London-based Nscale has secured a $1.4 billion Delayed Draw Term Loan backed by GPUs. The financing is led by funds managed by PIMCO, Blue Owl, and LuminArx Capital Management, with backing from additional asset managers and banks. In September last year, Nscale closed the largest Series B round in European history, securing $1.1 billion.
The capital will be deployed to purchase GPU systems tied to multiple signed customer contracts. Rather than speculative expansion, the facility’s design ensures that infrastructure investment aligns directly with secured demand.
Goldman Sachs & Co. LLC acted as Sole Structuring Agent and Sole Placement Agent for the transaction, structuring a debt facility that links capital deployment to GPU-backed assets.
Building a vertically integrated AI platform
Headquartered in Europe and operating globally, Nscale positions itself as an AI-native hyperscaler purpose-built for enterprise workloads. Its offering extends beyond raw compute. The company integrates networking, storage, managed software, and AI services within both owned and collocated data centres.
This vertical integration allows Nscale to control performance, efficiency, and service delivery fromthe infrastructure to the application layer. Enterprises running generative AI training, fine-tuning, and inferencing workloads gain access to tightly optimised compute environments rather than fragmented infrastructure stacks.
With enterprise AI adoption accelerating across industries, hyperscale GPU clusters are becoming critical infrastructure. Nscale’s model is built specifically to meet that need at scale.
Financing growth without slowing momentum
The GPU-backed Delayed Draw Term Loan enables Nscale to use structured debt to finance a portion of capital expenditure associated with large-scale GPU clusters. The facility supports both infrastructure tied to executed contracts and additional liquidity for clusters in its customer pipeline.
This balanced capital structure allows the company to scale without diluting equity while maintaining financial flexibility. It also signals institutional confidence in GPU assets as collateral in the emerging AI compute economy.
The announcement follows a year of notable milestones. Nscale secured contracts for multiple large-scale compute clusters worldwide, expanded its leadership team, and acquired Future-tech, a European data centre engineering consultancy. The acquisition strengthens its in-house engineering capabilities as it continues expanding its global footprint.
With GW+ greenfield data centre capacity under development and vertically integrated AI solutions already in operation, Nscale is positioning itself as a global hyperscaler engineered specifically for enterprise-grade AI.
As demand for generative AI infrastructure intensifies, the company’s combination of GPU-backed financing, renewable energy strategy, and end-to-end service model places it at the centre of the next phase of AI infrastructure growth.
Renewable energy as a competitive edge
A defining feature of Nscale’s expansion strategy lies in the geographic positioning of its data centres. The company operates in locations that provide access to some of the lowest-cost renewable energy sources globally.
This approach delivers two strategic advantages. First, it reduces operational costs, allowing Nscale to pass savings directly to customers. Second, it supports compliance with stringent regulatory requirements increasingly shaping enterprise AI deployment in Europe and beyond.
As energy consumption becomes a central concern in large-scale AI training, cost-efficient green power is a competitive differentiator. Nscale’s infrastructure strategy reflects that shift.
Josh Payne, Founder and CEO of Nscale, said, “We’re seeing massive demand for AI infrastructure to support the needs of businesses and consumers. This GPU debt financing is a key step in meeting that demand – backing infrastructure that can be delivered faster and more cost-effectively than industry norms, whether that’s large-scale hubs in Norway to smaller metro clusters built for low-latency workloads.”
