Small enough to say yes to

In Le Bourget, outside Paris, a proposed data center passed through two mayors and a municipal election in less than three months.

A building permit was issued on March 13, 2026. The newly elected mayor withdrew it on June 12. When the developer asked the administrative court to suspend that withdrawal, the court refused, finding that the project did not comply with height limits in the local planning rules. The court published its decision on July 28.

The legal issue was building height. The political dispute was much wider.

During the election campaign, residents and candidates raised concerns about noise, heat, land use, nearby schools, limited green space, and the number of lasting jobs the project might create. Reuters found candidates in at least 10 French towns campaigning against new data centers or calling for moratoriums and greater transparency. Its reporting from Le Bourget captured a tension now appearing across Europe: governments describe data centers as strategic infrastructure, while municipalities decide what that strategy will look like on the ground.

Demand for computing capacity is growing. Europe wants more infrastructure for cloud services, artificial intelligence, research, and public-sector systems. Yet the physical facilities needed to provide that capacity are becoming harder to treat as invisible background machinery.

Somebody has to host them.

And somebody has to say yes.

National strategy ends at a local address

A data center can be important to a national industrial strategy and still be difficult for a particular town to accept.

The benefits tend to be described at a broad scale: digital capacity, private investment, technological sovereignty, construction work, tax revenue, and access to AI infrastructure. The burdens arrive at an address.

Land is occupied in one municipality. Electricity is drawn from a particular part of the grid. Cooling equipment operates beside actual homes and workplaces. Water, roads, substations, transmission lines, and backup systems all have local conditions attached to them.

The UK Local Government Association describes a widening gap between national ambitions for AI and the “place-based realities” of delivering the infrastructure. Councils have to reconcile data-center proposals with housing, land use, water systems, grid capacity, environmental responsibilities, and public confidence. Its July 2026 response argues that local government cannot be treated as the administrative end of a decision made somewhere else.

The United States offers a warning about what happens when that gap grows. In a March 2026 Gallup survey, 71% of Americans opposed an AI data center being built in their local area, including 48% who strongly opposed it. Supporters most often cited economic benefits and jobs. Opponents raised energy and water use, pollution, noise, utility costs, traffic, land use, and quality of life. Gallup’s findings do not predict how every European community will respond, but they show how quickly infrastructure can lose legitimacy when its costs feel concentrated and its benefits remote.

These concerns should not be compressed into a familiar story about people resisting anything new.

A community may support data centers in principle and reject a particular site. It may welcome investment but dispute the proposed scale. It may accept the need for more computing infrastructure while objecting to the way one project uses power, water, land, or public money.

Water and power illustrate the same point from different directions: an industry-wide average cannot settle a local infrastructure decision.

The town is not being asked whether it approves of “the cloud.” It is being asked about a building, a power connection, a cooling system, a construction project, and a possible path to expansion.

That is the scale at which consent has to work.

Modularity changes the unit of commitment

“Modular data center” can describe several different things.

Some projects use prefabricated electrical or cooling systems inside a larger conventional facility. Others place the IT space, power distribution, cooling, controls, and supporting equipment inside integrated modules manufactured before they reach the site. Containerized data centers are one form of modular infrastructure, but the category is broader than a shipping container filled with servers.

The common principle is that more engineering, integration, assembly, and testing happens in a controlled manufacturing environment rather than entirely at the final location.

Uptime Institute traces prefabricated modular data centers back to telecommunications infrastructure in the 1980s and 1990s. The approach later moved into data-center electrical systems, technical spaces, and complete IT modules. Modularity is a construction and integration method, not one product shape.

Moving work into a factory can change the sequence of a project. A module can be built and tested while foundations, utilities, access, and other site work are prepared. Repeated designs can reduce the need to solve the same integration problem from the beginning at every location. Capacity can also be introduced in stages rather than requiring the eventual facility to be built before the first workload runs.

The site still matters. A modular unit needs foundations, power, network connectivity, transport and lifting access, security, drainage where applicable, permits, installation, commissioning, and testing. Modular construction moves and overlaps work; it does not erase it.

The important change is the unit of commitment.

A large campus can ask a community to accept years of construction and an eventual capacity several times larger than the first operating phase. A modular approach can allow a defined amount of infrastructure to be assessed first, with later additions considered as demand and local conditions develop.

