Heat reuse and community energy

Almost every kilowatt-hour that enters a data center leaves it as heat. A 20 MW facility rejects enough heat, continuously, to warm thousands of homes, yet most data centers simply dump it into the air through cooling towers and dry coolers. Reusing that heat in district heating networks, buildings, greenhouses or industry can displace fossil fuels and turn a data center into part of a community's energy system. Heat reuse is well established in the Nordic countries and is now required by law in parts of Europe, but it faces real obstacles: the heat is low in temperature, it is often far from users, and its supply and demand do not always match. This lesson covers the quantity and quality of data center heat, heat pumps, district heating, other uses, the business and contractual models, regulation, and how to design for heat reuse.

How much heat, at what temperature

The quantity of heat is simple: essentially all the electrical energy consumed by the IT equipment and most of the facility's overhead ends up as heat. The problem is its quality, meaning its temperature. Air-cooled data halls return air at perhaps 30 to 40 °C, and chilled water returns at around 20 to 30 °C, too cool to heat most buildings directly. Liquid cooling produces much warmer heat: direct-to-chip systems can return coolant at 45 to 60 °C or more, close to the temperatures used by modern low-temperature district heating networks (around 50 to 70 °C) and well suited to heat pumps. The trend toward liquid cooling for AI hardware therefore makes heat reuse much more attractive.

Heat pumps

To raise low-grade heat to useful temperatures, heat pumps are used. A heat pump uses electricity to move heat from a cooler source (the data center's return water) to a warmer sink (the district heating supply). Its coefficient of performance, the heat delivered per unit of electricity, depends on the temperature lift, as for chillers (Lesson 4):

where η, typically around 0.4 to 0.6, captures the difference between real heat pumps and the ideal Carnot cycle, and temperatures are in kelvins. The delivered heat is the source heat plus the electricity used by the compressor.

# Heat reuse: upgrading server waste heat with heat pumps for district heating
heat_MW = 10.0                 # waste heat available from the facility
target_C = 70.0                # district heating supply temperature
eta = 0.5                      # heat pump achieves about half of the Carnot COP

for source_name, source_C in [("air-cooled hall, 30 C return air", 30.0), ("liquid cooling, 45 C coolant", 45.0),
                              ("liquid cooling, 60 C coolant", 60.0)]:
    t_hot, t_cold = target_C + 273.15 + 3, source_C + 273.15 - 3     # 3 K heat exchanger approaches
    cop = eta * t_hot / (t_hot - t_cold)
    # Heat delivered = source heat + compressor electricity; electricity = delivered / COP
    delivered = heat_MW * cop / (cop - 1)
    elec = delivered / cop
    homes = delivered * 8760 * 0.5 / 15        # 50 % annual utilization, 15 MWh of heat per home per year
    print(f"{source_name:<34} COP {cop:4.1f}: delivers {delivered:5.1f} MW using {elec:4.1f} MW of electricity, ~{homes:,.0f} homes")

air-cooled hall, 30 C return air   COP  3.8: delivers  13.6 MW using  3.6 MW of electricity, ~3,977 homes
liquid cooling, 45 C coolant       COP  5.6: delivers  12.2 MW using  2.2 MW of electricity, ~3,557 homes
liquid cooling, 60 C coolant       COP 10.8: delivers  11.0 MW using  1.0 MW of electricity, ~3,217 homes

From 30 °C air-cooled return, the heat pump achieves a COP of about 3.8, delivering 13.6 MW of heat while using 3.6 MW of electricity. From 60 °C liquid cooling return, the COP is above 10, and only about 1 MW of electricity is needed. Warmer source heat means far less electricity per unit of useful heat, which is why liquid-cooled facilities are the best candidates for reuse. In every case, the delivered heat can supply thousands of homes, and the heat pumps also take over some of the data center's cooling duty, reducing its own cooling energy when the network takes heat.

