
To gain acceptance, data centers must also heat the city
This article is based on a talk by Max Croce at the event Powering AI, Protecting the Climate: Data Centres, Integrated Solutions and Verified Benefits, held at the Italian Pavilion at COP31 in Antalya.

A data center that releases its heat into the atmosphere while, just a short distance away, homes burn gas for heating represents a contradiction that industrial policy should learn to address. Technologies already exist to recover at least some of that energy, but what is still often missing is a plan to connect those who produce heat with those who need it, along with an incentive system that makes doing so economically viable.
The race to develop artificial intelligence makes this choice urgent. As of June 30, Terna reported data center grid connection requests totaling 18.2 GW in the province of Milan alone, compared with installed capacity of just over 0.5 GW at the end of 2025. Many of these projects will probably never be built, or will end up consuming far less electricity than requested, but the explosive growth trend is clear.
Attracting computing capacity is a reasonable goal. An increasing share of business competitiveness depends on it. Yet an investment can be strategic for the national economy while leaving the host community with an unsatisfactory balance of costs and benefits. Digital services reach users far away, while the additional demand for water and electricity is concentrated in a specific location. People living near a new facility have a right to know what costs they will bear and what benefits they will receive. Treating all opposition as a reflexive hostility to progress is therefore a mistake. Residents may be asking an entirely rational question: why should we allocate scarce resources to this activity? A credible answer must address the specifics of the project, going beyond the mere value of the announced investment.
The technical challenge is not trivial. The heat generated by servers may be at too low a temperature to supply a district heating network directly. Industrial heat pumps can raise that temperature, but they also consume electricity. Physical connections are needed, along with customers located sufficiently close to the facility. The environmental and economic balance depends on these conditions, as well as on the energy source being displaced.
This is why choosing a location is already a climate policy decision. An inexpensive plot of land can become a costly solution for the community if it requires new infrastructure and makes energy recovery impractical. A former industrial site close to a district heating network and adequate electricity connections may offer advantages that are not reflected in the price of the land. Assessing these factors before construction is far more effective than trying to incorporate them into a project that has already been designed. The same principle applies to water. Where technical conditions allow, using treated wastewater for cooling reduces pressure on higher-quality water resources. But this solution must be assessed at the local watershed level, taking other uses into account. A facility that appears efficient on paper may still pose problems in an area vulnerable to drought. Climate change also alters the conditions under which cooling systems will have to operate throughout their useful lives.
There are also concrete opportunities in electricity generation. When a data center is powered by gas turbines or engines, recovering heat from the generation process can make it possible to produce additional electricity without burning more fuel.
The challenge common to these measures is that they require upfront investment and agreements between parties operating on different time horizons. A data center operator wants to make computing capacity available quickly. A district heating network must plan infrastructure designed to last for decades. If each party makes decisions independently, a benefit that makes economic sense for the system as a whole may fail to attract anyone willing to finance it. Public policies are therefore needed to make the benefits sought for the community financially rewarding. Support for shared infrastructure could be tied to the amount of heat actually delivered to users and the emissions avoided. Site assessments should consider opportunities to reuse previously developed land. Permitting procedures with predictable timelines would help operators incorporate these investments into their plans rather than treating them as additions with uncertain outcomes.
The concern is global, so much so that on November 13, at COP31, Brazil, the UN, and UNESCO launched the first global framework on artificial intelligence and climate information integrity with the document Priorities for action on AI for climate information integrity. The document addresses the reliability of AI-generated information on climate change and transparency regarding AI’s own environmental impact.
Finance can also play a role. A proposal discussed within SDA Bocconi's Green Tech Initiative is to explore the use of verified carbon credits, recorded through digital tokens, to support efficiency improvements. This approach deserves experimentation, starting with the quality of measurement. A digital registry can make credits more traceable, but their credibility depends on the robustness of the emissions reductions they represent. It is therefore necessary to establish what emissions would have occurred without the intervention and demonstrate that the benefit is additional, while preventing the same reduction from being attributed to multiple parties. An existing subsidy or a legal requirement can change this assessment. Entrusting independent verifiers with checking the results helps protect both those purchasing the credits and those implementing credible projects.
A similar principle should guide relations with local communities. Before a project is authorized, commitments must be understandable and measurable. Once the facility is operating, actual results must be compared with the promises made. If public support rewards an environmental benefit, that support must also be able to decrease when the promised benefit fails to materialize.


