Nvidia is providing a $105 billion financial guarantee to underwrite OpenAI Ohio data center buildout, structuring the support around long-lived infrastructure and the residual value of the deployed hardware.
Nvidia is providing up to $105 billion in residual-value guarantees for OpenAI’s planned Ohio data center campus, positioning the chipmaker as a financial backer of the AI infrastructure buildout rather than merely a hardware supplier. The guarantees are explicitly tied to the long useful life of the facilities and the residual value of the computing hardware operating within them.
The commitments cover an initial 4.25 gigawatts (GW) of IT capacity at SB Energy’s PORTS-Pike Technology Campus in Pike County, Ohio. Nvidia holds an option to extend support for an additional 3.75 GW, which would bring the campus to approximately 8 GW of total IT capacity. Under a 20-year lease agreement, SB Energy will construct, own, and operate the facility for OpenAI.
Nvidia disclosed the arrangement in an August 17 Securities and Exchange Commission filing. While its aggregate payment obligation under the initial guarantees is capped at $105 billion, the company emphasized that this figure represents a contingent liability rather than a direct capital investment. Should specified default or insolvency events occur, Nvidia may be required to cover the shortfall between the guaranteed minimum lease values and the amounts SB Energy recovers through re-leasing or selling the assets. OpenAI has agreed to fully reimburse and indemnify Nvidia for any payments made under these guarantees. The definitive guarantee agreements have not yet been filed; Nvidia stated that detailed terms and guaranteed minimum values will be included in its quarterly report for the period ending July 26.
In announcing the deal, Nvidia framed the project around the extended operational lifespan of the infrastructure. “We are securing long-lived infrastructure for Nvidia compute so OpenAI can deploy the most productive AI factories that can be upgraded repeatedly with each new generation delivering more intelligence and better economics,” CEO Jensen Huang said in Nvidia’s announcement.
Industry analysts note that the structure reflects a broader shift in how AI hardware is financed. Steven Dickens, CEO and principal analyst at HyperFrame Research, observed that Nvidia is recognizing the enduring worth of its chips. “NVIDIA is realizing, in the same way that Ford Finance does with Ford’s vehicles, that GPUs have a residual value long after the initial three years of use,” Dickens said. He added that leveraging GPUs as an investable asset class is a strategic move that unlocks capital for continued AI infrastructure expansion. “Leveraging the GPU as an investable asset class is a smart move by Nvidia as it unlocks capital for the AI build-out,” he said. With AI hardware typically maintaining useful lives of six to seven years or longer, Dickens noted there is significant room for financing models that account for value retention well past initial deployment.
The Ohio project effectively integrates three critical components of the AI stack: Nvidia’s proprietary hardware, OpenAI’s sustained compute demand, and SB Energy’s physical infrastructure ownership. It is crucial to distinguish the $105 billion guarantee from Nvidia’s actual equity commitment. The company is investing $1.5 billion directly into SB Energy, while the guarantee remains a separate contractual obligation tied specifically to the initial 4.25 GW deployment. Nvidia confirmed it has already secured land, power allocations, and shell capacity at the site and will serve as the exclusive AI compute infrastructure provider. Because the guarantee is strictly conditional, it does not represent a $105 billion cash outlay for construction. Instead, it functions as credit support linked to the leased facilities, with payouts triggered only by specific lease-related events involving OpenAI.
This structure leaves a key question for future disclosures: precisely how much financing capacity does the residual-value guarantee actually unlock? The forthcoming guarantee agreements are expected to provide clarity.
Beyond the financial mechanics, Dickens also challenged the industry’s growing tendency to measure AI infrastructure primarily by power capacity. “What I struggle with is measuring the AI build-out in units of power,” he said. “I would prefer Nvidia and the wider IT infrastructure business stick to IT units such as core, GPU, racks or servers, as it is easier to quantify and visualize.”
Looking ahead, the initial OpenAI deployment at the Ohio campus will deliver 4.25 GW of IT capacity, with the optional expansion potentially reaching 8 GW. To support this massive compute footprint, SB Energy and SoftBank have separately announced plans to develop at least 10 GW of new energy generation dedicated to the project.