Nvidia’s financial guarantees tied to artificial intelligence infrastructure have climbed to $118 billion, up $9.5 billion from the previous quarter, as interim CEO Colette Kress clarified when the chipmaker could become liable for payments involving its data center partners.
The figure represents Nvidia’s maximum potential exposure under various financing arrangements, rather than money already spent or an immediate obligation to pay. According to the company’s latest financial commentary filed with the Securities and Exchange Commission, the guarantees would require Nvidia to make payments to third parties only if certain partners default on their obligations and the applicable conditions are met.
The total has increased from $108.5 billion disclosed in the previous quarter, primarily reflecting an additional $9.5 billion in power and systems credit support guarantees. Nvidia said it has not made any payments under its guarantees.
The largest exposure comes from a $105 billion arrangement supporting SB Energy’s PORTS-Pike Technology Campus in Ohio, a major AI infrastructure project that will host Nvidia computing systems under long-term leases to OpenAI.
The arrangement, originally disclosed in August, covers approximately 4.25 gigawatts of capacity across nine data center buildings. Each guarantee becomes effective only after its corresponding building is completed and ready for service.
Nvidia expects the first building to meet those conditions in fiscal 2029, with its potential exposure increasing in stages as additional facilities become operational. The guarantees are structured to decline over the 20-year lease terms as OpenAI makes payments.
The remaining $13 billion consists of $3.5 billion in guarantees supporting land, power and building infrastructure for other AI cloud operators, along with $9.5 billion in power and systems credit support.
Kress addressed the financing arrangements during Tuesday’s earnings conference call, responding to questions about whether Nvidia was changing how it supports specialized AI cloud providers, commonly known as neoclouds.
Bernstein analyst Stacy Rasgon questioned language in Nvidia’s latest financial disclosures stating that certain commitments would take effect only after the operators secured financing to purchase computing infrastructure and brought the relevant capacity into operation.
Kress explained that companies must first arrange the necessary debt or equity financing, complete construction and prepare the computing infrastructure for use before the relevant guarantees become effective.
She said construction delays or difficulties meeting those conditions could prevent the guarantees from taking effect. However, she emphasized that Nvidia had not introduced a new financing policy.
“Is it new? No. Did we change anything? No,” Kress told analysts, explaining that the company wanted to provide greater clarity about arrangements that were already in place.
The distinction is significant because Nvidia has increasingly supported companies that also purchase its processors and computing systems. These arrangements can help customers expand AI infrastructure while creating potential financial exposure for Nvidia if certain contractual obligations are not fulfilled.
The company separately disclosed $32 billion in future commitments associated with neocloud agreements, down from $36 billion in the previous quarter. These commitments are distinct from the $118 billion in guarantees and are subject to their own financing, deployment and operational requirements.
Under those agreements, Nvidia earns revenue from the initial sale of computing infrastructure and may also receive a share of revenue generated by cloud operators serving third-party customers, provided certain conditions are satisfied.
Kress also discussed how Nvidia would approach future financing arrangements, particularly those involving land, electricity infrastructure and the physical buildings required to house AI computing systems.
Responding to Cantor Fitzgerald analyst C.J. Muse, she identified land, power and building infrastructure as potential priorities for financial support, while indicating that Nvidia would carefully evaluate which companies receive assistance.
She also said the company would continue working with established customers, suggesting that greater scrutiny of future agreements does not necessarily mean Nvidia is withdrawing support from existing infrastructure partners.
In her prepared remarks, Kress said Nvidia would remain disciplined in its financial commitments as additional sources of capital become available to support the AI industry.
The disclosures follow Nvidia’s record quarterly earnings announced Tuesday, when the chipmaker reported $115.8 billion in revenue and projected further growth driven by demand for AI infrastructure.