How to Recover Data Center Heat for District Heating: The Nordic Case Mexico Can Copy

A 10 MW data center dissipates between 7 and 9 MW of residual heat into the environment. In Nordic countries (Sweden, Finland, Norway, Denmark), that residual heat is sold to the district heating network and pays for between 30% and 50% of the data center’s cooling operating cost. In Mexico, that same heat is vented into the atmosphere. Not because the technology doesn’t exist, but because regulation, tariffs, and urban planning culture in Mexico are not designed to capture that value. This article covers how heat recovery works in data centers, where it has been implemented well, and why the opportunity in Mexico is real but will take longer than it looks.

How residual heat recovery works

Residual heat recovery from a data center is technically simple. The data center, out of necessity of its cooling system, already transfers the heat from the servers to a fluid (water or glycol). That fluid, once it has absorbed the heat from the IT load, sits at a temperature between 25°C and 45°C depending on the design. For district heating, the typical required supply temperature is 55°C to 75°C in the district. The gap between 25-45°C and 55-75°C is covered by a heat pump that raises the temperature using additional electrical energy.

The TCO of the system depends on the heat pump’s COP: a COP of 3.5 means that for every 1 kW of electricity into the pump, 3.5 kW of useful thermal output is generated. Adding the energy already present in the data center’s water (the 7-9 MW from the example), the system delivers between 25 and 35 MW of useful thermal power for every 10 MW of IT consumed. That thermal energy is sold to the district heating network at a tariff that in Nordic countries runs from 30 to 80 EUR per MWh.

The Nordic case: how they did it

Nordic countries have the combination of 4 factors that make heat recovery economically viable:

1. Sustained cold climate: heating demand is continuous for 8 to 10 months of the year, not seasonal. The economics work if the heat sink is available.
2. Preexisting district heating network: Stockholm, Helsinki, Copenhagen, Oslo have district heating networks operated by municipalities or utilities that already carry hot water at 60-80°C to residential, commercial, and industrial buildings. The data center connects to that existing network as one more supplier.
3. Regulation that permits the sale of heat: there is a regulatory framework that defines the tariff, the technical requirements, and the purchase-and-sale contracts. The data center knows how much it will collect for its residual heat before investing.
4. Electricity tariff that penalizes heat waste: in Sweden, electricity carries a tax component on the CO₂ emitted in its generation. That raises the cost of energy, which makes investing in heat recovery to reduce consumption more profitable.

The Stockholm case is the most mature. The company Stockholm Exergi operates the district heating network, and data centers such as Bahnhof (operator of the data center that hosts The Pirate Bay) and other hyperscalers like AWS (in its eu-north-1 region) sell their residual heat to the network. AWS has reported that its Stockholm region recovers more than 90% of residual heat for district heating.

What is missing in Mexico for this to be viable

The technical opportunity exists. What does not exist is the combination of the 4 Nordic factors. Mexico has its own combination of barriers:

5. Climate with seasonal demand: most of Mexico (except the highlands of the north and the Altiplano) has mild winters with heating demand for 3 to 5 months of the year, not 8 to 10. The district heating economics work in CDMX or Guadalajara on a 6-month basis, not 12.
6. No preexisting district heating network: Mexico never built municipal heating networks. The construction culture assumes individual heating (gas, electric, or no heating at all). Building a district heating network from scratch is enormous CAPEX that no private party will do without public subsidy.
7. Regulation that does not permit the sale of heat: the CRE (Comisión Reguladora de Energía) and the Geothermal Energy Law and other Mexican norms have no specific framework for ‘residual heat sales service’ as a regulated activity. That does not prevent a data center from selling heat to a nearby industrial or commercial client under a private contract, but it does not scale.
8. Electricity tariff that does not penalize waste: the CFE industrial tariff in Mexico (GDMTH for high voltage, DIST for medium voltage) does not include a significant CO₂ tax component. That makes heat recovery pay back over 7-10 years, not 3-5 as in Nordic countries.

