Nebius to invest about £1.7 billion in UK AI infrastructure with new NVIDIA-backed deployments

Nebius, a specialist AI cloud company, said on Monday it plans to invest approximately £1.7 billion to expand its United Kingdom capacity, placing three new NVIDIA-powered infrastructure deployments into buildout. The company framed the move as a way to bring additional compute availability to UK customers and to extend the cloud capabilities it offers for agentic AI and enterprise AI use cases.

Nebius positions itself as more than a GPU reseller. In its public filings, the company describes a full-stack AI cloud platform that runs across the hardware and software stack, including data centers, in-house infrastructure, and software tools designed to support AI workflows from data preparation and model training to inference at scale. That approach is meant to let customers access specialized compute without having to operate the underlying systems themselves.

The UK expansion follows earlier steps by Nebius to establish a local infrastructure footprint. In June 2025, the company announced a UK deployment built around NVIDIA Blackwell Ultra GPUs, saying it would enhance domestic compute capacity and support work by UK firms, researchers, and public services. At the time, Nebius also emphasized the operational concept of combining “supercomputer power” with a hyperscaler-like experience, and it tied the effort to the UK’s broader push to grow domestic AI compute capability.

Agentic AI refers to systems that can take actions toward goals, often by planning steps, using tools, and iterating based on feedback, rather than only producing a single static response. NVIDIA has recently argued that agentic AI is driving a surge in compute needs and, in a separate partnership update involving Nebius, described Nebius’s approach as an AI cloud designed for the agentic era, with the company’s platform integrated across silicon-to-software and powered by NVIDIA’s accelerated computing roadmap.

For enterprises, the compute requirement is not only about training new models, but also about running inference, which is the ongoing process of generating outputs from deployed models inside production workflows. Nebius’s filings highlight that it expects inference compute requirements from enterprise customers to grow as more AI models move into practical, cost-effective deployments. The company also describes serving multiple customer categories, including enterprises, software vendors, AI-native tech companies, and research labs, with dedicated sales and customer success teams to support deployments.

In practical terms, adding more onshore UK capacity can reduce friction for customers that prefer local data handling, lower network latency, or tighter control over operational dependencies. It also deepens competition among “neocloud” providers, which aim to differentiate from general-purpose public clouds by focusing specifically on AI infrastructure and deployment support.

Nebius did not provide further specifics in the announcement details available for this report, including the exact locations of the three deployments, the scale in GPU terms, or the precise timelines for commissioning and customer availability. Investors and customers will likely look for those operational clarifications, along with any additional information on how the new infrastructure ties into Nebius’s existing UK offerings and enterprise service roadmap.