Nvidia's $530 Billion Question: What the Commitments Really Are, and What to Ask on Earnings Day

Nvidia’s $530 Billion Question: What the Commitments Really Are, and What to Ask on Earnings Day

What Nvidia's $530 billion in commitments really includes, which $164.5 billion backs customers, and four questions to ask on the November 17 earnings call.

One number has followed Nvidia through its leadership crisis like a shadow: $530 billion. Depending on who is talking, it is either proof that the AI boom is built on a hidden pile of guarantees, or a meaningless total that bears have inflated to scare retail investors.

Neither version is quite right. The figure is real, it is large, and it deserves scrutiny, especially now that the company is run by an interim chief executive. But most of it is not what the scariest headlines suggest, and the part that does carry genuine credit risk is smaller, more specific and easier to monitor than a single headline total implies.

With Nvidia reporting third quarter results on Tuesday, November 17, this is the moment to understand what that number contains and which questions actually matter.

Where the $530 Billion Figure Comes From

The total was popularised by the research firm SemiAnalysis, which added up six line items in Nvidia’s most recent quarterly filing. In its analysis of Nvidia’s growing backstop universe, SemiAnalysis noted that the combined figure jumped from about $184 billion in the prior quarter to roughly $530 billion as of the quarter ended in late July.

Two points are worth stating clearly before going further. First, Nvidia’s filing does not use the phrase “off-balance sheet” and does not present a single headline total. The $530 billion is an analyst’s sum of separate tables. Second, those tables contain very different kinds of obligation, and adding them together mixes routine purchasing with genuine financial support for customers.

Most of It Is Nvidia Buying Supply

According to a line by line breakdown from Xenospectrum, the ordinary commitments table totals about $366 billion. The largest single item is $279 billion committed to secure future semiconductor supply and production capacity. Smaller lines include about $29 billion of cloud service contracts used for Nvidia’s own research and about $25 billion of data centre leases that have not yet started, mainly for Nvidia’s own use.

The supply line is the main driver of the jump. SemiAnalysis reports it rose from $119 billion to $279 billion in a single quarter, mostly for memory. That is an increase of about 134 percent in three months.

This is not money lent to fragile startups. It is Nvidia locking in the components it needs to build its own products in a market where memory is scarce and prices are rising. It still carries risk. If demand for AI systems slows sharply, Nvidia could be left paying for capacity it no longer needs, and the company has taken large inventory charges before when export rules shifted. But it is a supply chain bet, not a hidden loan book.

The Part That Actually Supports Customers

The more important number for investors sits in two smaller tables. Xenospectrum’s breakdown identifies about $36 billion of AI cloud agreements, about $20 billion of third party data centre leases that Nvidia has signed and expects to hand to customers, and a guarantees table with a maximum exposure of $108.5 billion. Together, that customer support subset comes to roughly $164.5 billion.

The same analysis puts this at about 72 percent of Nvidia’s shareholder equity of roughly $229 billion. That is a large ratio by any standard. It also notes that Nvidia earned about $118 billion of net income in the first half of its fiscal year, more than the entire guarantees table.

So the honest framing is this. Nvidia has committed a very large amount of financial support to its own customers. It also generates cash at a scale few companies in history have matched. Nvidia’s second quarter results release shows operating cash flow of about $24.1 billion and free cash flow of about $21.3 billion for the quarter. By simple arithmetic, the customer support subset equals roughly two years of free cash flow at that quarterly rate. The risk is real, but it is a question of concentration and timing, not of imminent insolvency.

The Ohio Guarantee Explained

Almost the entire guarantees table comes from one deal. Nvidia agreed to guarantee up to $105 billion connected to OpenAI’s 20 year lease at SB Energy’s PORTS-Pike campus in Ohio. As Implicator’s account of the deal terms explains, the figure was cut from roughly $250 billion discussed in late July after Nvidia shares fell when those talks became public.

The structure matters more than the headline. According to the same report, the guarantee covers the value of completed data centres rather than OpenAI’s rent itself. If OpenAI walks away, SB Energy must first try to re-lease the site at the same price, then try to sell it, and only then does Nvidia pay the shortfall up to the cap. The first 800 megawatts are expected in 2028.

