A data centre in Denmark routes water from the municipal heating network through its cooling system, then uses heat pumps to lift the temperature high enough for household radiators — turning a by-product into a supply for around 11,000 homes.
Key takeaways
- A data centre in Denmark is reported to send recovered heat into a district heating network serving roughly 11,000 homes rather than discarding it into the atmosphere.
- Almost all the electricity a data centre consumes ends up as low-grade heat, which is normally rejected outdoors by chillers, dry coolers or evaporative systems.
- The recovered heat is too cool for radiators on its own, so heat pumps raise it to a usable supply temperature before it reaches households.
- The arrangement depends on a district heating network being physically close to the site, which is common in Denmark and Northern Europe but rare in most other markets.
- Heat reuse reduces the waste from computing but does not reduce the electricity a data centre consumes, and the specific efficiency figures for this site are not independently verified here.
What is actually happening at the site
The described set-up inverts the usual relationship between a data centre and its surroundings. Instead of treating heat as a problem to be expelled, the facility treats it as a product to be handed off. Water circulating in the city’s district heating network is routed through the data centre’s cooling equipment, where it absorbs heat from the servers. That water leaves warmer than it arrived. Because the temperature is still below what a domestic radiator circuit needs, heat pumps then lift it further, and the upgraded heat is distributed to households through the existing network of insulated pipes.
The number attached to the arrangement — enough heat for about 11,000 homes — is the figure circulating with the story. It is a capacity-style estimate rather than a measured constant: household heat demand varies with weather, insulation and the size of the dwellings, so the same output serves different numbers of homes in different conditions. The underlying engineering claim is the more durable one: the heat exists whether or not anyone captures it, and here it is being captured.
Why this is drawing attention now
Data centre construction has become a visible political and infrastructural issue in many countries, driven in large part by demand for AI training and inference hardware. That has sharpened public scrutiny of what these buildings consume — electricity, water, land, grid capacity — and what they give back to the places that host them. Heat recovery is one of the few answers with a tangible local benefit, which is why an operating example travels widely.
There is also a policy dimension. European energy rules have been moving towards requiring larger data centres to assess or demonstrate the recovery of waste heat where it is technically feasible, and to report energy performance data. A functioning installation in a country with dense district heating infrastructure serves as the reference case for what compliance can look like in practice.
The background a newcomer needs
Two separate systems have to meet for this to work. The first is the data centre’s cooling loop. Servers convert essentially all their electrical input into heat; that heat is picked up by air or liquid and transferred, usually via chillers and heat exchangers, to the outside world. The temperature of that rejected heat is typically low — warm rather than hot — because keeping chips within their operating range does not require carrying heat away at high temperature.
The second system is district heating: a network of buried, insulated pipes carrying hot water from central sources to buildings across a town or city. Denmark has one of the most developed district heating sectors anywhere, built out over decades and covering a large share of homes. That existing pipe network is what makes the data centre’s heat useful. Without it, there is no route from the building to anybody’s radiator, and no customer at the other end.
The bridge between the two is the heat pump. A heat pump moves heat from a cooler place to a warmer one using electricity, in the same way a refrigerator does in reverse. Feeding it water that is already lukewarm, rather than cold outdoor air, means it works across a smaller temperature difference and therefore uses less electricity per unit of heat delivered. This is why data centre waste heat is a useful input rather than a marginal one.
Who is affected, and how
For households on the network, the practical effect is a change in where their heat originates rather than in how it is delivered. Radiators, controls and billing arrangements stay as they are. Whether the heat is cheaper depends on the commercial terms agreed between the operator and the utility, which are not public in most such deals.
For the utility, recovered heat is an additional source that can displace some fuel burned elsewhere, though it also introduces a dependency: the supply exists only while the data centre is running at load. For the data centre operator, the benefits are partly reputational and partly practical — a route for rejected heat can simplify permitting and improve relations with a host municipality, and in some configurations it reduces the energy spent on conventional cooling.
For local authorities weighing new data centre applications elsewhere, the case matters as a precedent. It gives them something concrete to ask developers for, even where the answer turns out to be that the local conditions do not support it.
Where informed people disagree
The main dispute is over how much heat recovery really changes. Critics point out that reusing waste heat does not reduce a data centre’s electricity consumption or its grid impact by a single unit; it makes better use of energy already spent. A facility that heats homes is still a facility drawing significant power, and framing heat reuse as an environmental offset can overstate the effect. Supporters counter that the comparison should be against the alternative heat source being displaced, and that recovering otherwise-wasted energy is straightforwardly better than venting it.
