Summary
- Rexecode’s central scenario sees French installed data centre capacity increase tenfold between 2025 and 2035.
- Cumulative investment could reach approximately €210 billion as AI and cloud projects move into industrial-scale construction.
- Grid reinforcement, imported equipment, local opposition, water, and heat reuse will determine the economic and environmental outcome.
French data centre capacity could increase tenfold between 2025 and 2035 under a central scenario published by Rexecode, requiring approximately €210 billion of cumulative investment.
The economic research institute expects France to capture a larger share of cloud and AI infrastructure through its relatively low-carbon electricity system, central European position, fibre connectivity, and domestic market. The projected expansion would move data centres further into the territory of national energy, industrial, and land-use policy.
Annual electricity consumption, currently estimated at about 10TWh, could rise to approximately 15TWh–20TWh in 2030 and 23TWh–28TWh by 2035, depending on construction, hardware deployment, utilisation, and operating efficiency.
Construction reaches industrial scale
A tenfold increase would require far more than additional server capacity. New substations, transmission and distribution reinforcement, generators, UPS systems, chillers, liquid-cooling plant, steel, concrete, fibre, security infrastructure, and specialist labour would all be needed across a large national programme.
The €210 billion estimate covers a period in which data centre design is also becoming more capital-intensive. AI clusters carry greater rack densities, larger electrical blocks, and more direct liquid cooling than conventional enterprise and cloud halls. Efficiency gains at server or cooling-system level do not necessarily reduce the capital needed to deliver each energised megawatt.
France has attracted several proposals measured in hundreds of megawatts or gigawatts, although announced capacity is not equivalent to operating infrastructure. Grid studies, planning, environmental review, equipment supply, financing, and customer demand will remove or delay part of the pipeline before construction.
The Rexecode study sets out its investment and installed-capacity scenario.
Low-carbon electricity still needs a grid
France’s nuclear and renewable generation gives new data centres a lower operational-carbon baseline than facilities in many competing markets. Customers seeking lower-carbon computing can benefit from that electricity mix, provided sufficient power is deliverable at the chosen location.
National generation adequacy does not remove regional network limits. Several hundred megawatts of concentrated load may require new substations, cables, transformers, and transmission works whose programme extends well beyond construction of the data halls.
Electricity demand from data centres will also grow alongside the electrification of transport, heating, chemicals, steel, and other industries. France may have sufficient generation over the year while still encountering local congestion, winter peaks, maintenance periods, or delayed reinforcement.
Connection agreements may therefore include phased capacity, operational limits, storage, flexible demand, or other conditions. The most viable campuses will align construction with a realistic network programme rather than assuming that the final announced load can be delivered immediately.
Cooling and water will shape local acceptance. Direct liquid cooling removes heat from processors more efficiently than room air, but the energy still passes into chillers, dry coolers, cooling towers, or hybrid systems. Water use, fan noise, plume, and external plant size depend on climate, temperatures, and redundancy design.
Waste-heat recovery offers one route to local benefit, particularly near district-heating networks or industrial processes. Viable projects require nearby demand, compatible temperatures, pipe infrastructure, commercial agreements, and a party willing to fund and operate the connection.
Rexecode estimates that only about 30% of the direct value created may remain in France because much of the computing, networking, electrical, and cooling equipment is imported. Construction, energy supply, maintenance, engineering, and operations will create domestic activity, but the largest equipment purchases often flow through global manufacturing chains.
The economic return can improve where French and European suppliers participate in power distribution, cooling, automation, heat recovery, construction, and facility services. A large hosting footprint alone does not create a complete domestic AI supply chain, particularly while processors, memory, and cloud platforms remain concentrated among overseas suppliers.
Employment claims are likely to receive close examination. Data centres generate substantial temporary construction work and relatively small permanent operating teams. Local authorities will compare jobs and tax income with the land, power, water, road, and network capacity required by each project.
Opposition has already formed around several French developments, focusing on power allocation, water use, diesel generation, land consumption, and limited long-term employment. Accelerated national procedures may shorten parts of the approval process, but they do not remove the need for credible site selection and measurable local benefits.
France has the electricity mix, market, and policy ambition to host a larger share of European infrastructure. Reaching Rexecode’s central scenario would require grids, construction capability, equipment manufacturing, planning, and local consent to expand alongside the data centres themselves.
