Data centers spend enormous amounts of energy removing heat. The next sustainability advantage may come from finding somewhere useful to send it.

Every AI workload produces something besides intelligence.

It produces heat.

Electricity enters a data center, moves through processors, storage systems and networking equipment, and eventually leaves the computing environment as thermal energy. Cooling systems capture that heat and move it away from the equipment so the facility can continue operating.

In most data centers, that is where the story ends.

The heat is transferred into the surrounding air, water or atmosphere. It has served its technical purpose, and the facility pays to get rid of it.

But a different infrastructure model is beginning to emerge.

Instead of treating data-center heat as an unwanted byproduct, some operators, utilities and communities are capturing it and sending it to district-heating networks, public buildings, homes, industrial facilities and other nearby users.

This changes the sustainability equation.

The data center still generates heat. The difference is that the surrounding infrastructure gives that heat somewhere useful to go.

The Short Answer

Data-center waste heat is the thermal energy created by servers, processors and other computing equipment during operation.

Heat-reuse systems capture that energy from the cooling loop, transfer it through heat exchangers, raise its temperature when necessary and deliver it to a nearby customer or district-energy network.

The Concept Is Straightforward:

Capture the heat. Upgrade it. Move it. Use it again.

When the right infrastructure exists, waste heat can reduce the amount of energy required to heat buildings, lower reliance on fossil-fuel heating systems and create an additional source of value from electricity already being consumed.

But heat reuse is not automatic.

Its viability depends on temperature, distance, seasonal demand, infrastructure costs and whether a reliable customer exists to purchase or receive the heat.

The heat may be available.

The market for it still has to be built.

The Signal

Data-center sustainability has traditionally focused on the electricity entering the facility.

Where did the power come from? How efficiently was it used? How much carbon was associated with it?

Those questions remain important. But they examine only one side of the energy flow.

The next question is what happens after that electricity becomes heat.

The US Department of Energy’s data-center design guidance states that higher temperatures leaving servers create greater opportunities for reuse. It also identifies several conditions that improve project viability: a nearby heat customer, compatible temperature requirements, aligned ownership and supportive policies or incentives. US Department of Energy

That moves heat reuse beyond a cooling conversation.

It becomes a question of infrastructure coordination.

A data-center operator may control the servers and cooling system. A utility may control the district-energy network. A municipality may control rights-of-way. A developer may control the neighboring land. A hospital, university, factory or residential district may become the heat customer.

No single participant creates the opportunity alone.

The value appears when the systems connect.

What Happens to the Heat?

Most electricity used by servers eventually leaves the IT equipment as heat.

Cooling systems collect that heat and carry it away from sensitive components. Depending on the facility, heat may be transported through air, water or another liquid circulating through the computing environment.

From there, the facility normally rejects it through equipment such as chillers, cooling towers, dry coolers or other heat-rejection systems.

Heat reuse inserts another pathway before that final rejection.

Instead of immediately releasing the thermal energy, the facility directs some of it through a heat exchanger. A heat pump may then raise the temperature to a level suitable for buildings or a district-heating network.

The resulting heat can be delivered to a nearby user.

The data center still requires backup heat-rejection capacity because the outside customer may not need heat continuously. Demand can fall during warmer months, equipment can go offline and the receiving network may experience interruptions.

Heat reuse does not eliminate cooling infrastructure.

It gives the cooling system another destination.

The Four-Part Heat-Reuse Stack

A viable data-center heat project usually depends on four connected systems.

1. Heat Capture

The first system collects heat as close to the computing equipment as practical.

Traditional air cooling can support heat recovery, but liquid-cooling systems may improve the opportunity by capturing thermal energy directly from high-density processors.

Direct-to-chip cooling, rear-door heat exchangers and immersion systems can move heat into a liquid loop more efficiently than relying entirely on room air.

This does not guarantee that the heat will be reused.

It makes the resource easier to collect and transport.

2. Temperature Upgrade

Data-center heat is often considered low-grade heat because its temperature may be below what an existing heating network requires.

A heat pump can raise that temperature.

This step matters because temperature quality influences value. Warm water that is useful for a greenhouse or low-temperature heating loop may not be hot enough for an older district-heating system without additional equipment.

