Via robinhood.com
The Amsterdam-based AI cloud provider is pouring $2.26 billion into UK operations as part of a broader push to build out NVIDIA-powered computing capacity across Europe.
Nebius Group N.V. is setting up shop inside one of Europe’s largest data center campuses, announcing an agreement with Vantage Data Centers to deploy NVIDIA-powered AI infrastructure at the Newport facility in Wales.
The move is part of a roughly £1.7 billion ($2.26 billion) commitment to establish 65 MW of NVIDIA infrastructure across four UK sites by 2027. For a company that went public after spinning out of Yandex’s international operations, that’s a serious bet on Britain becoming a hub for AI compute.
What Nebius is actually building #
The Newport data center sits within the Cardiff CWL1 campus, originally developed by Next Generation Data. The facility can support approximately 150 MW of critical IT load, with two buildings already operational and a third in the planning stages.
It runs on 100% renewable energy and has its own dedicated 400 kV substation.
Nebius isn’t buying or building the physical infrastructure itself. Instead, the company is leaning into what it calls an “asset-light partnership model,” where partners like Vantage handle the bricks, concrete, and power distribution while Nebius manages the software stack and customer integration layer on top.
The Newport deployment adds to Nebius’s existing UK footprint, which includes prior infrastructure at Kao Data’s facility in Harlow. Together, these sites form part of what the company describes as a growing network of “AI factories” designed to serve enterprise clients who need serious GPU compute without wanting to build their own.
The NVIDIA connection runs deep #
Nebius announced a $2 billion investment partnership with NVIDIA back in March 2026. The infrastructure going into Newport is NVIDIA-powered, which in practical terms means racks of GPU clusters optimized for the kind of large-scale model training and inference workloads that have become the backbone of enterprise AI adoption.
Nebius has also cultivated relationships with Microsoft and Meta, suggesting the company is building a customer base among hyperscalers and large enterprises rather than chasing smaller startups.
Why Wales, and why now #
Power availability is the big one. AI workloads are extraordinarily energy-hungry, and finding locations with both sufficient grid capacity and renewable energy sourcing has become one of the primary bottlenecks for infrastructure expansion globally. Newport’s 150 MW capacity and dedicated substation solve that problem in a way that many urban data center markets simply cannot.
Geographic proximity to London matters too. Most of the UK’s financial services, technology, and enterprise decision-making happens within the M4 corridor or in the capital itself. Newport sits close enough to serve low-latency workloads for London-based clients while benefiting from lower land and operating costs.
What this means for the AI infrastructure race #
Nebius’s expansion reinforces a trend that has been reshaping the cloud computing landscape: the unbundling of infrastructure ownership from infrastructure operation. Nebius is betting that a lighter model, where it layers proprietary software and AI optimization on top of third-party facilities, can compete effectively while preserving capital.
The approach allows faster geographic expansion, since finding a partner with an existing facility is quicker than building from scratch. It also shifts the capital expenditure burden to companies like Vantage that specialize in financing and operating physical infrastructure.
But it also introduces dependencies. Nebius doesn’t control its own power contracts, cooling systems, or physical security at these sites. If Vantage raises prices or reallocates capacity, Nebius has fewer levers to pull than a company that owns its buildings outright.
Nebius trades on Nasdaq under the ticker NBIS.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our