In an op-ed published in Le Monde on 14 August 2026, Stéphane Troussel, president of the Seine-Saint-Denis departmental council, calls for a pause and for public consultation on the development of data centres. I agree with him, and I propose to dig further: to spell out what the words data centre cover, to show that public authorities are not only the referee of these projects but also their customer, to point to the concrete steps, from free software to a server on our own premises, by which each of us can weigh on these choices, and to propose a method for making these choices together, since no one person holds alone what it takes to make them.
From Le Bourget to Geneva, four data centres that have nothing in common
Stéphane Troussel is right: the development of data centres calls for a pause and for consultation, so that the public interest can reassert itself (« En France, le développement anarchique des data centers relève davantage du Far West que de l’eldorado numérique », Le Monde, 14 August 2026). His département alone has twenty-seven of them as of summer 2026, and the Paris region holds more than a third of the national total. I would like to extend his call and make it more precise, because each of us holds a power to act that we do not suspect: public authorities are not only the referee of these projects, they are also their customer, and so are we.
At Le Bourget, the contested project that Stéphane Troussel takes as his example involved a 35,000-square-metre site and an investment of 850 million euros; its developer, Segro, is a British specialist in logistics warehouses, and the municipality withdrew the building permit, a withdrawal that the Montreuil administrative court upheld on 28 July 2026. I have chosen three other examples, very different from that one, to show that everything depends on who owns the building and what is computed there: a private campus at Dugny, a public supercomputer at Saclay, a cooperative hosting company in Geneva. At Dugny, the campus belongs to Digital Realty, the American colocation giant, which rents its halls to the major platforms: these buildings consume the area’s water and electricity and discharge their heat into the atmosphere. At Saclay, thirty kilometres away, the public supercomputer Jean Zay, operated by the CNRS, makes its computing power available to researchers free of charge, and since late 2023 the campus heating network has been recovering the heat it gives off, enough to heat the equivalent of 1,500 households. In Geneva, finally, the hosting company Infomaniak, owned entirely by its employees, buried its latest data centre in early 2025 beneath a housing cooperative: no air conditioning, servers kept for fifteen years, all the heat fed into the district heating system, enough to heat 6,000 homes at full capacity. There, a data centre becomes a neighbourhood facility.
The words data centre thus cover owners, uses, ecological footprints and ties to the local area that have nothing to do with one another. A consultation should, in my view, begin with very concrete questions, which elected officials and residents are entitled to put to the developers of each project: who will own this centre, and under which law? What will it compute, and for whom? Will its heat go into the atmosphere or into a heating network, and will its water come from the drinking-water supply? How many lasting jobs, for what tax revenue? Depending on the answers, we will know whether we are dealing with a facility serving the public interest or a real-estate operation that benefits only its owner. I proposed the same questions about the uses of artificial intelligence in the article The Land We Cultivate: who owns, who decides, who profits from the computing.
Our emails, our hospitals, our town halls are already in there
These buildings do not only contain artificial intelligence; they also hold streaming films, social networks, online payments, networked video games, the outsourced IT of businesses. Add to that the data of public authorities: our emails, our hospital records, the servers of our town halls, the documents of our universities, because we entrusted them with these. In March 2025, the French ministry of education and higher education renewed for four years, and 152 million euros, their framework agreement with Microsoft (nearly a million workstations and servers), which in practice places these data under the Cloud Act, the American law; the National Assembly’s commission of inquiry into digital dependencies, in its July 2026 report, puts public purchases of non-European software at 1.5 billion euros a year, 80% of which goes to American companies. Through these subscriptions, we fill the data centres ourselves, and we end up with our hands tied: no one will ever ration the water of a building that hosts the hospital and the prefecture; this is what I called the functional hierarchy in the article Ecology and Artificial Intelligence. In summer 2025, Microsoft moreover admitted before the Senate that it could not guarantee that its French customers’ data would be beyond the reach of the American authorities; I described this situation in detail in the article Digital sovereignty, a democratic emergency.
Yet every institution, every local authority, even every business has a lever at its disposal: redirecting these budgets, at equal cost, towards free software and digital commons. The tools exist, at every scale. The French state offers public employees La Suite numérique free of charge, an open-source suite developed with Germany and the Netherlands, which more than 500,000 employees use every month in 2026. Around Lyon, nine local authorities share a free software suite hosted in regional data centres. The gendarmerie nationale switched to free software as early as the late 2000s. And in 2025 Denmark set about moving its ministries off American software; its minister for digital affairs, Caroline Stage Olsen, puts it this way: “We will never reach our goal if we don’t start somewhere.” I call this a micropolitical choice, in the sense of Deleuze and Guattari, and I applied this concept to digital tools in the article Sovereign Digital, a Micropolitical Choice: an act that seems marginal but which, carried out by collectives that spend this money every year anyway, produces a multiplier effect, and public money then goes towards building common goods instead of vanishing into subscriptions.