The hotter the recovered fluid, the less energy may be required to make it useful.

That is one reason higher-temperature liquid-cooling loops could become strategically important beyond the data center itself.

They may improve the economics of the infrastructure outside it.

3. Heat Transportation

Captured heat must physically reach the customer.

That usually requires insulated pipes, pumps, heat exchangers, rights-of-way, control systems and connections to buildings or a district-energy network.

Distance creates cost and energy loss.

A major heat customer located next door may create a realistic opportunity. A customer located many miles away may not.

This is why a glowing line on a conceptual sustainability diagram is not enough.

Someone must finance, permit, construct and operate the infrastructure underneath it.

4. Heat Offtake

The final requirement is demand.

A heat-reuse system needs a customer capable of using thermal energy at the right temperature, volume and time.

Potential Customers Can Include:

• District-heating networks

• Hospitals and healthcare campuses

• Universities

• Municipal buildings

• Residential developments

• Hotels

• Industrial facilities

• Greenhouses

• Aquaculture operations

• Domestic hot-water systems

• Neighboring offices and warehouses

The strongest opportunities involve stable, year-round demand.

A facility that needs heat only during the coldest part of the year may still participate, but the economics must account for lower summer demand and backup cooling requirements.

The heat customer is not an optional addition.

It is part of the infrastructure stack.

Liquid Cooling Changes the Opportunity

Liquid cooling is usually discussed as a response to rising rack density.

As processors become more powerful, moving heat through air becomes increasingly difficult. Liquid can capture and transport heat more efficiently, allowing facilities to support denser computing environments.

But there is a second-order opportunity.

When heat is collected in a liquid at a useful temperature, it becomes easier to move beyond the data center.

That does not mean every liquid-cooled facility will become a district-heating plant. Climate, location and nearby demand still matter.

It does mean cooling architecture may begin influencing regional energy planning.

A decision made inside the server room can affect what becomes possible outside the building.

This is where the water and sustainability pillars begin to overlap.

Liquid cooling does not necessarily mean high water consumption, especially when the facility uses sealed or closed-loop systems. Likewise, a liquid-cooling system does not automatically mean heat reuse.

The cooling loop determines how heat moves.

The surrounding infrastructure determines whether that heat becomes valuable.

The Market Is Responding

Some of the most advanced heat-reuse projects are emerging where large data centers intersect with established district-heating networks.

In Tallaght, Ireland, recovered heat from an Amazon Web Services data center supplies the Tallaght District Heating Scheme. AWS says the system serves approximately 55,000 square meters of public, residential and commercial space. The company provides the recovered heat to the network without charge. Amazon

Finland is approaching the opportunity at a much larger scale.

Fortum has built heat-pump plants designed to integrate waste heat from Microsoft data centers in Espoo and Kirkkonummi. The €225 million system includes water-to-water heat pumps capable of producing up to 180 megawatts of district heat, electric boilers and an 800-megawatt-hour thermal-storage facility.

Fortum expects recovered data-center heat to eventually cover approximately 40% of the area’s annual two-terawatt-hour district-heating demand, serving a network with roughly 250,000 users. Integration of Microsoft’s waste heat is expected to begin in phases in 2027 as the data-center campuses are commissioned. Fortum

These projects are notable because they extend far beyond cooling equipment.

They require utilities, data-center operators, cities, grid companies, heat-pump manufacturers, construction firms and customers to coordinate around a shared infrastructure plan.

This is what sustainability looks like when it becomes physical.

The Policy Signal

Europe is beginning to move heat reuse from a voluntary sustainability measure toward an infrastructure-planning requirement.

The European Union’s recast Energy Efficiency Directive requires member states to ensure that data centers with a total rated energy input above one megawatt utilize waste heat or another heat-recovery application unless the operator can demonstrate that doing so is not technically or economically feasible.

The directive also calls for cost-benefit analysis of factors including local heat demand, seasonal variation, connection to district-heating systems and cooling designs capable of capturing heat at a useful temperature. European Union Energy Efficiency Directive

That matters even beyond Europe.

Policies rarely stay isolated when they begin changing design standards, equipment markets and infrastructure expectations.

