The $105 billion credit and compute agreement
A securities filing revealed on 17 August 2026 that Nvidia will provide up to $105 billion in financing for a new OpenAI data center in Ohio. The credit supports an initial 4.25 gigawatts of computing capacity, with an option for an additional 3.75 gigawatts. This arrangement designates Nvidia as the exclusive chip supplier for the facility, intertwining hardware provision with direct capital deployment on an unprecedented scale within the artificial intelligence sector.
Capacity is expected to come online in phases starting in 2028. Nvidia will supply the compute hardware, effectively tying the financial loan to its own product revenue stream. This model suggests that the $105 billion commitment functions simultaneously as infrastructure financing and a guaranteed future sales channel. The arrangement positions Nvidia not merely as a component vendor but as a primary creditor underwriting the physical expansion of its largest downstream customer.
The deal represents Nvidia's latest move to support the sprawling artificial intelligence data center buildout. By supplying both the capital and the physical compute infrastructure, Nvidia absorbs the financial risk of the facility's construction while securing long-term demand for its hardware. This vertical integration of finance and silicon supply indicates a structural shift in how major technology acquisitions are funded, moving away from traditional institutional debt toward vendor-backed capital allocation.
SB Energy facility scale and 20-year lease terms
SoftBank subsidiary SB Energy will build and manage the data center at the PORTS-Pike Technology Campus in Pike County, Ohio. OpenAI will lease the facility under a 20-year agreement, locking in infrastructure costs over a two-decade horizon. The site will ultimately house around 8 gigawatts of IT capacity, making it one of the largest single-site artificial intelligence infrastructure projects currently documented in public real estate and technology sector filings.
The 20-year lease term provides OpenAI with long-term operational stability while transferring construction and management execution to SB Energy. This division of labor separates physical real estate development from artificial intelligence model training. The fixed-duration commitment also suggests that OpenAI is prioritizing predictable infrastructure costs over the volatility of short-term cloud computing contracts, hedging against future compute scarcity by securing physical capacity decades in advance.
Nvidia CEO Jensen Huang stated that land, power, and building shells now constitute the biggest bottlenecks in the AI buildout. The Ohio project directly addresses these constraints by delegating site development and power procurement to SB Energy. By financing the compute hardware separately, Nvidia circumvents the physical real estate bottlenecks Huang identified, allowing OpenAI to secure both the physical shell and the internal compute capacity through two distinct, specialized corporate partnerships.
Off-balance-sheet AI infrastructure commitments
The Wall Street Journal reports that major technology companies currently hold around $3 trillion in artificial intelligence commitments off their balance sheets. Analysts warn that investors can barely gauge these companies' actual debt levels anymore. The Nvidia-OpenAI financing structure exemplifies this trend, as the $105 billion credit arrangement represents a massive liability that may not appear as traditional debt on the primary tenant's financial statements, obscuring the true cost of infrastructure expansion.
This off-balance-sheet positioning allows companies like OpenAI to rapidly scale physical infrastructure without immediately impacting their debt-to-equity ratios. However, it creates a systemic opacity where the actual financial exposure of artificial intelligence developers remains hidden from standard equity analysis. The reliance on vendor financing from entities like Nvidia further complicates risk assessment, as the compute hardware serves as both collateral and operational necessity, blurring the line between capital expenditure and operational leasing.
The open question is whether these off-balance-sheet commitments will eventually require reclassification under stricter accounting standards. If analysts cannot accurately gauge debt levels, the $3 trillion figure suggests a systemic underpricing of risk across the technology sector. Should compute demand falter, the intertwined nature of vendor credit and hardware supply means losses would cascade directly back to Nvidia, transforming an artificial intelligence buildout bottleneck into a balance sheet liability for the chipmaker itself.
Sources
- cnbc.com — Nvidia Backs $105B Ohio Data Center for OpenAI in Record Infrastructure Deal
- the-decoder.com — OpenAI signs record Ohio data center lease with Nvidia backing up to $105 billion (2026-08-17)
- datacenterknowledge.com — Nvidia Backs OpenAI’s Ohio Data Center Buildout With $105B (2026-08-20)
- dealroom.co — Nvidia backs OpenAI's Ohio data center with up to $105B ... (2026-08-18)
- New Horizon daily digest, 2026-08-23 (topic provenance)