Nvidia Expands AI Cloud Financing With Revenue-Sharing Backstops

Nvidia is turning parts of its AI chip business into a financing-and-revenue-share model, offering credit support to AI cloud partners that buy its systems while seeking a slice of the cloud revenue those systems generate. The approach could widen access to scarce GPU capacity, but it also raises new questions about customer dependence, balance-sheet exposure and the quality of demand behind the AI infrastructure boom. Nvidia described the arrangement in July as a revenue-sharing and credit-support model in which AI cloud providers procure NVIDIA infrastructure, sell NVIDIA-powered cloud services and give Nvidia both standard product revenue and a share of revenue from supported capacity. The first named participants included Sharon AI, which Nvidia said is deploying up to 40,000 NVIDIA Grace Blackwell GB300 GPUs, and Firmus, which is building a NVIDIA DSX AI factory campus in Batam, Indonesia, expected to scale to 360 megawatts and up to 170,000 Nvidia GPUs. Nvidia later described the mechanics as a take-or-pay commitment on part of a facility's capacity: a minimum revenue guarantee helps lenders underwrite a project, and Nvidia shares in neocloud revenue above that floor. The company said independent capital still underwrites each deal, while the model could produce billions of dollars of medium- to long-term revenue on top of one-time hardware sales through a recurring usage-linked stream. Nvidia disclosed $36 billion of AI cloud agreement commitments, typically six years in duration, plus $20 billion of data center leases not yet commenced for third parties, for $56 billion of additional commitments tied to customer infrastructure support as of July 26, 2026. The same filing also disclosed $108.5 billion of maximum gross exposure from guarantees, including $3.5 billion for select AI cloud partners' land, power and shell obligations and $105 billion tied to SB Energy infrastructure for an OpenAI affiliate's leases covering about 4.25 gigawatts of IT load at the PORTS Technology Campus in Ohio. The program sits beside a larger capital-market push. Nvidia announced memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute financing platforms designed to mobilize more than $500 billion of third-party capital for AI infrastructure over time; Nvidia later emphasized that the $500 billion figure is not Nvidia revenue, not one fund and not a commitment to one customer in its AI factory financing explanation. The strategic rationale is clear: Nvidia reported $96.2 billion of fiscal second-quarter revenue, up 106% year over year, with Data Center revenue of $89.0 billion, up 117%, and its filing shows that three direct customers represented 16%, 15% and 13% of first-half fiscal 2027 revenue. Extending financing support, longer payment terms and minimum revenue guarantees can help Nvidia reach AI startups, model builders, enterprises, research organizations and sovereign customers that cannot fund hyperscale AI infrastructure on their own. The reception is mixed. Developers and infrastructure buyers see the model as a faster route to full-stack accelerated computing without waiting through site selection, power procurement, construction and hardware bring-up. Supporters also argue that distributed AI cloud capacity could reduce reliance on a handful of hyperscalers and make large-scale inference available to more model builders and agent platforms. Critics focus on the other side of that same structure: if Nvidia supports the financing of customers that buy its chips, reported demand can look more dependent on Nvidia's own balance sheet, and customers may become less independent in how they allocate capacity. Reports of partner pushback have centered on capacity-allocation limits over which customers could rent supported GPUs and a preference that capacity be spread across multiple smaller AI firms rather than one large buyer; Nvidia has said the July model remained in place and continued to evolve because of high demand. Nvidia's own risk language is also cautious: if AI clouds fail to sell committed capacity, Nvidia may have to purchase it, and lower-than-expected compute demand or pricing could reduce the revenue it receives. The near-term test is whether third-party demand, not only financing architecture, absorbs the new capacity at profitable utilization rates.