Developers operating globally may begin evaluating heat-recovery readiness earlier. Cities may incorporate potential heat demand into data-center planning. Utilities may consider data centers as distributed heat sources. Equipment manufacturers may design systems around higher-temperature recovery.

The sustainability conversation is moving from reporting what happened to planning what should happen next.

The Thermal Proximity Principle

Heat behaves differently from electricity or data.

Electricity can move across regional grids. Information can travel through fiber across continents. Low-temperature heat is most useful when the customer is relatively close.

This creates the Thermal Proximity Principle:

Recoverable Heat + Compatible Demand + Delivery Infrastructure = Thermal Value

A data center located near a dense heating network may have a significant advantage over an identical facility surrounded by customers that cannot use its heat.

The same is true in reverse.

A community planning a hospital district, university campus, industrial park or large residential development may create additional value by coordinating that development with nearby digital infrastructure.

Land-use planning becomes energy planning.

The distance between the heat producer and heat customer becomes an economic variable.

The Heat Buyer Changes the Site-Selection Map

Data-center site selection usually begins with power, fiber, land, water, permitting and construction readiness.

Heat reuse adds another layer.

Developers May Begin Asking:

• Is there an existing district-heating network?

• Where are the largest nearby thermal loads?

• Are those customers operating year-round?

• What temperatures do they require?

• Who owns the land between the facility and the customer?

• Can insulated pipes be installed within existing rights-of-way?

• Is a municipal utility willing to operate the network?

• Are grants or incentives available?

• Could future development be planned around the heat source?

• Who will purchase the heat, and under what contract?

These questions will not determine every project.

But in regions where sustainability expectations, energy costs and community scrutiny are rising, they could influence which sites become more attractive.

The data center may still be built for computing.

Its thermal output can become part of the regional development strategy.

Company Watch: Equinix

Equinix is demonstrating how heat export can move from an individual demonstration project into a broader operating program.

The company reported exporting 14.5 gigawatt-hours of residual heat from its data centers in 2024, an increase of 245% from the previous year. Its heat-export initiatives include facilities in Helsinki, Toronto and Paris, where recovered thermal energy is supplied to district networks and community facilities. Equinix

This makes Equinix worth watching for more than its data-center portfolio.

The company operates facilities in dense metropolitan markets where data centers, buildings and energy networks are located close together. That geography could make certain campuses well positioned for heat partnerships.

The Larger Opportunity Includes Several Categories of Companies:

• Data-center developers and operators

• District-energy utilities

• Heat-pump manufacturers

• Cooling-system providers

• Engineering and construction firms

• Thermal-storage companies

• Building-controls providers

• Municipal energy companies

Everyone is watching who supplies electricity to AI.

The next watchlist may include the companies that move the heat away from it—and sell that heat twice.

A Metric That Deserves More Attention

Power Usage Effectiveness, or PUE, has become the most widely recognized measure of data-center energy efficiency.

It compares the total energy entering a facility with the energy used by its IT equipment.

PUE remains useful, but it does not fully capture the value created when energy leaving the facility is used somewhere else.

Energy Reuse Effectiveness, or ERE, attempts to account for that recovered energy.

The US Department of Energy includes ERE among the metrics that operators can use alongside PUE, Water Usage Effectiveness and Carbon Usage Effectiveness to develop a more complete view of facility performance. US Department of Energy

That distinction may become increasingly important.

A data center can be highly efficient within its property boundary while still releasing large amounts of usable thermal energy.

The next generation of sustainability metrics may need to examine the entire regional system.

Why Heat Reuse Will Not Work Everywhere

Heat recovery is promising, but it should not be presented as a universal solution.

Several barriers can prevent a project from working.

The Heat May Be Too Cool

Low-temperature heat may require a large heat pump, increasing capital costs and electricity use.

The Customer May Be Too Far Away

Longer pipe networks cost more to build and lose more heat during transportation.

Demand May Be Seasonal

A district may require substantial heat in winter and very little in summer. The data center produces heat throughout the year.

The Network May Not Exist

Many communities do not have district-heating infrastructure. Building a new network can require major investment and coordination.

