Nebius Announces £1.7B UK Nvidia Infrastructure Expansion
Nebius announced it will invest about GBP 1.7 billion to build three new Nvidia-powered AI infrastructure sites in the U.K., per the company's newsroom and a BusinessWire release. Nebius says the three new deployments, together with its first U.K. site launched in November 2025, will reach 65 MW of capacity when fully ramped in 2027, using Nvidia's latest full-stack AI platform and building on an earlier Blackwell Ultra deployment. The company says customers including Revolut and AI biotech startup Prima Mente already run production workloads on the platform, and reporting from Data Center Dynamics notes a 10-year, 22 MW lease at Kao Data's Harlow campus. Nebius frames the build-out as adding domestic compute aligned with the U.K. Government's AI Opportunities Action Plan. Market coverage reported Nebius shares rose about 4% premarket and Nvidia about 2% on the news, per Seeking Alpha.
What happened
Nebius announced an investment of approximately GBP 1.7 billion to deploy three new Nvidia-powered AI infrastructure sites in the U.K., per the company's newsroom and a BusinessWire release. Nebius says the three new deployments, combined with its first U.K. site launched in November 2025, will reach 65 MW of capacity when fully ramped in 2027, using Nvidia's latest full-stack AI platform and building on an earlier deployment of Nvidia Blackwell Ultra. Reporting from Data Center Dynamics adds that Nebius signed a 10-year lease for 22 MW at Kao Data's Harlow campus as part of the plan.
Editorial analysis - technical context
Industry context
regional GPU build-outs typically cite two drivers, access to lower-latency high-throughput clusters and control over power and cooling at scale. An added 65 MW of onshore Nvidia capacity materially increases regional headroom for large-model training and inference, easing queueing pressure on hyperscale clouds. Deployments described as Nvidia full-stack combine GPU compute, networking fabrics, and software components that together shape throughput, model-parallelism choices, and cost-per-token.
Nebius frames the expansion as adding domestic compute aligned with the U.K. Government's AI Opportunities Action Plan, and names customers including Revolut and AI biotech startup Prima Mente as already running production workloads, per the company. More regional capacity tends to widen options for enterprises and research groups with data-residency, latency, or regulatory constraints, and for customers seeking non-U.S. cloud footprints.
What to watch
- •Service offerings, SLAs, networking options, and managed MLOps tooling as the new sites come online, since those will drive commercial differentiation.
- •Power ramp timelines against the 2027 target, because grid connection and power availability are common bottlenecks for large GPU sites.
- •U.K. hiring and partnerships, which affect the local ecosystem's capacity for managed services and enterprise integrations.
Bottom line
This is a notable regional infrastructure expansion that increases U.K. onshore GPU capacity and may relieve some demand pressure for regional customers and research teams. Key questions are execution on power ramp timing, commercial packaging of Nvidia hardware and software, and how quickly the new capacity is absorbed.
Key Points
- 1Nebius committed about GBP 1.7 billion to three new Nvidia-powered U.K. sites, reaching 65 MW of capacity when fully ramped in 2027, per the company and BusinessWire.
- 2The expansion adds onshore U.K. GPU capacity, including a 10-year 22 MW lease at Kao Data's Harlow campus (per Data Center Dynamics), easing data-residency and latency constraints for regional AI workloads.
- 3Nebius frames the build-out as aligned with the U.K. AI Opportunities Action Plan, citing customers such as Revolut and Prima Mente as already running production workloads.
Scoring Rationale
A substantial, region-scale GPU expansion backed by the company's own announcement and a BusinessWire release increases onshore U.K. compute and matters to practitioners planning training and deployment footprints. It is a notable single-company infrastructure commitment rather than a frontier-model or regulatory event, placing it in the notable band.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems