Nebius to build 240MW AI-focused data centre near Lille, France

Nebius plans to redevelop a former Bridgestone tyre plant in Béthune, near Lille, into a roughly 240 megawatt, GPU-dense data centre campus aimed at AI training and inference workloads. Delivery will be phased: some capacity is expected by late summer and the company targets about 50% of the campus operational by the end of 2026. The site follows Nebius’s recent commercial expansion supplying large U.S. cloud customers and is intended to host concentrated GPU and CPU clusters for large-model compute. Building to this scale requires heavy upfront capital for land remediation, civil works, power hookups, redundant cooling and high-density IT infrastructure; when aggregated these items point to a multibillion-euro program. The Béthune brownfield should simplify some planning steps compared with greenfield builds, but integrating high rack density, heat rejection and N+1 redundancy across a former industrial site remains technically and logistically demanding. The announcement arrives amid two parallel market dynamics: hyperscalers increasingly choose to own and site GPU-heavy campuses to control energy and interconnection, while specialist developers and financiers are mobilising purpose-built capital and credit structures to underwrite third-party AI capacity. Those market responses — from corporate balance-sheet builds to private-credit facilities and leasing models — will shape how quickly projects like Béthune can scale and the commercial terms they offer customers. Immediate bottlenecks are expected in GPU procurement, skilled electrical and mechanical contracting, and grid interconnection timelines; local transmission upgrades or dedicated substations may be required to meet peak demand. Environmental and planning scrutiny will focus on water use for cooling, noise, and the carbon profile of electricity sourcing, which could influence permitting and the mix of onsite versus grid-based generation. Economically, the project should generate construction jobs and some operational roles, while potentially attracting AI software and services firms seeking lower-latency, on-continent compute. For regional planners and utilities, Béthune will be another test case in coordinating rapid AI-capacity expansion with transmission planning and renewable procurement strategies. Commercial risk is partially mitigated by Nebius’s existing anchor contracts, but overall execution depends on aligning grid upgrades, supply chains and any financing constraints that affect deployment pacing.