Nvidia's Form 10-Q for the second quarter of fiscal 2027, filed with the U.S. Securities and Exchange Commission (SEC), reveals the scale of the supply commitments and customer support underpinning its AI infrastructure business. As of July 26, 2026, its supply and capacity commitments stood at $279 billion, up $160 billion from the previous quarter. Nvidia has also committed to purchasing computing capacity held by AI cloud providers, and in August it entered into a guarantee agreement of up to $105 billion for a data center project used by OpenAI.
These deals show a business structure in which Nvidia does not simply sell semiconductors but also supports customers' procurement and operation of AI infrastructure. On November 12, Nvidia announced that Jensen Huang would step down as CEO and that Colette Kress would become interim CEO, raising the question of how the new leadership will manage the company's financial support and supply contracts.
The 10-Q Shows Supply Commitments Jumping by $160 Billion in One Quarter
According to Nvidia's 10-Q, supply and capacity commitments rose from $119 billion at the end of April 2026 to $279 billion on July 26. The $160 billion increase is about 1.7 times the quarter's revenue of $96.221 billion. Using the Federal Reserve Board's November 6 exchange rate of 158.23 yen to the dollar, the increase alone comes to roughly 25.3 trillion yen.
Nvidia describes these commitments as long-term arrangements to secure the supply needed for current and future products. They mainly cover memory and manufacturing capacity for data center infrastructure products, with scheduled payments of $92 billion for the rest of fiscal 2027, $87 billion for fiscal 2028, and $88 billion for fiscal 2029. Some contracts can be canceled, rescheduled, or adjusted in quantity before firm orders are placed, though doing so may incur additional costs.
However, the $279 billion is a total of scheduled payments spread over several years, so it cannot be compared directly with quarterly revenue. The growth in commitments suggests Nvidia has secured supply in anticipation of future demand, but if actual sales fall short of plan, inventory and purchase obligations could become a burden. The 10-Q itself cites the risk that demand fluctuations and manufacturing complexity could lead to higher inventory provisions, volatile revenue, and rising manufacturing costs.
The financial statements also reflect the supply expansion. Inventory rose from $21.403 billion on January 25, 2026, to $31.575 billion on July 26, and accounts receivable climbed from $38.466 billion to $63.059 billion. Meanwhile, the second-quarter gross margin held at 75.0%, and provisions for inventory and excess inventory purchase obligations came to $985 million. These are costs tied to inventory valuation and purchase obligations, not losses arising from the $279 billion in supply commitments themselves.
Nvidia Supports Customers' Financing, Not Just GPU Purchases
Alongside the supply commitments, Nvidia's financial support for customers is a key issue. A structure in which a chipmaker provides financial backing to customers who then buy its products is sometimes called "circular financing." Because chip sales and customers' capital spending depend on each other in such deals, the sustainability of the business is affected not only by end demand for computing but also by financing terms and future utilization rates.
The 10-Q describes AI cloud agreements that illustrate this structure. AI cloud providers procure Nvidia's data center products, while Nvidia contracts to buy the providers' computing capacity as a cloud service. As of July 26, 2026, the outstanding balance under these agreements was $36 billion, with terms typically of six years. If a provider can sell capacity to third parties on more favorable terms, it can unilaterally stop providing services to Nvidia.
In other words, Nvidia encourages customers to buy GPUs while also taking on part of the demand for the computing capacity built with them. If third-party usage grows, Nvidia's purchase obligations shrink, and under certain conditions Nvidia also receives a share of the cloud providers' third-party revenue. Nvidia explains that this model is meant to make computing capacity more accessible to emerging AI companies and research institutions.
Separately from these AI cloud agreements, the 10-Q as of the end of July also lists guarantees of up to $3.5 billion for AI cloud providers' data center lease obligations. These cover land, power, and buildings, with Nvidia assuming the burden if a customer fails to meet its contractual obligations.
A Guarantee of Up to $105 Billion for OpenAI-Related Data Centers
The largest of Nvidia's financial support arrangements is its contract involving SB Energy, a SoftBank Group subsidiary. On August 17, 2026, Nvidia filed with the SEC the conclusion of a guarantee agreement for the PORTS Technology Campus in Pike County, Ohio. An OpenAI affiliate will be the tenant, and the plan calls for deploying Nvidia AI infrastructure in nine data centers with a combined IT load of about 4.25 GW.
The agreement caps Nvidia's total guarantee at $105 billion (about 16.6 trillion yen). The guarantee takes effect in stages as facilities are completed and leases begin, with the first facility expected to come online in Nvidia's fiscal 2029. Each lease runs for 20 years, and the guarantee covers only a portion of payments related to land, power, and buildings.
Nvidia would owe payment only if conditions set out in the contract are met, such as the OpenAI affiliate defaulting. The amount would be determined by taking the guaranteed minimum value and subtracting sums recovered through re-leasing or selling the facilities. The guarantee also ends if OpenAI obtains a specified credit rating, among other conditions, so the $105 billion is a conditional maximum exposure rather than an amount Nvidia expects to pay.
Nvidia also has the option to provide additional guarantees for facilities totaling about 3.8 GW. Combining the existing $3.5 billion for AI cloud providers with the $105 billion for SB Energy, the maximum total guarantees listed in the 10-Q at the end of July come to $108.5 billion. In return for the guarantee, the covered facilities are to use Nvidia AI infrastructure, with some exceptions.
Financial Support Expanded in Spring, Then Began to Shrink After August
Nvidia was reportedly debating internally how to assess the risks of such customer support. According to exclusive reporting by The Wall Street Journal (WSJ) on November 13, in spring 2026 Nico Caprez, Huang's son-in-law who became a vice president in January that year, and others proposed financial support to emerging AI cloud providers and data center developers. Companies hoping to cut borrowing costs by leaning on Nvidia's creditworthiness reportedly flocked to the offer, and Nvidia is said to have asked for a share of revenue in return for its support.
