Heat Reuse & Waste-Heat Recovery (Engineering)

A 100 MW IT load rejects very close to 100 MW of heat: essentially all the electrical energy entering the building leaves as low-grade warmth. That source runs year-round next to cities that burn gas to stay warm, so the pull to sell it is obvious, and in 2026 it is increasingly a legal obligation. The constraint is grade, not quantity. Heat at 30 °C needs paid electrical lift to serve a customer asking for 80 °C, while heat delivered at the right temperature can be a tradable commodity if the customer needs it when you make it. The grade you can deliver was largely set one chapter upstream, when you chose the facility-water supply temperature in Chapter 5.7.

The engineering comes down to four questions. How are the grades defined, and what is each good for? How does a heat pump upgrade a sub-usable return to a district-heating supply, and what does the lift cost in COP and electricity? Where does the reuse heat exchanger tie into the loop, and how does warm-water DLC change what you can capture? And how do you survive the offtake chicken-and-egg, the scheduling and counterparty problem that kills more heat-reuse projects than any thermodynamic limit? Economics, contract structures, carbon accounting and regulatory treatment sit in Chapter 15.5; this chapter engineers the heat.

Waste heat is graded by temperature because temperature, not energy content, is what an offtaker's network and emitters require. A district-heating network has a supply-temperature spec set by its generation. On the IEA DHC reference temperatures (2017), 2nd-generation networks run above 100 °C and 3rd-generation networks at 80–100 °C. Modern 4th-generation networks typically supply 50–70 °C (Yuan et al., 2025). Emerging 5th-generation networks are ambient or near-ambient loops that put a heat pump at every building. Actual seasonal schedules vary, so check the one you would feed. Your captured stream is useful only if it can meet that network's temperature, directly or after a lift. Everything below follows from that matching condition.

The cooling modality dictates the grade you can capture, which is why the cooling chapters sit upstream of this one. Air cooling and rear-door heat exchangers (Chapter 5.2, Chapter 5.3) give you a diffuse, low-grade stream, return air or RDHx water around 30–40 °C, spread over large flows and surface areas that are expensive to concentrate. Direct-to-chip liquid (Chapter 5.4) concentrates almost all the heat into one small-flow, higher-temperature stream. NVIDIA's Rubin liquid-cooling example describes about 45 °C coolant entering and about 55 °C leaving the fully liquid-cooled chip, while other D2C products set their own OEM TCS/FWS limits and ΔT. That concentration, not any extra heat, lets DLC deliver usable grade in a form a heat exchanger can collect cheaply, and it shortens the lift a heat pump must make. Liquid cooling alone does not guarantee that grade, because the supply temperature sets it. NREL's ESIF data center feeds its cold plates at about 24 °C and gets 35–40 °C back, low-grade water it uses to heat the building's offices and labs (NREL, 2020).

The table sorts halls by cooling modality. An air-cooled hall sits in the top row: a low source temperature, a large lift, and an offtaker who must be close or forgiving. A warm-water DLC hall sits in the second row, where a warmer source shortens the lift. A modest lift from the DLC return to a 4th-generation network's 50–70 °C supply can be worth selling if the COP at those temperatures supports it and the customer takes the heat. Air-cooled recovery can pay too: a 2026 IEA Heat Pump Conference study of a 4 MW air-cooled colocation hall found heat-pump reuse cost-competitive with other renewable heat under its assumptions. Check your own source temperatures, the COP at those temperatures, the annual heat the customer will take and the connection terms before you rule out either row. Even so, the cooling decision in Chapter 5.4 and the loop-temperature decision in Chapter 5.7 do most of the work. By the time you reach this chapter, the grade is mostly set.

Absorption chilling sits in the bottom row because a single-stage lithium-bromide chiller wants hot water at about 200–240 °F (93–116 °C). It returns only about 0.7–0.8 units of cooling per unit of that heat (US DOE, 2017). Lifting data-center heat that far to cool your own hall costs more electricity than an electric chiller would. Absorption pays only when another process has already heated the drive water that hot, leaving the chiller no lift to add.