Cases where heat recovery DOES work in Mexico

Although the Nordic district heating model does not apply, there are specific applications in Mexico where heat recovery does close financially:

9. Data center adjacent to an industrial client that needs process heat: a data center near a food plant, a chemical plant, or a greenhouse can sell steam or hot water directly, without going through a municipal network. The typical case: a data center in Querétaro adjacent to an industrial park with process steam demand. Low connection CAPEX, shared OPEX.
10. Domestic hot water heating for a campus or large development: on a university campus, a hospital, or a mixed-use development, the data center can feed the domestic hot water heating system. The heat pump COP is favorable because the target temperature (45-55°C) is lower than the district heating one.
11. Heating of swimming pools or municipal pools: target temperatures of 25-30°C are easily reached with the data center’s water without an additional heat pump, which maximizes the economic efficiency.
12. Greenhouse agriculture: a data center in a semi-arid zone adjacent to greenhouses can sell residual heat to the agricultural operation. Greenhouses need controlled nighttime temperatures during winter, and data center residual heat is ideal.

The TCO calculation for an implementation in Mexico

Take a 5 MW data center with 3.5 MW of recoverable residual heat, adjacent to an industrial client with continuous heat demand 12 months of the year. The typical numbers:

13. CAPEX of the heat pump and distribution system: between 8 and 15 million pesos for 3.5 MW thermal, depending on the distance to the client and the complexity of the heat exchanger.
14. Electrical OPEX of the heat pump: between 1.0 and 1.5 million pesos per year (assuming COP 3.5 and the GDMTH tariff).
15. Revenue from heat sales: between 4 and 8 million pesos per year, depending on the rate negotiated with the industrial client (typically 200-400 pesos per thermal GJ, comparable to industrial natural gas).
16. ROI: 2.5 to 4 years, without considering the value of emission reductions for ESG reporting. If the industrial client pays a premium for ‘green’ heat (replacing natural gas with zero CO₂), ROI drops to 1.5-2.5 years.

Why Mexico will not copy the Nordic model (in the short term)

The political and regulatory reality in Mexico makes it unlikely that we will see Nordic-scale district heating in the next 10 years. The initial investment of a municipal network is on the order of 50 to 100 million pesos per kilometer, and no Mexican municipality has that in its budget nor is going to subsidize it. The real opportunity for data centers in Mexico is in adjacent cases: industrial, campus, agriculture, hospital. Those specific cases can close financially with ROI of 2-4 years, but they do not scale to the municipal network model.

The verdict

Recovering heat from the data center in Mexico is technically viable and financially attractive in adjacent cases, but the Nordic model of urban-scale district heating will not be replicated in Mexico in the short term. The real opportunity for operators in Mexico is to identify industrial clients, campuses, or greenhouses adjacent to the site and design the cooling system from the start with heat recovery in mind. The incremental investment is low (10-15% over the cooling system CAPEX), and ROI is 2 to 4 years. If your site is in a zone with continuous thermal demand and an adjacent client, the question is not whether you should do it, but how to structure the heat sales contract so the client pays and the data center recovers the investment. Regulation will arrive later; the first movers in this space will have a competitive advantage that followers will not easily match.

Sources

ASHRAE TC 9.9 — Thermal Guidelines for Data Processing Environments (5th Edition, 2021, heat reuse concepts) —.https://www.ashrae.org/technical-resources/bookstore/thermal-guidelines-for-data-processing-environments.

Uptime Institute — 2024 Global Data Center Survey (heat reuse trends and adoption rates) —.https://uptimeinstitute.com/resources/asset/2024-global-data-center-survey-results.

The Green Grid — Data Center Heat Reuse Best Practices —.https://www.thegreengrid.org/.

Stockholm Exergi — District heating technical specifications and pricing —.https://www.stockholmsenergi.se/.

CRE — Regulatory framework of the Comisión Reguladora de Energía (residual heat sales, district heating) —.https://www.gob.mx/cre-historico.