Summary

• USD.AI arranged a $106 million senior secured facility with Silver Point Capital and Victory Park Capital.
• The financing covers 216 NVIDIA HGX B300 servers containing 1,728 GPUs in Finland.
• Shipment and associated funding are expected in the fourth quarter, while the four-year loan follows contracted inference demand.

British AI infrastructure provider NexGen Cloud has secured a $106 million lending facility to finance GPU computing capacity in Finland, using contracted demand for AI inference services to support investment in new hardware.

The four-year senior secured loan was arranged by USD.AI, with Silver Point Capital and Victory Park Capital participating as lenders. It covers 216 NVIDIA HGX B300 servers containing 1,728 graphics processing units, which are intended to operate at a Tier III data centre in Finland.

The facility was closed on 29 October, but the servers have not yet been delivered. Funding from the escrow arrangement is aligned with equipment shipment, currently expected during the fourth quarter of 2026, while the computing capacity has been contracted to an unnamed AI inference platform.

The customer agreement follows a take-or-pay structure, under which contracted payments support the facility over its four-year term. Loan repayments are designed to amortise over the same period, connecting the financing obligations with expected revenue from the computing service.

Hardware investment follows contracted demand

Building AI computing capacity requires substantial expenditure before equipment begins generating income. Operators must acquire processors, servers, networking equipment, and supporting infrastructure, then complete installation and testing before the resulting service becomes available to customers.

For companies outside the largest hyperscale cloud providers, financing that expenditure can restrict expansion even when demand for computing is already established. Purchasing the equipment outright requires significant capital, while ordinary corporate borrowing may not reflect the value of the assets or the income expected from a particular deployment.

The NexGen Cloud transaction uses a different approach by linking secured lending to the hardware and a contract covering its future use. The underlying computing equipment provides collateral, while the customer agreement offers a defined source of expected payments.

A take-or-pay arrangement requires the customer to pay for contracted capacity regardless of whether it uses every available computing resource. That can create greater revenue certainty for the infrastructure provider, particularly where services are designed around a customer requiring dedicated capacity over several years.

However, the arrangement remains dependent on the equipment being delivered and made operational. The commercial outcome also depends on the customer’s ability to meet its contractual obligations and the operator’s capacity to maintain the service.

The interest rate and detailed financial conditions of the loan have not been disclosed, making it impossible to establish the complete cost of borrowing or compare the transaction precisely with other forms of infrastructure finance.

A specialised market for GPU-backed lending

USD.AI arranged the facility through its GPU Loan Exchange, known as GLX, which connects infrastructure operators seeking finance with institutional investors prepared to lend against computing assets.

The platform was launched on 1 October and is designed to originate, service, and place loans supported by GPU hardware. Its role in the NexGen Cloud transaction included arranging the financing and acting as administrative agent for the lending group.

Silver Point Capital and Victory Park Capital’s involvement brings institutional private credit into a transaction focused on specialised computing equipment. This approach is intended to give operators below the largest cloud companies another route to finance hardware that has already attracted commercial demand.

USD.AI says it has originated more than $500 million of GPU financing during 2026, with more than $200 million arranged through GLX following the new facility. These are company-reported figures covering the lender’s wider activity rather than the value of equipment deployed in Finland.

NexGen Cloud has previously used the same financing provider. A $34 million facility agreed in June supported NVIDIA B200 infrastructure in Sweden, bringing the combined value of the two arrangements to $140 million across the Nordic markets.

The transactions reflect a shift towards financing individual computing deployments through identifiable assets and commercial contracts, rather than relying solely on a provider’s balance sheet or successive equity investment rounds.

Equipment value creates financing challenges

GPU servers differ from many conventional industrial assets because their commercial value can change quickly as new processor generations enter the market. A machine may continue functioning reliably while demand for its capabilities changes relative to newer systems.

That creates challenges for lenders assessing the value of collateral over several years. The expected resale value of processors, their energy consumption, compatibility with current AI software, and demand for particular workloads can all influence the economics of an investment.

Contracted revenue can help support the financing case because the equipment is associated with an identified commercial service. However, the duration and payment obligations of the contract need to be considered alongside the possibility of hardware failures, service interruptions, and changes in operating costs.

Aligning repayments with the customer contract reduces a potential mismatch between the periods in which revenue is earned and the debt must be serviced. The financial structure does not establish that the equipment will retain its original market value throughout that period.

Maintaining computing performance also requires expenditure beyond the purchase of GPUs. Cooling systems, networking, electricity, maintenance, and operational support contribute to the cost of running the infrastructure, while customers may require contractual performance and availability commitments.

Inference demand supports Nordic expansion

The Finnish deployment is intended to provide inference capacity, which supports the operation of AI models after their initial training. Inference workloads generate responses to requests and may operate continuously as businesses incorporate AI into customer services, internal applications, and other software.

That demand creates a market for infrastructure with predictable performance and capacity. Some customers may prefer dedicated computing arrangements when their workloads are substantial or require particular performance characteristics, rather than competing for resources through shared services.

Finland provides a location for NexGen Cloud’s further European expansion, following the earlier financing associated with its Swedish deployment. Individual infrastructure projects must still obtain suitable electricity, cooling, telecommunications connectivity, and operating arrangements.

Customers also need to consider where their data is processed and what contractual protections apply to their computing services. The location of a facility can influence those decisions, although operating within Europe does not itself establish that every sovereignty or regulatory requirement has been satisfied.

The planned 1,728-GPU installation is a defined financing transaction rather than confirmation of the company’s entire future hardware capacity. Other expansion plans should be assessed separately according to their own funding, delivery, and operating milestones.

NexGen Cloud’s immediate next step is the shipment and installation of the financed servers. The facility will support a contracted inference service once the necessary equipment is deployed and operational, with payments from that service intended to support the four-year lending arrangement.