1. Nvidia off balance sheet exposure reaches $530 billion
The Nvidia off balance sheet total of commitments and guarantees climbed to roughly $530 billion in its fiscal second quarter, which ended July 26, 2026, nearly triple the $184 billion recorded one quarter earlier. Research firm SemiAnalysis arrived at the total by adding up the commitment and guarantee disclosures in Nvidia's quarterly Form 10-Q filing.
The jump is notable both for its size and its speed. Obligations of this kind usually build up gradually as a company signs new contracts. In Nvidia's case, the figure grew by about $346 billion in a single three-month period, a change that reflects how aggressively the chipmaker has been securing supply and supporting the construction of artificial intelligence infrastructure.
It is important to be precise about what the number represents. Nvidia does not itself report a single $530 billion figure or describe these items as off balance sheet. The total is SemiAnalysis's aggregation of several separate tables and disclosures in the filing. It describes future obligations and maximum potential exposures, not money Nvidia has already spent or losses it expects to incur.
2. What does off balance sheet mean?
Off balance sheet refers to financial obligations that a company has agreed to but that do not appear as liabilities on its balance sheet. They are typically disclosed in the notes to a company's financial statements rather than in the headline figures investors see first.
These items can take several forms. A purchase commitment is a promise to buy goods or services in the future. A guarantee is a promise to cover someone else's payments if they fail to make them. Neither creates a liability on the balance sheet at the moment it is signed, because the payment or the default has not happened yet, but both can turn into real cash outflows later.
Accounting rules require companies to disclose material commitments and guarantees so that investors can judge the risks they carry. The 10-Q, the quarterly report that U.S. public companies file with the Securities and Exchange Commission, is where these details are published between annual reports. For a company of Nvidia's scale, the notes to the 10-Q can contain obligations that are large relative to the numbers on the balance sheet itself.
3. Supply and capacity commitments jump to $279 billion
The largest piece of the total is supply and capacity commitments, which rose to about $279 billion from $119 billion in the prior quarter. That increase of roughly $160 billion accounts for close to half of the overall jump on its own.
These commitments represent agreements by Nvidia to purchase manufacturing capacity, components and finished products from its suppliers. Nvidia does not operate its own chip factories. It designs processors and relies on outside partners for fabrication, advanced packaging, memory and the assembly of complete AI systems, so locking in that capacity ahead of time is central to its ability to meet demand.
The scale of the increase suggests Nvidia is reserving supply far into the future. In a market where the components needed to build AI accelerators have been in short supply, securing capacity early can be a competitive advantage. It also means Nvidia has committed to paying for that capacity whether or not demand holds up at its current pace.
The commitments also show how far ahead Nvidia is planning. Building new chipmaking, packaging and memory capacity takes suppliers years, and they generally want firm orders before investing. Long-dated purchase agreements give those partners the confidence to expand, while giving Nvidia a claim on output that rivals might otherwise secure.
4. The $105 billion SB Energy lease guarantee
The single most striking item is a guarantee of up to $105 billion tied to leases at SB Energy's data center campus in Ohio, which is leased to an affiliate of OpenAI. Nvidia agreed to backstop those lease payments on behalf of the customer.
That guarantee makes up nearly all of Nvidia's total maximum gross exposure under guarantees, which stands at $108.5 billion. In practical terms, it means that if the OpenAI affiliate failed to meet its lease obligations, Nvidia could be called on to cover them, up to the stated cap.
Arrangements like this are designed to help large infrastructure projects obtain financing. Developers and lenders funding a massive data center campus want assurance that the tenant will pay over the life of the lease. A guarantee from a company with Nvidia's financial strength can make a project easier to finance than it would be on the tenant's credit alone, and in turn helps create demand for the computing hardware Nvidia sells.
The guarantee also concentrates risk. A large share of Nvidia's guarantee exposure now rests on one counterparty and one project. That is a very different profile from a diversified set of smaller commitments spread across many customers.