District heating

District heating networks distribute hot water through insulated pipes to many buildings from central sources. They are common in Scandinavia, Central and Eastern Europe, and some North American city centers and campuses, and they are the most natural outlet for data center heat, because they can absorb large, steady quantities. Several well-known examples exist: data centers in Stockholm feed the city's network under a program that pays for recovered heat; large facilities in Finland and Denmark supply municipal networks, including a major project in the Helsinki region; and modern "fifth-generation" networks operate at low temperatures, sometimes near ambient, letting each building use its own heat pump and exchange heat with others. Where no network exists, building one is costly and slow, so heat reuse is easiest when a data center is sited near an existing network or a new development planned with one.

Other uses of data center heat

Where district heating is unavailable, other users may be nearby:

• Buildings on the same campus or next door: offices, residential buildings, swimming pools and schools.
• Greenhouses and vertical farms, which need steady warmth year-round in cool climates.
• Aquaculture (fish farming), which benefits from warm water.
• Industrial processes needing low-temperature heat, such as drying of timber, food and biomass.
• Snow melting and de-icing of pavements in cold climates.

Each use has a different temperature requirement, scale and seasonal profile, and many projects combine several users.

Matching supply and demand

Data center heat is steady year-round, while heating demand peaks in winter and falls to domestic hot water levels in summer. As a result, a data center can typically supply only part of its heat to a heating network over a year, unless the network has other year-round loads or seasonal storage. The data center must also be able to reject all its heat itself whenever the network does not need it, because cooling cannot depend on a third party. Heat reuse systems therefore sit alongside, not instead of, normal heat rejection, and the energy reuse factor (ERF, Lesson 2) is usually well below 1.

Business models and contracts

Heat reuse involves at least two parties, the data center and the heat user or network operator, with different priorities. Typical arrangements include the network operator owning and operating the heat pumps and buying heat at a price linked to its alternative fuel costs, or the data center selling heat at the fence. Contracts must address heat quantity and temperature commitments (usually "as available" from the data center's side), liability if heat delivery stops, metering, pricing, the cost of connection, responsibility for the heat pumps, and what happens when the data center changes its cooling technology or load. Public support, such as grants for connections or carbon pricing that raises the value of displaced fossil heat, often makes projects viable.

Regulation and planning

European regulation is pushing heat reuse. The recast EU Energy Efficiency Directive requires large new data centers to assess and, where feasible, use their waste heat, and some member states go further: Germany's Energy Efficiency Act requires new data centers to reuse a rising share of their waste heat over the coming years. Several cities make heat reuse a condition of planning approval.

In the United States, there is no federal requirement for data center heat reuse, and district heating is less common than in Europe, mostly found in older city centers, university campuses and military bases. A few projects use data center heat for campus buildings, greenhouses or nearby developments, and some states and cities are studying requirements or incentives as data center growth draws attention to wasted heat.

Designing for heat reuse

Heat reuse is far easier when planned from the start. Designers can choose sites near existing or planned heat networks or users; select cooling systems with warm return temperatures (liquid cooling, warm-water chilled systems); provide space, pipework connections and heat exchangers for future heat export; meter heat flows; and engage early with municipalities and network operators. Even where no user exists today, a "heat-ready" design with connection points and space for heat pumps keeps the option open for the facility's decades-long life.

Measuring and verifying reused heat

Heat delivered to a network should be metered with calibrated heat meters, which combine flow rate and the temperature difference between supply and return to calculate energy. Metered heat supports billing, the ERF calculation (Lesson 2), and any regulatory or certification claims. Where heat pumps are involved, their electricity use must be metered too, so that the net benefit can be calculated: the fossil heat displaced, minus the emissions from the heat pump's electricity. On a clean grid, the net benefit is large; on a carbon-intensive grid, it depends on the heat pump's COP and on what fuel the network would otherwise burn.

Even where full heat reuse is impractical, smaller steps help: heating the facility's own offices and generator rooms with recovered heat, and preheating domestic hot water, use heat that would otherwise be wasted.