The Contracts May Be Complicated

The data-center operator, heat-network owner and final customer may be different organizations with different investment timelines and risk requirements.

Reliability Must Be Protected

The data center cannot depend on the heat customer always being available. Backup heat-rejection systems remain necessary to protect the computing equipment.

These limitations do not invalidate the opportunity.

They explain why site selection, partnerships and early planning matter so much.

Sustainability Across the Eight Pillars

Waste-heat reuse connects nearly every pillar of the Infrastructure of Intelligence™.

It begins with power because electricity entering the data center eventually creates heat. It intersects with water through cooling-system design. It influences land because heat customers and pipe routes must be located nearby.

Connectivity supports the control systems coordinating the facility, heat pumps, thermal storage and district network. Workforce determines whether engineers and technicians are available to design and maintain those systems.

Policy can require feasibility studies, support infrastructure investment or establish utility rules. Capital determines who finances the heat plant and distribution network. Sustainability measures the broader result.

The eight pillars are not separate boxes.

Heat makes their interdependence visible.

What Leaders Should Start Asking

Data-center operators, utilities, developers and communities should begin evaluating heat before a facility is fully designed.

They Should Ask:

• How much recoverable heat will the campus produce?

• At what temperatures will that heat be available?

• How will the cooling architecture affect heat quality?

• Which potential heat customers are located nearby?

• What are their hourly and seasonal demand profiles?

• Is there an existing district-energy network?

• Would a heat pump be required?

• How much electricity would the heat pump consume?

• Who would finance the recovery and distribution infrastructure?

• Who would own and operate it?

• What rights-of-way would be required?

• What happens when the heat customer is unavailable?

• Can thermal storage improve the match between supply and demand?

• Are policy incentives or infrastructure grants available?

• Should neighboring development be planned around the heat source?

The earlier these questions are asked, the more options remain available.

Frequently Asked Questions

Can Waste Heat From an AI Data- Center Heat Homes?

Yes. Data-center heat can be transferred into a district-heating system and used for space heating or hot water. A heat pump is often required to raise the temperature before distribution.

Does Liquid Cooling Make Heat Reuse Easier?

It can. Liquid can capture heat closer to high-density processors and transport it efficiently. Higher-temperature liquid leaving the server can improve reuse potential, but a nearby customer and delivery infrastructure are still required.

Does Heat Reuse Eliminate the Need for Cooling Towers or Chillers?

Not necessarily. Facilities generally need backup heat-rejection systems for periods when the outside customer cannot accept the heat. Certain designs may reduce reliance on traditional cooling equipment, but operational redundancy must be preserved.

Why Is Data-Center Heat Reuse More Common In Europe?

Many European cities already have district-heating networks, dense urban development and policies encouraging waste-heat recovery. Those conditions make it easier to connect large heat producers with nearby customers.

Can Data-Centers Sell Recovered Heat?

Potentially. Commercial arrangements vary. An operator might sell the heat, provide it without charge, share infrastructure costs or enter a long-term agreement with a utility or district-energy provider.

The Golden Nugget

Heat is only waste when no infrastructure exists to use it.

The data center produces the thermal resource either way.

The opportunity is created by what surrounds it: compatible customers, heat pumps, pipes, policies, capital and coordinated planning.

The IOI Take

Waste heat could become one of the clearest examples of sustainability evolving from a reporting exercise into an infrastructure strategy.

The old model measures the energy entering the data center and the efficiency of the facility consuming it.

The emerging model follows that energy through the entire system.

• Where did the electricity come from?

• How efficiently did the processors use it?

• How much water was required to manage the heat?

• Where did the thermal energy go afterward?

• Could another building, business or community use it?

The strongest AI infrastructure regions will not evaluate data centers as isolated buildings. They will examine how power, water, heat, land and neighboring development can function together.

This is the larger sustainability opportunity.

Not simply consuming fewer resources.

Creating more value from the resources already being consumed.

One Last Thought

AI turns electricity into intelligence.

It also turns electricity into heat.

For years, the industry has treated one output as valuable and the other as something to remove.

That distinction may not last.

The next infrastructure advantage may come from recognizing that the energy leaving the data center still has work left to do.

Workforce

Policy

Capital

Sustainability