Data centres: “Laws and public governance struggle to keep up with the AI revolution” (E.Morello, POLIMI)

ProdAction: how did it start, and what are its goals?

ProdAction began in 2024 as a continuation of "Aree produttive, aree proadattive" ("Productive areas, proadaptive areas"), a project funded by the Ministry of the Environment that I had coordinated in our department. At the time, the aim was to define recommendations for regenerating industrial areas in a sustainable way, improving their environmental quality and resilience, working as a living lab with companies and public administrations in a number of municipalities around Milan. With ProdAction, however, we identified new territorial impacts and new forms of production, and we immediately focused on data centres.

The project started out as a collection of information, which is still scarce and fragmented today: a job that should fall to the public administration, in particular Regione Lombardia, and that we have partly filled in ourselves. Alongside the mapping, we listened to operators (developers, designers) and administrations. As urban planners, our goal is to support public decision-making.

At first the dialogue was positive: data centres were seen as an opportunity to regenerate disused sites in need of remediation, and municipalities considered them less impactful than logistics, a "box" with computers that generates no traffic. Then came the wave of what is happening abroad, from the United States to Northern Europe, where Ireland has already saturated the market. Operators move to where conflict has not yet emerged, and when the impact becomes visible, conflicts arise there too.

Who are the main recipients of this kind of mapping?

We designed the mapping primarily with regional and metropolitan governance in mind, which is currently lacking: we wanted to make visible a phenomenon that is still invisible, because until we map it, we can't understand its scale and impacts.

The reasoning is "I won't do it in my municipality, the one next door will". But if there are 20 data centres around me, I suffer externalities too. There are no models for negotiation and concertation: the metropolitan city has run some experiments, but without effective results, particularly on compensation, which concerns not only the host municipality but also the surrounding ones. We need data and scientific quantification to calculate the presumed impacts.

The second purpose is to inform citizens. The awareness that is emerging is partly slowing the system down and has pushed operators to manage the conflict, sit down at a table and improve their offers to the territories in terms of environmental compatibility and sustainability.

How many data centres are there in Italy and, specifically, in the Milan area?

Nationally, we counted about 200 data centres built, of which 32 are active in the metropolitan city of Milan, which however accounts for about 68% of national nominal capacity: the largest and latest-generation data centres are all here, because it is the first area where the phenomenon developed.

In addition to the 32 already operating, 10 are under construction, 29 are under evaluation and 6 have been announced. To meet the growing energy demand, 10 new Terna substations are also being built (3 approved, 7 pending approval), plus another 17 announced in Lombardy, about which Terna has released little information. The figures are constantly changing, and it is hard to keep up with the sector's development.

Everything starts from Milan, in particular from the surrounding municipalities, which operators find easier to "assault": in the city itself it is very hard to find space for a large data centre or a campus. Development starts from the metropolitan city and expands southwards.

Why do data centres tend to cluster in the same area? And why Milan in particular?

Milan combines more positive characteristics than other areas. It is a metropolitan setting open to innovation, with technical know-how and, for decades, the presence of major ICT clients, meaning the headquarters of big companies: proximity to clients is a key factor. It is well served in terms of ICT networks (fiber optics, and the location of the MIX, the Milan Internet eXchange, Italy's main Internet Exchange Point), energy, mobility and logistics. It lies on a plain, so there is no seismic risk, and it is rich in water. It has no particular landscape constraints and is a territory with a strong industrial vocation, with numerous disused and already urbanised areas. For all these reasons, it is an ideal location.

Governance also matters, as it is rather fragmented: operators approach small municipalities that, at the outset, did not know what to expect from this kind of development, or how and on what to negotiate (somewhat more so today), and so found the ground easy.

There is also an agglomeration effect: setting up where operators already have a data centres is faster. In the North-West there are municipalities with more than ten data centres, where the paperwork, technicians and procedures are well known. Like all industrial districts, data centres cluster where they find the conditions and the institutional capacity to build them. In a municipality or region that does not know what a data centre is or which procedures to follow, the process would be slower, because a range of stakeholders would have to be brought on board. Speed of delivery is another decisive factor for operators, and it makes Milan a well-equipped setting.

What are ProdAction's biggest limitations and challenges?

We began with a listening phase that also involved operators and designers. With recent developments, however, they have realised that the topic has become a matter of public debate and is generating conflict among citizens. We therefore hope to resume the conversation with the sector to consider together how to set up more open and collaborative participatory and deliberative processes. As a university, we could be a more objective actor, able to bring the different stakeholders around a table and try to reconcile significant development with territorial compatibility, moving beyond a head-on standoff. The first challenge is to attempt a dialogue with a sector that needs to deploy strategies and tools for more concerted negotiation, rather than taking an increasingly defensive stance.

Another challenge is the speed of AI, which is growing exponentially: as long as the mapping covered 10, 20 or 30 data centres we could keep up, but now that they number in the hundreds, that is no longer possible. Public support is needed to take this on.

The real challenge is rigorous scientific monitoring of impacts. On soil, energy, water, air, heat, PFAS and other issues, we are working largely on hearsay: the literature yields little information, and we have no evidence on projects in the metropolitan and national context. Public support is needed, including from environmental agencies. This is also in the sector's interest: if operators claim that water is not consumed and heat is not released, they can prove it on the basis of more robust monitoring.

In a way, we are a “crutch” for the public sector, which has neither the strength nor the institutional capacity for this kind of mapping, nor the leverage to negotiate to make projects more compatible with the context in which they are located. Laws and public governance struggle to keep pace with the AI revolution. We are supporting, or in effect substituting for, work that a supra-local public authority could carry out.