AI's suppliers now fund their customers, and Nvidia's filing shows $99 billion of stakes

The companies that sell the chips behind artificial intelligence are increasingly also paying for the buyers' side of the trade. Nvidia (Nasdaq: NVDA) reported equity investments of $99 billion and further equity commitments of $25 billion as of July 26, 2026, in its quarterly report, and TSMC's board approved its first stakes in two AI model developers on November 10.

Taken together with the funding mix CoreWeave disclosed on Tuesday, the filings show money moving in both directions along the AI supply chain: suppliers invest in, guarantee and buy capacity from their customers, while those customers lean on prepayments and debt to buy the suppliers' products.

Nvidia's private stakes grew twelvefold in a year

The clearest figure is in Nvidia's notes on non-marketable equity securities, its holdings in privately held companies. They were carried at $47.9 billion at the end of July, against $3.8 billion a year earlier, an increase of 12.6 times (computed). Net additions in the quarter alone were $13.1 billion, and Nvidia recorded $4.9 billion of unrealized gains on these holdings in the same three months.

The company describes the purpose in its own words: it has made "investments and commitments in our ecosystem to enhance our growth opportunities, cultivate our ecosystem, and strengthen our competitive position." The filing does not list the companies behind the $99 billion figure in this section.

The gains already show up in profit. Other income, the line where investment gains are recorded, was $24.1 billion in the first half of fiscal 2027, against $3.0 billion a year earlier, and equaled 17.1% of income before tax (computed). Revenue in the second quarter was $96.2 billion, up from $46.7 billion.

Guarantees and cloud contracts reach further than equity

Equity is only one channel. In August 2026, Nvidia agreed to guarantees capped at $105 billion "to provide credit support on a land, power, and shell buildout" with affiliates of SB Energy on behalf of an affiliate of OpenAI, covering leases for about 4.25 gigawatts at a campus in Pike County, Ohio. In exchange, the filing says, "the site will exclusively host NVIDIA AI infrastructure, subject to limited exceptions." Including smaller guarantees for AI clouds, Nvidia put its maximum gross exposure at $108.5 billion.

Nvidia also became a buyer of the capacity it sells. Under what it calls a new business model with select AI cloud partners, those clouds buy Nvidia's data center products and Nvidia commits to cloud service agreements in return. Those commitments, "typically six years in duration," totaled $36 billion at the end of July. The same month, Nvidia signed memorandums of understanding with capital providers to set up financing platforms meant to mobilize "more than $500 billion of third-party capital over time."

Added up, the four channels in the filing (equity investments, equity commitments, cloud service commitments and guarantees) come to $268.5 billion (computed). They are different kinds of exposure and would not all turn into losses at once, but the sum shows their scale next to the $177.8 billion of revenue Nvidia reported for the first half of the fiscal year.

TSMC takes a smaller, first step

TSMC's version is far smaller. Its board approved buying up to $500 million of Anthropic shares "at its per share initial public offering price" and up to $500 million of OpenAI preferred stock, alongside $40.93 billion of capital appropriations, as FinanceInTime reported. The Anthropic purchase is part of a $26 billion private placement at the offering price with seven suppliers and partners, Nvidia and TSMC among them, according to FinanceInTime's report on the IPO range, which rests on TechCrunch and CNBC reporting.

The board also set aside up to $500 million for a new investment company "for the purpose of making strategic investments in the semiconductor industry and related ecosystem," without naming targets.

At the other end, buyers borrow and customers prepay

CoreWeave shows the buyer's side of the same flows. The AI cloud provider spent $9.44 billion on property and equipment in the third quarter against $1.98 billion of cash from operations, and its total debt reached $46.9 billion, as FinanceInTime reported. Its chief executive, Michael Intrator, said growth is supported by "financing backed by customer contracts, operating cash flow, and approximately $9 billion in customer prepayments and deposits expected in the next three months alone."

Prepayments are customers financing their supplier. Seen from CoreWeave's position in the chain, the money arrives from three sides: lenders, customers paying in advance and the chip supplier's own commitments to rent capacity, a model Nvidia's filing describes in general terms without naming partners.

What the filings do not show

The documents leave the most important questions open. Nvidia's report does not say which customers hold its cloud service commitments, how much of its revenue comes from companies it has invested in, or what the stakes cost against their current carrying values beyond the totals above. TSMC's resolutions give ceilings, not completed purchases. And the data does not show whether these arrangements raised demand that would not otherwise exist or only changed who pays for it; none of the companies says so.

We think supplier money is a support, and a concentration, at once

FinanceInTime's view is that the AI buildout has entered a phase in which the most profitable companies in the chain are lending their balance sheets to the least profitable ones. That is rational for a supplier with Nvidia's margins: a guarantee that locks a 4.25-gigawatt campus to its hardware, or a cloud contract that keeps a partner buying, protects future sales at a cost that is small next to $96.2 billion of quarterly revenue.

The threat is that the same structure concentrates risk. If demand for AI computing slows, Nvidia would feel it in orders, in the value of $47.9 billion of private stakes, in guarantees that turn into payments and in cloud capacity it has committed to rent. TSMC's $1 billion is too small to matter in that way, but it is a first step in the same direction. CoreWeave, with interest at 23.1% of revenue, depends on all of these sources continuing.

The opportunity is real too: supplier support lets capacity be built faster than buyers could finance alone, and if the end customers' revenue keeps growing, the stakes and guarantees become profitable rather than costly. Our forecast, conditional on the AI developers' revenue growth continuing over the next two quarters, is that these commitments keep rising in Nvidia's next filing. We would change our view if unrealized gains turn into impairments, or if the share of supplier money in the buyers' funding rises while their own cash flow does not.

The view in this final section is FinanceInTime's editorial opinion, for information only. It is not investment, tax, legal or other advice, and it does not take your circumstances into account. Forecasts are uncertain and past results do not guarantee future ones.