Key Notes
- Nvidia’s disclosed gross commitments and guarantee exposure add up to roughly $530 billion.
- Most of the total consists of future commercial commitments, with guarantees reported separately.
- Colette Kress inherits decisions over customer credit support, cloud capacity and supply commitments.
Nvidia (Nasdaq: NVDA) is facing fresh scrutiny over its financing of chip customers after a Wall Street Journal report described how Colette Kress questioned the deals before replacing Jensen Huang as interim chief executive.
The company’s latest quarterly disclosures contain roughly $530 billion in gross commitments and guarantee exposure when several categories are added together. The amount spans supply purchases, cloud services, leases, investments and contingent guarantees, rather than a single borrowing obligation.
Supporting customers can generate hardware sales and investment returns. It can also leave Nvidia paying for unwanted capacity or covering certain obligations if those same customers falter.
What the $530 Billion Actually Covers
Adding the relevant categories in Nvidia’s fiscal 2027 second-quarter filing produces approximately $530.5 billion:
- $366 billion in future commitments, including $279 billion for supply and manufacturing capacity, plus cloud services, leases, equity investments and capital spending.
- $56 billion in additional commitments, comprising $36 billion of AI cloud agreements and $20 billion of leases intended for third parties.
- $108.5 billion in maximum gross guarantee exposure, including a $105 billion Ohio backstop and $3.5 billion of other AI cloud guarantees.
The report covers the quarter ended July 26 but also discloses the Ohio guarantees signed in August. Some payments run into the 2030s and beyond, and some supply arrangements can be adjusted before firm orders.
Nvidia reported $91.3 billion of balance-sheet liabilities, including about $33.4 billion of debt. Comparing that balance with the gross commitment total does not measure an equivalent amount of hidden debt or expected losses.
Why Kress Challenged the Deals
The Journal reported that Nico Caprez, Huang’s son-in-law and a vice president since January, helped pitch credit support to cloud startups and data-center developers. His team told prospective partners that arrangements could be organized within weeks, with Nvidia seeking a share of cloud revenue in return.
Kress pressed for a method to value the guarantees and calculate Nvidia’s return, according to people cited by the newspaper. At an August meeting, Huang interrupted her and called her “a deer in the headlights.”
The Ohio talks initially involved a $250 billion backstop. The Journal credited Kress with reducing the proposal before the smaller deal was signed in August. It also reported an August pause in some financing deals as executive and board scrutiny increased. Kress nevertheless publicly defended the strategy.
How Customer Financing Can Feed Chip Sales
The commercial incentive is straightforward. A developer with access to cheaper funding can build a larger campus, while a cloud operator can buy more accelerators. Nvidia can earn hardware revenue as that infrastructure is delivered and may receive investment returns or revenue sharing later.
Its fiscal second-quarter CFO commentary describes customers growing faster than their balance sheets and credit profiles can support. Nvidia says selected financing arrangements help bridge that gap and secure sites for multiple generations of its computing systems.
The concern for shareholders is how much demand depends on that support. If a customer cannot obtain financing without Nvidia’s backing, the chipmaker is exposed to more than the sale: it may also bear investment losses, unused-capacity costs or obligations under a guarantee. Those exposures arise through different contracts and do not make Nvidia a direct lender in every transaction.
OpenAI’s Ohio Campus Is the Largest Backstop
Under the Ohio arrangement, SB Energy is building and operating the PORTS-Pike campus for OpenAI. Nvidia’s August announcement also committed $1.5 billion to SB Energy and made Nvidia the exclusive compute infrastructure provider, subject to limited exceptions.
The $105 billion guarantee covers defined lease and power obligations associated with an initial 4.25 gigawatts of capacity. It becomes effective in phases as facilities enter service, with the first expected in fiscal 2029. Exposure falls as OpenAI makes payments, and specified conditions can end the guarantees, including a satisfactory credit rating.
This ties Nvidia’s potential chip sales to the financing and performance of a long-lived property development. The guarantee ceiling is potential contractual exposure, not $105 billion already paid.
CoreWeave Shows the Overlapping Roles
CoreWeave (Nasdaq: CRWV) illustrates the relationship at a cloud operator. Nvidia is its equipment supplier, shareholder and a buyer of computing capacity. In January, Nvidia invested $2 billion in its shares as the companies expanded their partnership.
A September 2025 agreement, with an initial value of $6.3 billion, obliges Nvidia to purchase residual cloud capacity CoreWeave has not sold to other customers through April 13, 2032. Delivery, service availability and termination provisions apply.
That gives the operator a potential buyer for otherwise idle capacity, while linking Nvidia to the economics of systems built around its hardware. CoreWeave’s recent funding pressures show why investors are examining those ties alongside revenue growth.
The Risk Can Work in Both Directions
Supply commitments add another obligation to the customer-financing exposure. Nvidia committed $279 billion to supply and capacity, compared with $119 billion in the previous quarter. That spending is meant to secure future production. If customer funding slows, Nvidia could face weaker orders while still working through purchases arranged for stronger demand.
The company is also trying to bring in outside investors. Its proposed independent financing platforms target more than $500 billion of third-party capital over time. That is a fundraising ambition under partnerships subject to final agreements, separate from the commitments already disclosed.
Kress’s task is to keep customers building without leaving Nvidia responsible for too much of the funding. Tightening support could constrain buyers that depend on it; extending more support would leave the chipmaker carrying additional risk. After her appointment on November 12, the next scheduled chance for investors to hear how she plans to manage that trade-off is Nvidia’s November 17 earnings call.
