Nvidia (NASDAQ: NVDA) enters its first earnings week without Jensen Huang facing a question larger than the headlines surrounding his departure: what happens to its $530.5 billion web of AI supply commitments, cloud agreements and potential guarantees? The sum can be reconstructed from the company’s fiscal second-quarter 2027 Form 10-Q. It is not $530.5 billion of debt, but it shows how closely Nvidia has connected its chip business to financing and building the AI ecosystem.
Huang resigned as chief executive and director, with longtime finance chief Colette Kress named interim CEO, Nvidia announced November 12. The issue for shareholders is whether the next leader can sustain AI sales and supplier advantages without expanding the company’s financial backstops as aggressively.
Update, November 15, 2026, 12:30 p.m. ET: Nvidia’s disclosed financial commitments and maximum gross guarantees total $530.5 billion, including a $105 billion conditional guarantee tied to an OpenAI-affiliated Ohio campus. The terms put capital allocation and counterparty risk at the center of Colette Kress’s interim leadership.
The $530.5 billion number: what Nvidia actually committed to
The quarterly filing separates ordinary forward spending, additional commitments supporting AI infrastructure, and maximum potential payments under guarantees. They belong in the same risk discussion, but they are not interchangeable accounting liabilities.
| Commitment or guarantee | Amount | What it represents |
|---|---|---|
| Supply and capacity | $279B | Future memory, manufacturing and other production capacity |
| Cloud service agreements | $29B | Compute used for research and development |
| Data-center leases not yet started | $25B | Planned engineering and testing facilities |
| Equity investment commitments | $25B | Future investments, subject to conditions |
| Capital expenditure commitments | $8B | Equipment and infrastructure |
| AI cloud agreements | $36B | Minimum cloud capacity and revenue-support arrangements |
| Third-party data-center leases not yet started | $20B | Leases Nvidia expects to reassign |
| Land, power and shell guarantees | $108.5B | Maximum gross exposure if specified defaults occur |
| Combined gross total | $530.5B | Different obligations across multiple years |
The arithmetic is $366 billion of scheduled commitments, plus $56 billion of additional agreements, plus $108.5 billion of maximum gross guarantees. Research firm SemiAnalysis estimated that the corresponding aggregate was roughly $184 billion in the previous quarter. The increase is about $346.5 billion, or 188%. Crucially, the current figure includes an Ohio guarantee signed in August, after the July 26 fiscal-quarter end and disclosed as a subsequent event.
The biggest exposure is upstream, not a loan to customers
Nvidia’s supply and capacity commitments jumped from $119 billion to $279 billion in one quarter, a 134% increase. Roughly $267 billion, or 96%, is scheduled through fiscal 2029. The 10-Q explains that these agreements primarily secure memory and manufacturing capacity for multiple product generations, with some contracts allowing cancellation, rescheduling or adjustment before firm orders are placed.
That is a powerful competitive position when AI accelerator demand outruns component supply. Secured production slots can protect shipment schedules and give Nvidia bargaining leverage. In a demand slowdown, however, the same arrangements can leave the company paying adjustment costs, carrying excess inventory or accepting weaker margins to clear capacity. The filing does not identify a contractual HBM4 price for Nvidia or its rivals, so precise supplier-price claims should not be treated as established economics.
The $105 billion Ohio guarantee is conditional, but strategically important
The most striking customer-side arrangement is a guarantee capped at $105 billion for an SB Energy project serving an OpenAI affiliate at the PORTS Technology Campus in Ohio. Together with $3.5 billion of other AI-cloud lease guarantees, the maximum gross exposure reaches $108.5 billion. According to Nvidia’s disclosure, the Ohio support covers defined portions of land, power and building-related lease payments, not the full project cost.
The guarantees generally become effective phase by phase when leases start. The first of nine phases is expected in fiscal 2029; payments would be triggered by specified tenant defaults. Exposure is expected to decline over the 20-year lease terms and may terminate sooner if OpenAI reaches a satisfactory credit rating. In return, the campus is expected to use Nvidia AI infrastructure almost exclusively. The economic trade is clear: conditional credit support helps secure a long-lived hardware customer.