In his own post, Securing the Infrastructure of Intelligence, Jensen Huang argued the deal is not circular financing because OpenAI pays the lease. He also wrote that the support phases in as facilities enter service and that Nvidia’s exposure falls as lease payments are made. Those are the company’s own claims, but they describe an obligation that is conditional, staged and years away from peak exposure.

Why the Circular Financing Debate Still Matters

None of this means the critics are wrong to worry. The concern is not that Nvidia will run out of money. It is that some of the demand behind Nvidia’s record revenue is supported by Nvidia itself, which makes that demand harder to judge.

The Bank for International Settlements put numbers on the wider pattern in its October bulletin on circular relationships among AI firms. It found that 55.2 percent of incoming investment into AI firms between 2021 and 2025 came from other AI firms, and that these arrangements, while often economically rational, carry macroeconomic risks and reduce transparency.

Nvidia’s management is aware of the perception problem. On the August earnings call, as Gizmodo reported, Colette Kress acknowledged that some would call the support circular financing and said the company sees it differently, arguing that Nvidia’s computing platform can be redeployed to other customers. The same report cited figures showing five customers accounted for about 70 percent of Nvidia’s receivables at the end of that quarter, up from 56 percent a year earlier.

That redeployment argument is the crux. It holds if demand for AI computing stays strong enough that any capacity a customer abandons can be filled by someone else. It weakens if many customers struggle at once.

The Customers Carrying the Debt

The firms most exposed to that scenario are the specialised AI cloud providers that borrow heavily to build capacity. CoreWeave is the bellwether. Its third quarter results showed revenue of about $3.66 billion and a revenue backlog of roughly $132.5 billion, both strong figures. But net interest expense was $846 million against operating income of $207 million, the quarter ended with a net loss of $709 million, and the company guided fourth quarter interest expense to between $1.05 billion and $1.15 billion.

That is the pressure point. Rapid growth paired with interest costs running at roughly four times operating income works while capital markets stay open and contracts convert to cash. It becomes fragile if funding costs rise or customers delay. Nvidia’s backstops exist precisely to make this kind of borrowing cheaper, which ties part of Nvidia’s risk to the credit health of these firms.

What to Listen For on November 17

Nvidia’s earnings call notice confirms results on Tuesday with a call at 5 p.m. Eastern. Kress, who defended the support programmes in August, now delivers the numbers as interim CEO following the leadership change announced on November 12. We looked at what that change means for the stock in Nvidia Without Jensen Huang: What the Founder Discount Is Really Pricing. Four questions matter more than the headline beat or miss.

Did the customer support lines grow again? The new quarterly filing will update the commitment tables. Growth in AI cloud agreements, third party leases and guarantees deserves more attention than growth in supply commitments.

Is the support being paused or reshaped? SemiAnalysis noted reports that Nvidia had paused its AI cloud partner programme. Any confirmation, and any explanation of why, would signal how the new leadership views this exposure.

How concentrated are receivables? A further rise in the share of receivables owed by a handful of customers would strengthen the bear case. Stable or falling concentration would weaken it.

What does management say about redeployment? If the company can point to capacity being reassigned smoothly between customers, the core defence holds. Vague answers would be telling.

The Contrarian Take

The lazy bear case treats $530 billion as a hidden debt bomb. The lazy bull case dismisses it as accounting noise. Both miss the useful middle.

Roughly two thirds of the total covers Nvidia’s own supply, research computing and facilities in a tight market. The genuinely sensitive part, about $164.5 billion, is concentrated in a few customers and one large, conditional, multi-year guarantee. It is large relative to equity, small relative to Nvidia’s earnings power, and most exposed to the credit health of heavily indebted cloud providers.

That gives investors something better than fear or faith: a short list of numbers to watch each quarter. If customer support keeps rising faster than cash flow while neocloud borrowing costs climb, the risk is building. If support stabilises and receivables broaden, the alarm is overdone.

This article is for information only and is not financial advice. Do your own research or speak to a licensed adviser before making investment decisions.

Mark Cannon
Mark Cannon
Articles: 410