Published on: 2026-07-28
Circular financing begins when a seller’s capital or credit helps a customer buy the seller’s products. AI companies are committing hundreds of billions of dollars to chips and data centres, while some suppliers are also investing in customers, guaranteeing loans or absorbing unused capacity.
Nvidia’s reported talks to support roughly $250 billion of financing for an OpenAI data-centre project have renewed scrutiny of how much AI demand is independently funded.
Circular financing gives a customer the means to buy a supplier’s products using the supplier’s own capital or credit.
A sale can be genuine even when the customer’s ability to pay depends partly on the seller.
Nvidia’s proposed $250 billion OpenAI backstop, its CoreWeave agreements and AMD’s OpenAI warrants show that supplier support takes several forms, not one.
Cash conversion, receivables, customer funding, guarantees and capacity use together indicate whether demand can eventually stand without supplier support.
The evidence so far does not show the whole AI boom is self-funded. It shows some suppliers are keeping more of the risk behind customer purchases.
Circular financing is an arrangement in which a supplier’s own capital or credit gives a customer the buying power to purchase the supplier’s products. A normal sale runs the other way: the customer pays with money it earned or borrowed from an independent lender. Circular financing changes the source of that buying power.
Picture a chipmaker that invests $10 billion in a cloud company. The cloud company spends part of that capital on servers built around the chipmaker’s processors. The chipmaker now holds an investment in its customer and books revenue on equipment the customer bought with money the chipmaker supplied.
The worry is not that the sale is fake. It is that reported revenue can look like independent demand when the supplier quietly helped fund the purchase. For as long as its balance sheet allows, a supplier that helps fund its own demand can keep sales rising even when end-user demand is soft. Revenue records what was sold; it does not show who made the purchase financially possible.
A circular deal can be entirely real and still overstate independent demand. The chips ship, the data centre gets built, and the supplier collects payment. The open question is whether the customer could have made the same purchase without the seller’s help.
The timing gap is the crux. The supplier books revenue now, while the customer may need years of subscriptions, cloud usage or enterprise contracts to earn back the investment. Financing pulls orders forward: a guarantee or cheap equity lowers the customer’s cost of capital and reassures lenders, so purchases that depend on cheap money arrive sooner than end-user demand alone would justify.
In a capital-hungry business like AI infrastructure, that timing gap can make a young market look far deeper than its cash generation supports.
The seller’s exposure does not end at delivery. An equity stake, a guarantee, extended payment terms or a promise to buy unused capacity can leave the supplier on the hook if the customer or the project stumbles. The deal lifts today’s revenue and parks tomorrow’s risk on the supplier’s balance sheet.
Supplier support for customer purchases takes four broad forms, each carrying a different risk back to the seller.
Structure |
How it supports demand |
Risk the supplier keeps |
|---|---|---|
Loan or extended credit |
Customer receives cash or more time to pay |
Default and collection risk |
Equity investment |
Customer gains capital to expand |
Loss on the stake in the customer |
Financing guarantee |
Lenders rely on the supplier’s backing |
Payment if the customer fails |
Warrant or capacity agreement |
Customer gains equity upside or a buyer for spare capacity |
Dilution or utilisation risk |
The labels matter. Vendor financing usually means direct loans, payment terms or guarantees from a supplier. Circular financing is broader: it can bundle several connected arrangements, from equity stakes to capacity deals.
Round-tripping is a heavier accusation, describing transactions with little real commercial purpose beyond inflating each side’s numbers.
Supplier financing is not automatically a red flag. It can be the rational way to seed a market that lenders do not yet understand.
Financing can be commercially sensible when:
New infrastructure needs unusually large upfront capital.
Customers have little credit history.
Independent lenders stay cautious about an unproven market.
The supplier reads the technology and its likely demand better than a traditional bank.
Used this way, financing can build a market that later funds its own growth through operating cash flow or independent credit. The signal turns negative when each new order needs a bigger cheque: larger guarantees, fresh investment or richer incentives just to close. The dividing line is simple to state and hard to game. Does supplier support shrink as the customer matures, or does it become a permanent condition for growth?
No single number proves that demand is propped up. The pattern shows up across several financial signals at once.