Why the neocloud agreements matter more than their headline value
The separate $36 billion of AI-cloud agreements represent a newer model: cloud operators buy Nvidia infrastructure, while Nvidia commits to purchase or backstop a portion of the resulting cloud capacity. Customers can redirect that capacity to third parties on more favorable terms, which reduces Nvidia’s exposure. If certain conditions are met, Nvidia also earns a share of third-party cloud revenue. The mechanism is described in the filing; it is not equivalent to Nvidia simply lending $36 billion to buyers.
Another $20 billion covers leases Nvidia expects to transfer to third parties after construction. These deals address a recurring problem for smaller GPU-cloud operators: lenders want contracted utilization before they finance expensive equipment and facilities. A backstop from the chip supplier can help a project get built. Yet it also ties Nvidia’s economics to its customers’ occupancy, rental rates and creditworthiness. Reuters reported in August that Nvidia had paused some proposed revenue-sharing arrangements, while the company said the model remained in place and was evolving.
The risk is visible at the customer level. CoreWeave’s third-quarter results paired $3.661 billion of revenue with $846 million of net interest expense, illustrating how much financing costs can matter even when GPU demand grows quickly. The larger ecosystem risk is explored in our comparison of AI data-center debt exposures. Strong demand and expensive financing can coexist.
A $99 billion investment portfolio adds another layer of sensitivity
Nvidia also disclosed approximately $99 billion of equity investments and another $25 billion of future investment commitments. The existing equity holdings are assets on the balance sheet, not an extra $99 billion to add to the $530.5 billion total. Their valuations can nevertheless move with the same AI infrastructure cycle that drives Nvidia’s sales. In addition, the company reported roughly $33.4 billion of debt outstanding at July 26, after issuing $25 billion of senior notes in June. See the financial statements.
This is the correlation shareholders must watch. If AI-capacity economics weaken, Nvidia could face lower new hardware orders at the same time that the value of strategic investments falls and some customer-support commitments become more costly. The company can benefit from several parts of a growing ecosystem, but it can also be exposed to several parts of a downturn.
Kress may change the next deal, but old agreements remain
Kress joined Nvidia in September 2013 and previously held senior finance positions at Microsoft and Cisco. She remains Nvidia’s chief financial officer while also serving as interim CEO, giving her direct knowledge of the obligations being reassessed. Reuters reported November 13 that Kress and her finance team had challenged some of Huang’s larger financing initiatives. Sources described a proposed $250 billion Ohio support plan that was scaled back to $120 billion. That account suggests a more disciplined approval process could emerge, but it does not establish that existing commitments will be canceled or that Nvidia will abandon customer financing.
The board has said it expects operations, customer commitments and the product roadmap to continue without interruption. Nvidia’s leadership announcement names Kress as interim CEO and says the permanent search will include internal and external candidates. That creates two separate investment questions: whether old deals still earn acceptable returns, and whether new deals will clear a higher hurdle.
Nvidia plans to report fiscal third-quarter 2027 earnings on Tuesday, November 17, at approximately 4:20 p.m. ET, with its call at 5:00 p.m. ET, according to its investor-relations announcement. Investors should focus on management’s explanation of financing policy as much as revenue and margin guidance.
- Supplier exposure: Updated purchase and capacity commitments, their payment schedule and any contract changes.
- Cloud financing: New guarantees, minimum-revenue commitments, leases and whether previously announced structures are expanding or contracting.
- Cash conversion: Operating cash flow, working capital, capital spending and debt versus cash resources.
- Demand quality: Data Center growth, gross margins, customer concentration and evidence that cloud utilization supports profitable GPU purchases.
For context on the initial leadership shock, see our coverage of Nvidia’s November 11 selloff and Huang’s exit. The financial test now extends beyond the succession itself.
Bottom line
Nvidia has not inherited a $530.5 billion debt bomb. It has inherited an enormous, layered set of supply commitments, capacity arrangements, investments and contingent guarantees designed to keep the AI infrastructure cycle moving. The next management team must show that these arrangements produce durable cash returns without leaving shareholders disproportionately exposed if AI cloud economics deteriorate.
Data as of 12:30 p.m. ET, November 15, 2026. For informational purposes only; not investment advice.