Summary
- Nvidia is reportedly in talks to provide a roughly $250 billion financing guarantee linked to OpenAI data centre capacity.
- The structure would support a proposed 10GW data centre project in southern Ohio being developed by SoftBank’s energy subsidiary.
- The talks show how AI infrastructure financing is moving beyond conventional data centre leasing and into energy-scale capital structures.
Nvidia is reportedly in talks to provide a financing guarantee of roughly $250 billion linked to OpenAI data centre capacity, a structure that would pull chip supply, tenant credit, power development, and project finance into the same industrial frame.
The reported guarantee would support OpenAI’s leasing of a 10GW data centre project in southern Ohio being developed by SoftBank’s energy subsidiary. The total cost of the project, including chips, could exceed $500 billion.
The talks remain reported discussions rather than a completed financing package. Even so, the scale points to a market where the largest AI data centre projects are beginning to stretch beyond conventional real estate finance and into structures closer to energy megaprojects.
The tenant is not the only risk
Traditional data centre finance rests on a familiar set of inputs: land, powered shell, operator capability, lease terms, tenant credit, and lender appetite for long-life infrastructure. AI campuses alter that balance because chips, power, cooling, and grid infrastructure can dominate the cost and execution profile.
A 10GW project cannot be understood as a larger version of a standard cloud campus. It needs access to generation, transmission capacity, substations, transformers, water or low-water cooling strategies, backup systems, internal roads, construction staging, and specialist labour across several years. The hardware refresh cycle also creates a different risk profile from buildings alone.
If a chip supplier guarantees a large share of financing, it becomes more deeply exposed to the physical capacity pipeline that drives demand for its products. That could help move projects through capital markets, especially where the tenant’s future cash flows, hardware supply, and power availability are tightly connected. It also concentrates risk across a smaller group of AI companies, suppliers, and investors.
The arrangement would underline how AI infrastructure is no longer separated into neat vendor, developer, customer, and lender roles. The entities providing accelerators, consuming compute, developing energy-backed campuses, and arranging capital are increasingly tied together by the same delivery bottlenecks.
Europe has a different delivery environment
The reported Ohio project sits outside Europe, but the financing pressure it illustrates will shape European competition for AI capacity. If US projects can organise chip supply, energy development, and financing at multi-gigawatt scale, European markets will need credible alternatives for power-backed compute or risk losing the largest deployments to regions with faster permitting and larger power corridors.
Europe’s challenge is not only the cost of capital. The region has planning constraints, stronger sustainability reporting, stricter public scrutiny in many markets, and grid queues that can slow even well-funded schemes. Ireland, the UK, the Netherlands, Germany, northern Italy, and parts of France all show different versions of the same problem: demand is moving faster than the infrastructure that permits capacity to be built.
European lenders and infrastructure investors may also need to adapt. AI data centre projects can carry tenant concentration, technology-refresh risk, power price exposure, and equipment supply risk at levels that conventional colocation portfolios did not. Financing structures may increasingly need anchor customers, supplier support, government coordination, or dedicated energy arrangements to reach scale.
The reported Nvidia-OpenAI talks show how far the market is moving from the language of cloud expansion. AI capacity is being treated as strategic industrial infrastructure, with capital structures to match. The companies able to align power, hardware, site delivery, and long-term credit will move faster than those relying on lease agreements and grid applications alone.