For investors, the guarantee also highlights how closely Nvidia's fortunes are now linked to OpenAI's expansion plans. The tenant's ability to keep paying over the life of the leases depends on its own growth and access to capital. As long as that holds, the guarantee stays in the background as a contingent promise. If it ever came under strain, the exposure would move from the footnotes toward Nvidia's income statement, which is why analysts tracking the company are likely to watch this single item closely in future filings.
5. AI cloud agreements add $36 billion
Another $36 billion comes from what Nvidia calls AI cloud agreements. SemiAnalysis characterizes these as take-or-pay commitments with so-called neoclouds, the newer cloud providers that specialize in renting out GPU computing capacity for AI workloads.
A take-or-pay arrangement obliges the buyer to pay for a minimum amount of a product or service whether or not it actually uses it. In this context, Nvidia has committed to purchasing a floor level of computing capacity from these cloud partners. The structure gives the neoclouds predictable revenue, which they can use to finance the purchase of more hardware, much of it from Nvidia.
Neoclouds have become an important channel in the AI market. They compete with the largest cloud platforms by offering access to the latest accelerators, often to AI developers that need large amounts of capacity quickly. Nvidia's willingness to stand behind part of their demand helps these companies grow, but it also ties a portion of Nvidia's obligations to their success.
6. How SemiAnalysis reached the total
The $530 billion figure is built from three groups of disclosures in the filing: Nvidia's main table of commitments, an additional table of commitments and the table of guarantees. Added together, they produce a total of roughly $530 billion, compared with $184 billion on the same basis a quarter earlier.
Because the figure combines different kinds of obligations, it should be read with care. Purchase commitments are amounts Nvidia expects to pay as it receives goods and services it intends to use. Guarantees, by contrast, are contingent: Nvidia pays only if another party defaults. Adding them together gives a sense of the overall scale of Nvidia's promises, but it does not describe a single type of risk.
SemiAnalysis's work highlights how much of the AI build-out is being financed through structures that sit outside the balance sheet. Nvidia is not only selling chips. It is also helping arrange and underwrite the supply chains, data centers and cloud capacity that consume them.
7. Why Nvidia is backing its own customers
Nvidia has become one of the most valuable companies in the world by supplying the processors that power AI models. Demand for those processors depends on customers being able to build and operate the data centers that house them, which requires enormous amounts of capital, power and land.
By guaranteeing leases and committing to buy cloud capacity, Nvidia reduces the financing hurdles its customers face. That can accelerate the construction of new AI infrastructure and, in turn, sales of Nvidia's hardware. The approach effectively uses Nvidia's strong balance sheet to support demand for its own products.
The strategy carries a trade-off. When things go well, the commitments help sustain growth and keep Nvidia at the center of the AI supply chain. When demand slows, the same commitments can become a burden, because Nvidia would still owe payments for capacity and could be exposed to customers who struggle to meet their own obligations.
8. What the commitments mean for Nvidia investors
Off balance sheet does not mean free of obligation. These items are potential future liabilities, and the guarantee figure in particular is a maximum gross exposure: a ceiling on what Nvidia could owe, not an estimate of what it expects to lose. A guarantee only turns into a payment if the party being backed fails to pay.
Still, the size and speed of the increase give investors reason to pay attention. A jump from $184 billion to $530 billion in a single quarter shows how deeply Nvidia has tied its future to continued spending on AI infrastructure. The concentration of more than $100 billion of guarantee exposure on a single OpenAI-linked project adds a specific counterparty risk on top of the broader bet on AI demand.
The coming years will test whether that bet pays off. If AI investment keeps growing, Nvidia's commitments will look like a well-timed effort to secure supply and support its customers. If demand cools, the company could find itself paying for capacity the market no longer needs. Future 10-Q filings, and how these figures change from one quarter to the next, will be the clearest way to track which path Nvidia is on.