That can make the decision more proportionate. It can also create a problem if the eventual scale remains hidden behind a series of smaller applications.

A first module may look modest while shared power, cooling, and network systems have been designed for a much larger build. Local authorities therefore need to understand the planned ceiling, the conditions for expansion, and which infrastructure is being installed for future phases.

Uptime Institute’s guidance on modular and phased construction makes a similar point from an operational perspective. Later phases have to be designed so they can be commissioned without compromising infrastructure already serving live workloads.

The legitimate advantage of modularity is proportionality, not concealment.

A smaller first decision should come with a clearer account of what may follow.

Smaller infrastructure still has neighbors

In June 2026, residents of Langenhorn in northern Germany attended a town hall meeting about a proposed 150 MW data center.

The project would occupy around 55,000 square meters, with two buildings reaching 27.5 meters and the possibility of increasing the site’s connection capacity later. Residents questioned the building height and the transparency of the planning process. The municipality postponed its decision. Data Center Dynamics reported that the developer had chosen the area partly for its renewable-energy resources, a national-scale advantage that did not settle the local questions.

The example warns against treating “modular” and “small” as synonyms.

A large campus can be assembled from modules. A compact deployment may contain highly integrated conventional systems. Modularity describes how infrastructure is organized and built; local impact still depends on total capacity, equipment, site, and operating conditions.

A smaller first deployment may use less land and require less on-site construction than a major campus. Several deployments can still create substantial cumulative demand across a region. They also create several local conversations.

Each site has its own grid connection, neighbors, roads, planning rules, water conditions, acoustic environment, and path to expansion. Distributed infrastructure spreads capacity, but it also spreads responsibility.

Noise makes this especially clear. A manufacturer may state how loud one machine is at a specified distance. A community experiences the combined sound of cooling, transformers, power equipment, vehicles, alarms, and other systems under actual conditions. Data-center noise has to be assessed where people hear it.

Site selection brings the same details together. Power availability can make a location attractive, but inadequate fiber, unsuitable land, difficult access, nearby housing, flood exposure, planning restrictions, or insufficient cooling options may make it unworkable. A viable data-center site is one whose constraints fit together.

A smaller project does not deserve less scrutiny because it is modular. It deserves scrutiny proportionate to what will actually be built.

Calling infrastructure “small” or “prefabricated” does not reassure a local authority given little information. Showing the physical unit, site requirements, operating conditions, expansion limit, and responsible parties gives the authority something it can assess.

The advantage is not invisibility.

It is legibility.

A clearer proposition for a real place

Policloud develops and deploys physical, modular data-center infrastructure for defined sites.

That creates a different proposition from asking a town or site owner to commit immediately to the eventual scale of a large campus. Capacity can be considered in units. More infrastructure can be integrated and tested before delivery. A first deployment can be evaluated against the power, network, land, workload, and operating model that exist now.

Future expansion still has to be planned honestly. The total possible scale matters, as do the shared systems installed at the beginning and the conditions under which more units may follow.

Modularity also does not justify claims such as “plug-and-play,” “deployable anywhere,” or “no permits.” Production, transport, site preparation, utilities, access, approvals, installation, commissioning, and testing remain part of the work.

The commercial benefit is narrower and stronger: a modular deployment can make the proposal easier to describe in physical terms.

What is being installed? How much capacity is included in this phase? Which resources does it require? What work will happen at the site? What operates continuously? How can the deployment grow? Who owns and operates each part? What approval would later expansion require?

Those questions turn an infrastructure ambition into a decision.

Some sites will not fit. Power may be insufficient. The network may be inadequate. Noise or visual effects may be unacceptable. Planning rules may rule the project out. A community may decide another use of the land matters more.

A credible modular model has to leave room for that answer.

The lesson from Le Bourget is not that local opposition will defeat every large data center. The court’s decision rested on a specific planning violation, while the public debate involved a wider set of questions.

The lesson is that the building eventually becomes more real than the strategy used to justify it.

National governments may speak about AI capacity, investment, competitiveness, and sovereignty. A municipality still has to decide where the walls go, how high they rise, what connects to them, and how the place around them changes.

A modular data center still has to earn a yes.

Its advantage is that it can give people a clearer thing to say yes to.