A heat pump moves heat from a cooler source to a warmer sink and uses electricity to do it. For waste-heat upgrade, the source is your DLC or RDHx return and the sink is the district-heating supply. The governing number is the coefficient of performance (COP): useful heat delivered per unit of electrical work. COP falls as the temperature lift rises, which is Carnot's tax and cannot be avoided. The lift that sets COP is the refrigerant's, from evaporating to condensing temperature, and it is wider than the water-side difference by each exchanger's approach plus the amount the source water is cooled. Lifting a 50 °C source to a 65 °C sink is a 15 K lift in the water but about 26 K for the refrigerant.

You can screen a scheme before any vendor quote. The ideal heating COP is the condensing temperature divided by the condensing-minus-evaporating difference, both in kelvin. A 2017 review by Arpagaus et al. found most commercial high-temperature heat pumps running at 40–60% of their ideal COP, mostly on larger lifts than the cases below. Half of the refrigerant-side ideal, which already counts the exchanger approaches, is a fair screen rather than a guaranteed floor. Lifting a 50 °C facility return to 65 °C, with the return cooled by 5 K, needs the refrigerant to evaporate near 42 °C and condense near 68 °C, so the ideal is about 13 and the real COP about 6.5. For a 30 °C air-side source cooled by 5 K and lifted to an 80 °C network, those temperatures are near 22 °C and 83 °C: ideal 5.8, real about 3. Yuan et al.'s 2025 review of data-center district-heating projects reports a COP of 3.2–5.0 for most cases, between those two. The screen tells you which schemes deserve a quote; the COP you design to comes from the chosen machine's rating at your source and sink temperatures and load.

COP sets both the electricity you burn and the carbon you account for, so it dominates the economics. Assume a COP of 5: you deliver 5 MW of heat for 1 MW of electricity, and the captured 4 MW of waste heat plus 1 MW of pump work leave as sellable heat. Assume a COP of 3: a fixed 5 MW delivery needs 5/3 MW of electricity, about two-thirds more than at COP 5, and on a power-bound site that extra power is capacity you could have spent on accelerators. A low-COP scheme is a tax on the building's IT capacity, and the warmer the source water, the smaller that tax. Designing the DLC loop warm (Chapter 5.7) is therefore the biggest lever on heat-reuse economics, because it shrinks the lift before the pump ever runs. The machine's refrigerant is a regulatory choice too. The HFOs common in high-temperature heat pumps, R-1234ze(E) and R-1233zd(E), fall within the EU's proposed universal PFAS restriction, which has not been adopted. Ammonia, butane and CO₂ machines avoid it but bring toxicity, flammability or high-pressure design (Chapter 5.8 covers refrigerant rules).

Heat reuse adds a third thermal domain to the building. The technology-cooling loop (chip to CDU) and the facility-water loop (CDU to heat rejection) already exist from Chapter 5.6 and Chapter 5.7. Reuse adds a heat-recovery loop that runs from the reuse HX, optionally through a heat pump, into a hot-water buffer and across an isolation heat exchanger to the offtaker's network. That interface, where your water meets the district utility's water, is a hydraulically and contractually critical plane. It is almost always a plate heat exchanger that keeps the two networks chemically and pressure-isolated. Neither side wants the other's water, chemistry or pressure transients, so the HX is the demarcation.

The hardest integration problem is that source heat and customer demand vary independently: the data center keeps throwing off heat while the city may stop asking for it. An AI hall rejects near-baseload heat 8,760 hours a year. District-heating demand swings from a January peak to a July trough as space-heating loads fade, and the summer trough can fall below the heat the data center produces. Demand can also drop to zero during an outage or interruption, and the GPUs do not stop. The plant must therefore reject the heat anyway when there is no buyer. The dry coolers and towers of Chapter 5.8 never leave the building, because heat reuse is an opportunistic sink layered on a mandatory one, never a replacement for it.

Warm-water DLC keeps that fallback cheap. When nobody is buying, compressor-free dry coolers can often carry the whole plant duty alone at the design ambient and within the permitted FWS supply temperature. While a customer is taking heat, the recovery exchanger supplies whatever grade that customer can use. Count on dry rejection only where the coil's capacity and approaches carry the whole load at those conditions. Where they cannot, keep the trim or alternative rejection method from Chapter 5.8. A passive dry cooler running against cold winter air removes heat but never raises its temperature, so any lift the network's supply temperature needs comes from a heat pump you size and power.