Kress, then CFO, reportedly raised questions about how recipients were selected, how guarantees were valued, and how returns on investment were calculated. According to the WSJ, on July 26 it was reported that Nvidia was considering a $250 billion guarantee for OpenAI's Ohio data center plan, and the stock fell 5% the next day. Kress then reportedly reviewed the contract and cut Nvidia's initial guarantee exposure by about half, and the deal was concluded on August 17.
The paper also reported that in late August Nvidia paused some financial support deals, and in October began tightening support further. Approvals now require review by executives and the board, making it harder to close deals quickly as before. These accounts are based on anonymous sources, and the details of individual negotiations cannot be confirmed from SEC filings.
Kress herself acknowledged in late August that Nvidia's financial support could be called circular financing, while saying the company sees it differently. She expressed the view that the return on equity investments relative to capital deployed would be very good, positioning customer support as an investment rather than an inflation of demand.
A Soup Shop Anecdote Points to Pricing Talks With Major Customers
The WSJ also reported that at a conference in Berlin, Huang used a Japanese soup shop as an example to explain how to run a business to about six CEOs of cloud providers, Nvidia's major customers. The shop owner Huang described reportedly turned down additional orders even when long lines formed, keeping prices high and spreading sales across customers rather than selling out limited stock to a few. Huang is said to have praised the owner's approach and told the cloud executives that they had not fully established the value of their own products.
What matters in this anecdote is whom Huang was addressing. His audience was not the consumers or retailers who buy GeForce cards but cloud providers that procure Nvidia GPUs in bulk and sell the computing capacity to businesses and AI developers. In context, it reads as encouragement to avoid dependence on a few large customers and to sell computing capacity at appropriate prices.
This thinking also relates to Nvidia's own customer mix. According to the 10-Q, a single direct customer accounted for 16% of revenue in the second quarter of fiscal 2027, so dependence on customers remains significant. As large cloud providers adopt alternative chips, helping emerging cloud providers secure financing to widen the pool of buyers could work to reduce reliance on major customers. However, no figures have been disclosed that directly show how such financial support has affected the selling prices of Nvidia GPUs.
GeForce Price Spikes and Data Center Financing Should Be Considered Separately
Meanwhile, prices of consumer GeForce cards have swung sharply in autumn 2026. According to ITmedia PC USER's November 14 survey of Akihabara retailers, the Palit GeForce RTX 5090 GameRock at Dospara's Akihabara flagship store fell from 1,288,000 yen the previous week to 999,800 yen. Some ASUS models, on the other hand, exceeded 1.5 million yen, creating price gaps of more than 500,000 yen among products with the same GPU.
Domestic price swings have also hit the RTX 5080. In Gazlog's November 14 price survey, the lowest domestic price for the RTX 5080 rose from 267,700 yen on October 13 to 345,000 yen on October 31, then fell to 299,800 yen by November 13. The RTX 5090 has also dropped from about 1.29 million yen in early November to around 1 million yen, though prices vary widely by product and retailer.
Nvidia's financial support mainly targets data center AI infrastructure, and its transaction structure differs from retail GeForce sales. Consumer GPU prices involve multiple factors, including manufacturer and distributor shipping prices, inventory, exchange rates, and distribution channels. If Nvidia prioritizes manufacturing capacity and memory for data center products, it could affect the supply of consumer products, but the 10-Q does not reveal what share is allocated to each.
It is therefore difficult to explain Nvidia's financial support for AI cloud providers and the surge in GeForce retail prices through the same mechanism. What the disclosures on supply contracts and guarantees do show is that Nvidia is involved in customers' financing and in the sale of computing capacity in order to support demand for data center GPUs.
Three Metrics to Watch in the November 17 Earnings
Nvidia is scheduled to announce its third-quarter fiscal 2027 results at around 1:20 p.m. Pacific Time on November 17 (around 6:20 a.m. Japan time on November 18). The earnings call will begin at 2:00 p.m. the same day (7:00 a.m. Japan time on the 18th). Because the third quarter covers the three months through October 25, it offers a chance to see how much the scaling back of financial support since August is reflected in results and commitment balances.
The first point to watch is the balance of supply and capacity commitments and inventory provisions. If supply commitments decline from the previous quarter's $279 billion while provisions for inventory and excess inventory purchase obligations rise sharply from $985 million, the burden of the supply expansion may be starting to materialize. Conversely, even if the commitment balance stays high, it would support Nvidia's explanation of securing supply in line with demand if revenue and inventory growth are commensurate.
The second is the relationship between the scale of financial support and profitability. At the end of the previous quarter, maximum total guarantees stood at $108.5 billion and AI cloud agreements at $36 billion. As these balances and terms change, it will be necessary to see how revenue, gross margin, and cloud-related earnings move. Nvidia has guided third-quarter revenue of $108 billion (±2%) and gross margin of 74.0% (±0.5 percentage points), so how actual results compare with that outlook will also be a factor.
The third is the allocation of supply between data center products and GeForce. With Nvidia prioritizing manufacturing capacity for AI infrastructure, what it says about consumer GPU supply will be important. In particular, specific information on GeForce shipments and inventory would make it easier to judge how much the surge in retail prices is tied to supply constraints.
Nvidia has supported the expansion of AI infrastructure through massive supply contracts and customer support, but with the November 12 leadership change, capital allocation and risk management now pass to interim CEO Kress. The November 17 earnings will test how far the supply contracts and financial support built up so far have translated into actual revenue and profit.