What to check |
Warning signal |
What it can reveal |
|---|---|---|
Operating cash flow |
Cash lags revenue |
Sales are turning into cash more slowly |
Receivables |
Receivables grow faster than sales |
Easier payment terms may be propping up orders |
Customer investments |
Supplier funding rises alongside purchases |
Supplier cash may be returning as revenue |
Guarantees and backstops |
Maximum exposure widens |
Credit or utilisation risk has shifted to the seller |
External usage |
Demand lags new capacity |
Capacity is growing faster than paying use |
Rising supplier support paired with weakening cash conversion is the combination that deserves the closest look. No metric alone proves manipulation, but several deteriorating together can show that reported growth leans more and more on financial support.
Nvidia is the counterexample to any claim that supplier support automatically signals weak demand. Its operating cash flow stays strong relative to revenue, a sign that it keeps converting most of its sales into cash, according to its own filings.
Yet its investments, commitments and customer concentration still deepen its exposure to the very companies driving future orders.
The proposed Nvidia-OpenAI structure is straightforward in outline: outside lenders finance a large OpenAI data-centre project, Nvidia backs part of that financing, and the finished site runs on computing systems built around Nvidia hardware.
According to The Wall Street Journal, Nvidia is discussing a roughly $250 billion financial backstop for a proposed 10-gigawatt OpenAI data-centre project in southern Ohio. The support would reportedly sit behind financing vehicles tied to lease and construction debt rather than the chips themselves.
Separately, Nvidia is said to be weighing financing that could support as much as $350 billion of OpenAI chip purchases. Terms are not final, and the proposal could still fall through.
A guarantee would not send $250 billion out of Nvidia’s door on day one. It would place Nvidia’s credit behind specified obligations, creating losses only if the relevant borrowers or lease vehicles failed to pay. That is the mechanism that turns a supplier’s balance sheet into part of the financing behind its own future sales.
Circular financing does not always look like a loan or a guarantee. Nvidia’s CoreWeave agreements and AMD’s OpenAI warrant show two quieter versions of the same idea.
Nvidia invested $2 billion in CoreWeave in January 2026. A separate agreement, worth $6.3 billion at the outset, lets Nvidia buy qualifying cloud capacity that CoreWeave has not sold to other customers, and it runs to April 2032. The structure leaves Nvidia exposed to CoreWeave’s build-out and to a slice of its unused capacity, even if CoreWeave still finds plenty of outside buyers.
AMD granted OpenAI warrants to buy up to 160 million AMD shares at $0.01 each, with vesting tied to purchases that scale from an initial one-gigawatt deployment to six gigawatts of AMD GPUs. The order can be real. The equity handed over to win it is still part of its cost.
The late-1990s telecom bust is the clearest precedent, and its lesson is about credit, not technology. Equipment makers lent young network operators the money to buy their gear.
When credit markets tightened, some of those operators stopped ordering and could not repay what they had already borrowed. Lucent later disclosed that it wrote off customer financings, sold others at steep discounts and set aside large reserves against what remained. Product sales and customer credit had become two sides of the same bet.
The internet survived. Many balance sheets built on always-available financing did not. The takeaway travels well: a supplier can face falling product demand and losses on the financing used to prop up that demand at the same moment.
Circular financing is when a supplier funds a customer, and the customer uses that funding to buy the supplier’s products. The money makes a loop: it leaves the supplier as investment or credit and returns as revenue.
No. Loans, guarantees and strategic investments are ordinary commercial tools. The trouble starts when a company hides concessions, overstates how much it will collect, or books deals with no real economic substance.
Not by itself. Products can be delivered and revenue recorded correctly. The real question is how much of the customer’s buying power came from the supplier.
Vendor financing usually means direct credit, loans or guarantees. Circular financing is wider, folding in equity stakes, warrants, capacity commitments and other arrangements that help fund customer purchases.
Not necessarily, and circular financing alone does not settle it. A technology can prove genuinely important even if some of its infrastructure is overbuilt, badly financed or bought at prices that never earn an adequate return.
Circular financing comes down to one question: who carries the risk? It does not by itself create fake revenue or prove that a market is a bubble. What matters is whether customers can eventually fund their purchases through their own cash flow, independent financing and real end-user demand rather than continued supplier support.
Cash conversion, receivables, customer funding, guarantees and unused capacity are the gauges that show whether sales are becoming self-sustaining or still lean on the seller.
The question is not whether the sale is real. It is whether the customer can keep buying once the supplier stops helping to pay.