Most data-center waste heat is still dumped for want of coordination, not for want of thermodynamics. Reuse needs two parties with very different planning horizons to arrive at the same place at the same time. The data center is built in 18–36 months against an interconnection slot and a depreciation clock. A district-heating expansion, a new greenhouse or an industrial tie-in is a multi-year municipal or industrial capital project with its own permitting, financing and committee. Each side hesitates to commit capital first: the utility will not lay pipe to a heat source that may never deliver, and the operator will not buy heat pumps and reuse HXs for an offtaker that may never connect. That is the offtake chicken-and-egg. It is an engineering constraint as well as a commercial one, because the loop you build must survive the period when one side exists and the other does not.

The engineering responses are concrete. Plumb for reuse without committing to it. Install the tie-in stubs, reserve plant-room footprint and electrical headroom for a future heat pump, and route the pipe rack so a reuse HX can drop in. Loop taps, space and power are cheap to provide during construction and costly to retrofit, the same reversible-versus-irreversible logic as Chapter 1.1. Size the plant to make money with zero offtake on day one. The mandatory rejection plant must stand alone, so the reuse loop can sit dormant for a year or three until the offtaker materializes. Split the capital contractually so neither party fronts the other's risk. The model that broke the deadlock at scale is Stockholm Exergi's Open District Heating: the operator invests in the heat pumps, the utility invests in the network connection, and the heat trades as a standardized, temperature-indexed product. Eurelectric’s June 2026 report (p. 48) says that platform connects 30-plus data centers across 16 providers, and a single site can earn on the order of SEK 2M per MW of delivered heat per year for heat it would otherwise have thrown away. Stockholm Exergi’s heat-recovery page gives that payment example and makes it dependent on outdoor temperature. Write the agreed availability, temperatures, metering boundary and connection date into your own contract (15.5). Another city’s payments do not set the price of your connection.

Two 2026 rules bound the engineering. EU EED Article 26(6) requires Member States to ensure that data centers above 1 MW of total rated energy input use their waste heat unless a cost-benefit assessment shows that is not technically or economically feasible. Article 26(7)(d) separately makes newly planned or substantially refurbished data centers above 1 MW run an installation-level cost-benefit assessment, with exemptions under Article 26(8). Germany's EnEfG sets a minimum energy-reuse factor of 10% for data centers starting operation from 1 July 2026, rising to a planned 15% and 20% for the 2027 and 2028 cohorts. A government bill pending in the Bundestag (Drs. 21/8027, heard on 5 October 2026) would keep those cohorts and allow four years instead of two to reach them. It would also add an exemption when the cost-benefit analysis finds that a link to a heat network, existing or planned, is unreasonable on technical and economic grounds. Check its status before fixing a commissioning plan.

Chapter 15.5 works out which heat-reuse duty applies, the cost-benefit assessment and the offtake contract. It needs these inputs from the engineering: source and delivered heat, temperature limits, how much of the year supply and demand coincide, electrical headroom, fallback capacity and the connection schedule. A reuse obligation does not find you a buyer or set the heat pump's COP. You still have to engineer and negotiate both.

Unit input = 320/4.00 = 80.0 kW; recovered heat = 320 − 80.0 = 240 kW; source flow = 240 × 60/(4.18 × 10.0) = 344 L/min. The source balance is 240 kW recovered + 83.0 kW rejected = 323 kW. The delivery balance is 240 kW source + 80.0 kW electrical = 320 kW.

Annual source recovery is 480 MWh, delivery 640 MWh, unit electricity 160 MWh and auxiliary electricity 20.0 MWh, so system COP = 640/(160 + 20.0) = 3.56. Available annual source heat is 323 × 8,760 = 2,829 MWh, and normal rejection receives 2,349 MWh. At zero offtake the full 323 kW goes back to the independent sink, so a sink sized for 83.0 kW would fall short by 240 kW.

Now hold the design to the 90.0 kW electrical allowance, with full fallback in place. The crossover unit COP, where input just meets the allowance, is 320/(90.0 − 10.0) = 4.00. The assumed machine sits exactly on it, so the design has no electrical margin. At COP 3.00 the total input is about 117 kW, 27 kW over the allowance, so cut the delivered duty or change the machine. Chapter 15.5 values the offtake, and Chapter 13.5 tests the fallback rejection at commissioning.

Commit to heat recovery only once you have worked the source, delivery and auxiliary-energy balance and found a sink whose hourly demand overlaps the heat. Keep full rejection for the hours with no offtake, because a seasonal heat sale does not remove a year-round cooling obligation.