Choices that are more political than technical
We still need to look at who makes these choices, and how. In most organisations, public and private alike, they are made by IT departments alone, and the same argument is often heard there: the tool everyone uses is the best choice. I am not putting these departments on trial; most of the time they do their best, pushed towards subscriptions by shrinking investment budgets (subscription is the only formula that counts as operating expenditure) and by rules decided above them. Yet this choice commits the sovereignty of the organisation, the law that applies to its data, the destination of its money: it should no longer fall to them alone. I propose treating it like any other budget, by submitting it to deliberation on its political stakes and not only its technical ones.
A town council can debate the renewal of its subscriptions before they expire, instead of renewing them by tacit agreement. A university can bring together teaching staff, administrative staff and students to think through free alternatives together. A public library can host a server running open artificial-intelligence models, such as those of Mistral, and make it an object of mediation, since libraries have always given access to what is too expensive for a single person. An association, a school, a business can bring their data back within their own walls: a storage server the size of a book, at modest cost (a NAS), holds files, calendars, contacts, phone backups, document sharing, even the website, within reach of non-technicians, provided you do your own backups and keep at it. Our data then physically come back home, like our archives of old: that is the condition on which we will still find them in twenty years’ time. And every file, every computation brought back home is one less to host and cool in a distant warehouse. These deliberations take trust, work and time, but they make these choices democratic, give IT departments the mandate they lacked, and can change things faster than one might think.
Equipping ourselves for collective intelligence
A town council debating its subscriptions, a university bringing together its teaching staff, administrative staff and students, a département organising a public consultation on twenty-seven data centres: in each case, no one around the table holds alone what needs to be known. The IT director knows the contracts, the councillor knows the budget, the resident knows the water table and the noise of the air conditioners, the lawyer knows the Cloud Act, and each of them is unaware of a good part of what the others know. These choices are therefore not the responsibility of each individual, but a collective responsibility. To exercise it, I would like to share here a way of working that I have practised for years: I run collective-intelligence processes for local authorities and cultural organisations, on digital technology and on artificial intelligence, for instance the AI monitoring committee of the Forum des images, whose findings I published in the article Webinar presenting the findings of the AI Monitoring Committee of the Forum des images, and I run them with sovereign digital tools, on free software. I have described this way of working in the articles Leading Collective Intelligence in the Face of Crises, Activating collective intelligence with digital tools and Cultivating empathy and controversy : my proposals for human governance of AI in cultural institutions. I am not saying this is how it must be done; these are ideas I have tested in other settings, and which I put forward here, applied to data centres, first of all to convey what collective intelligence is:
- Bring together people who do not usually talk to one another: elected officials, IT department staff, lawyers, residents, associations, local businesses, and the operators of the centres themselves; then agree together, from the first session, on the group’s operating rules, because a group keeps the rules it has made and endures those handed to it ready-made.
- Prepare each person beforehand, individually or in a small group, so that they know precisely what they want to say; then, in session, give short and equal speaking times, with a clock visible on the table.
- Take notes as a mind map projected on the screen while people speak, so that each person sees their words taken up and connected to those of others, and a shared line of thought takes shape before the group’s eyes.
- Write during the sessions and between them on a contribution platform running free software, hosted in the area, and create moments of collective emulation, even by video call, which is also a way of practising what is being discussed.
- Document disagreements instead of smoothing them over: a report that presents three well-argued contradictory positions, the operator’s, the neighbours’, the local authority’s, is worth more than a vague consensus in which no one recognises themselves.
- Make a common object that can be shown: the questionnaire to put to project developers, a map of the area’s data centres showing what each one computes and for whom, a specification for the subscriptions up for renewal. This object, made public, turns the group’s words into an act addressed to the local community.
An expert report is no substitute for this work, because this subject will never have a definitive answer. Artificial-intelligence models change every six months and computing needs with them, American and European laws change too, a centre announced for hosting websites may end up computing artificial intelligence, a moratorium will be circumvented by a new legal arrangement. We will therefore need others in order to understand what comes next, and a group that has learned to work together on this subject remains able to do so when the situation changes, which no report can guarantee.
I therefore support the pause and the consultation that Stéphane Troussel is calling for; the commission of inquiry I cited moreover proposes a moratorium on data-centre projects that do not meet an imperative of sovereignty. But the political question does not stop at where the buildings are sited: it bears on the choice of our tools, on which depends what these buildings will contain. This choice, no one can make alone, neither an IT director, nor a mayor, nor a user, and that is precisely what makes it political. Each of us can play a part, by switching to free software at home, by asking within our own organisation that these decisions be debated rather than endured, by pushing for public money to support digital commons and the associations that develop them. But each person’s share will not be enough: what it takes to move forward on data centres is collective intelligence, that is, groups that are set up, equipped and facilitated, in départements, municipalities, universities and businesses, which learn together to understand what is happening and to decide what to do about it, and which remain in place when the subject changes shape. The money it would take, we already spend every year on subscriptions to a few big capitalist monopolies.
