump with mixing.
So you’re saying to just let the servers get as hot as they want and ignore it? What is this heat pump going to be heating?
If you don’t have anything you are wanting to heat up, adding a heat pump doesn’t help.
Modern heat pump systems are reversible, so you would be drawing the heat energy from the room and releasing it outside. During this exchange air is run over a compressed coolant which cools the air and reintroduces it back to the room.
That’s just called an air conditioner?
An AC can only cool the home, while a heat pump can both cool and heat a home. They are different things, and worth looking into.
That’s just an air conditioner. The difference between an air conditoner and a heat pump is that a heat pump has a reversal valve that allows the refrigerant to run backwards through the system.
My point is someone is using a whole bunch of words to say “I’d air-condition my server room”. It’s not exactly a ground breaking idea.
A heat pump still costs energy to run… And running it in reverse to cool a room is the exact same process as running an ac to cool a room. It’s all phase change cooling which we’ve been doing for decades. The only possible innovation that could be had here is using the waste heat instead of just releasing it to the outside environment, but that requires wanting to heat something else while you need to cool the server room.