British AI infrastructure firm Nscale has signed a landmark deal with Microsoft to supply around 200,000 Nvidia AI chips, in one of the largest AI infrastructure agreements of 2025. Valued at up to $14 billion, the multi-year contract marks a major step in scaling global AI compute capacity amid soaring demand for machine learning and generative AI services.
Under the agreement, Microsoft will receive 104,000 Nvidia GB300 GPUs at Nscale’s Texas data centre and 12,600 at the Start Campus in Portugal, with deliveries beginning next year. Dell Technologies will assist in deploying the infrastructure, building on an earlier $6.2 billion contract with Nscale to supply 52,000 GPUs from its Norway-based AI campus, developed in partnership with Aker.
Nscale, which emerged from an Australian Bitcoin mining spin-out just over a year ago, has rapidly positioned itself as a leading AI infrastructure provider. It raised $1.1 billion last September from investors including Aker, Nokia, Nvidia, Dell and Fidelity, reaching a valuation of $3 billion. The company’s ambition to become a national AI infrastructure champion for the UK is now underpinned by plans to invest up to £2.5 billion in domestic data centre development.
A key part of that expansion is Nscale’s new greenfield site in Loughton, Essex. The facility is set to deliver 50 megawatts of high-performance computing capacity, scalable to 90 megawatts. Construction is expected to complete by the end of 2026, with tens of thousands of GPUs to be installed. Chief executive Josh Payne has outlined plans for a potential public listing by late 2026, subject to progress on construction and chip delivery milestones.
Yet challenges remain. Hyperscale data centre development must navigate complex regulatory environments, environmental standards, and power constraints. AI workloads are energy intensive, and Nscale’s ability to secure stable, green power sources will be closely scrutinised. Logistics and chip supply also pose risks, with global shipping delays or Nvidia capacity constraints potentially impacting delivery schedules.
The deal reflects how technology giants such as Microsoft are scaling their physical infrastructure to meet AI demand. Microsoft’s recent launch of a supercomputer-scale GB300 NVL72 cluster using 4,608 Nvidia GPUs on its Azure platform shows the scale of investment underway. The system delivers 92.1 exaFLOPS of inference performance—significantly reducing AI training times.
Alongside internal build-outs, Microsoft is leasing AI compute from multiple partners—including Nscale, CoreWeave, Nebius and Lambda—bringing its total access to more than 100,000 GB300 GPUs as part of a broader $33 billion cloud infrastructure push. While this helps meet near-term demand, it has raised concerns over energy usage, grid pressure and regulatory oversight, particularly regarding the company’s ties with Nvidia and OpenAI.
Nscale’s rapid evolution from crypto mining to AI infrastructure highlights the strategic role compute capacity plays in the next wave of innovation. With strong investor backing and a UK expansion grounded in sustainability, the firm is well placed to support the UK’s ambitions as a global leader in responsible AI development.
Created by Amplify: AI-augmented, human-curated content.
