Nvidia discloses up to $118B in AI infrastructure guarantees
Data center revenue reached $107.6 billion, up 110% from a year earlier. Supply commitments climbed to $390 billion, and Vera Rubin NVL72 systems entered volume production.

Nvidia's data center revenue rose 110% from a year earlier to $107.6 billion, TechTarget reported on Wednesday, as the company also laid out $118 billion in maximum guarantee exposure tied to AI infrastructure buildouts. Total revenue reached $115.8 billion, up 103%, and net income climbed to $65.3 billion from $31.9 billion a year earlier.
Most of that exposure sits behind the PORTS-Pike Technology Campus in Ohio, run by SB Energy. Nvidia is providing credit support for roughly 4.25 gigawatts across nine buildings there, and the site will host Nvidia systems leased by OpenAI for 20 years. Each guarantee starts only when its building is finished and ready for service; the first completion is expected in fiscal 2029. The Ohio backstop would peak at $105 billion before declining as OpenAI makes lease payments. Nvidia said no guarantee payments had been made as of Oct. 25.
The $118 billion ceiling also includes $9.5 billion in credit support guarantees for power and systems, and $3.5 billion in land, power and shell guarantees for neocloud providers. Nvidia separately disclosed $32 billion in potential commitments on neocloud contracts, contingent on those customers obtaining financing, purchasing Nvidia gear and deploying capacity. Stephen Sopko of HyperFrame Research said the largest piece of the guarantee exposure is a single campus leased to OpenAI, so the total should not be read as direct neocloud financing.
Nvidia's supply-chain obligations rose to $390 billion from $279 billion the previous quarter, with most of the $111 billion jump coming from multiyear memory purchases. New arrangements with Kinsus, Samsung Electro-Mechanics, Siliconware Precision Industries and Amkor Technology will add capacity. Kinsus expects to expand advanced ABF substrate capacity set aside for Nvidia to more than 3.5 times its current level by the third quarter of 2028.
The company also moved Vera Rubin NVL72 systems into volume production, calling it the fastest product ramp in its history. Early deployments are live with Nebius, CoreWeave, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure, and SpaceXSI is next in line. On the MLPerf Endpoints test, Vera Rubin generated ten times the tokens per watt of the GB300.
For the fourth quarter, Nvidia expects revenue of about $130 billion, give or take 3%, assuming no data center compute sales in China. It sees gross margins between 71% and 72% for the quarter.
The disclosures follow Nvidia's earlier moves to back neocloud construction. Five Nines reported on Nov. 2 that Nvidia was in talks to provide financial support for Nscale's 460-megawatt West Virginia campus being built for Anthropic. That backstop, estimated above $5 billion, would cover power and construction for a building containing Nvidia Vera Rubin hardware.
Sources: TechTarget, Five Nines
Arjun Mehta covers data centres, AI infrastructure and networks for Five Nines. He writes from company announcements, filings and published reporting.
Corrections. Report an error to corrections@thefivenines.com. We list substantive corrections on the corrections page.