A second disagreement concerns replicability. Denmark’s combination of dense district heating, cool climate and supportive regulation is unusual. In regions with little or no heat network infrastructure, retrofitting pipes to reach a data centre would be a large civil engineering project with uncertain economics — and in hot climates, the demand for domestic heating is limited in the first place.
A third is technical: how far cooling systems should be redesigned to produce hotter output. Running servers with warmer liquid cooling loops yields heat that needs less upgrading, but it constrains hardware choices and cooling design. Where that trade-off lands is site-specific and contested.
The practical implications
The lesson most transferable is one of siting. Heat recovery has to be designed in early, when the location is chosen and the cooling architecture set, because the value of the heat depends almost entirely on proximity to a customer. Retrofitting a network to an existing building rarely pencils out.
It also shifts what a data centre is in planning terms. A facility with a heat offtake agreement is a piece of local energy infrastructure with a counterparty, obligations and a long-term contract, not a self-contained industrial box. That changes the negotiating position of municipalities, which can treat heat as something to bargain for alongside jobs and tax revenue.
Finally, it points to liquid cooling. As chip power densities rise, air cooling becomes less viable, and liquid systems happen to produce heat at higher, more useful temperatures. The technical direction the industry is already taking for its own reasons makes heat recovery easier as a side effect.
What to watch next
Watch whether heat recovery requirements harden from assessment obligations into binding conditions in European planning and energy rules, and whether any equivalent appears outside Europe. Watch how many new-build data centres in countries with existing heat networks announce offtake agreements, since those are the sites where the economics work without new pipe. Watch for published performance data — how much heat is actually delivered across a full year, as opposed to nameplate capacity — because that is the figure that will show whether these schemes deliver at the scale claimed. And watch whether operators begin designing cooling loops specifically for heat export rather than treating recovery as a bolt-on.
Frequently asked questions
Why do data centres produce so much heat?
Electricity supplied to servers, storage and networking equipment is almost entirely converted into heat as it does computational work. Very little leaves the building as anything else. A data centre drawing a large, continuous electrical load therefore generates a correspondingly large, continuous flow of heat that must be removed to keep hardware within safe operating temperatures.
Why are heat pumps needed if the water is already warm?
Data centre cooling systems typically produce water that is warm but not hot — well below the temperature a domestic radiator circuit requires. A heat pump uses electricity to raise that temperature to a usable level. Because it starts from lukewarm water rather than cold outdoor air, it works across a smaller temperature gap and consumes less electricity per unit of heat delivered.
What is district heating?
District heating is a system in which hot water is produced centrally and distributed through insulated underground pipes to homes and businesses across a town or city, replacing individual boilers. Denmark and several other Northern European countries have extensive district heating coverage built up over decades, which is what allows recovered industrial or data centre heat to reach households.
Does heat recovery make a data centre carbon neutral?
No. Recovering waste heat does not reduce the electricity a data centre consumes or the emissions associated with generating that electricity. Its benefit lies in displacing some other heat source that would otherwise have been used, so the environmental gain depends on what it replaces. Describing heat reuse as making a facility neutral overstates what the technique does.
Could this be done anywhere?
Not easily. The approach requires a heat distribution network close enough to the data centre to connect to, a sustained local demand for heating, and commercial terms acceptable to both operator and utility. In regions without existing heat networks, or with warm climates and little heating demand, the conditions that make the Danish arrangement work are largely absent.
Does exporting heat make the data centre less efficient?
Not necessarily, and it can help. Transferring heat to a network can replace some conventional cooling work, reducing the energy spent on chillers or fans. Designing for heat export may, however, constrain cooling architecture and hardware choices. The net effect is site-specific and depends on the temperatures involved, so it cannot be generalised from one installation.
Sources and further reading
- Danish Energy Agency — public material on the structure and coverage of the country’s district heating sector.
- European Commission — energy efficiency legislation covering waste heat recovery obligations and reporting duties for data centres.
- International Energy Agency — analysis of data centre electricity demand and cooling technologies.
- Technical and trade press covering data centre infrastructure — reporting on liquid cooling, heat reuse projects and district heating connections.
Surfaced from the reddit:technology signal “data centre heat reuse”. AI-assisted draft, editorially reviewed.
