Assess an AI company’s financing and revenue as one connected network, not as separate headline numbers. Trace who supplies capital, who buys the product, who ultimately uses it, whether the company collects cash for delivered services, and who bears the risk if demand or funding falls away. Then measure concentration separately across revenue, receivables, contracts, financing, suppliers, and committed capacity. A link between an investor, supplier, and customer is a reason to examine the terms—not, by itself, proof of misconduct.
What circular financing means in practice
“Circular financing” is most useful as a description of a specific funding-and-purchase loop, not as a label for every strategic investment or customer relationship. For example, a company may provide funding, credit, or another form of support to a customer that then uses that support to buy the company’s products or services. The diligence question is whether the customer’s demand and ability to pay are independently supported, or depend materially on capital supplied by the seller, its investor, or a connected party.
As an Amazon Associate I earn from qualifying purchases.
AI commercial arrangements can involve several parties: an AI developer, its investors and lenders, a cloud provider, a chip or hardware supplier, an AI cloud, a direct buyer, and an end user. Capital and commercial commitments may run in different directions. Equity, debt, warrants, guarantees, cloud credits, prepayments, vendor financing, purchase commitments, capacity reservations, and revenue shares can each affect who ultimately funds demand and who bears losses.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minute- Connected transactions are not automatically improper. Examine contract terms, services delivered, payment conditions, collectability, cash settlement, and risk transfer before drawing a conclusion.
- Accounting revenue is not the same as cash collection. Compare recognized revenue with invoices, receipts, aging receivables, credit losses, deferred revenue, advances, and any noncash consideration.
- A direct buyer may not be the ultimate user. A cloud provider or reseller can sit between the AI supplier and the end customer, obscuring whose demand supports the sale.
Map the parties and follow the money
Start with a relationship map of material counterparties and, where available, their beneficial owners. For each entity, note whether it is an investor, lender, customer, supplier, reseller, cloud operator, or end user. Mark confirmed ownership or financing links separately from unknown ones; do not assume two entities are independent merely because they have different names.
Draw each material flow in both directions. Include equity and debt funding, convertible instruments and warrants; loans, guarantees and cloud credits; hardware purchases and cloud services; prepayments and capacity reservations; purchase or offtake commitments; and revenue-sharing terms. Record the payer and payee, amount or range, timing, recourse, termination rights, conditions to payment, and related performance obligations.
Then ask whether a single party appears on multiple sides of a commercial loop. A strategic investor might also lend to a customer; a supplier might commit to buy unsold capacity from a cloud partner; a customer might receive a loan connected to infrastructure needed to serve its own contracted demand. These structures can transfer or concentrate demand risk in ways that headline revenue alone will not show.
Check whether sales are independent and collectible
Identify the ultimate customer
Ask who pays the AI company, who uses the service, and whether those are the same entity. If a direct customer is a cloud provider or reseller, seek the end-customer mix and the basis for attributing indirect sales. Public disclosures may estimate indirect exposure without naming the underlying buyer. NVIDIA’s FY2026 Form 10-K, for the fiscal year ended January 25, 2026, says it estimates some indirect customer exposure using purchase-order information, product specifications, internal sales data, and other sources; it also reported that an AI research and deployment company contributed a “meaningful amount” of revenue through cloud services purchased from NVIDIA customers, without naming that end customer or quantifying the amount. Read NVIDIA’s FY2026 Form 10-K.
A named direct buyer therefore does not establish that end demand is diversified or independent. Where the ultimate buyer is undisclosed, treat attribution as limited rather than filling the gap with an assumption.
Reconcile revenue to cash and delivery
For each important customer arrangement, compare recognized revenue with invoices and actual cash receipts. Review receivables aging, credit losses, deferred revenue, advances, and noncash consideration, including customer-related warrants. Check whether the revenue corresponds to services actually delivered and whether the price or the customer’s ability to pay depends on funding supplied by the seller, its investor, or a related party.
Contract-specific terms and the applicable accounting guidance matter. The disclosures discussed here do not establish one accounting rule that resolves every connected financing arrangement; a reviewer needs the underlying contract and company-specific accounting analysis.
Measure concentration by exposure type
Calculate concentration for comparable periods and consistent denominators. For revenue, record the largest customer’s share and the combined share of the top three and top five. Separately measure receivables, cash collections, bookings or remaining performance obligations, financing sources, suppliers, data-center capacity, and committed purchases. Note whether customers are named or anonymized, whether related entities are grouped, and whether a direct buyer is only an intermediary.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Do not combine quarterly and annual percentages as though they describe the same period. A large revenue share can signal reliance on a buyer, while a large receivables share signals that cash collection is especially exposed to a particular payer. Backlog and commitments add a forward-looking view, but do not by themselves establish that a customer will pay or that capacity will be used.
| Disclosure example | What was reported | What it helps assess |
|---|---|---|
| NVIDIA, Form 10-Q for the quarter ended July 26, 2026 | One direct customer accounted for 16% of second-quarter revenue. Three direct customers accounted for 16%, 15%, and 13% of first-half revenue. NVIDIA also said some indirect customers may each represent at least 10% of revenue; indirect attribution is estimated, and one AI research and deployment company was described as contributing a “meaningful amount” through cloud-service purchases, without a quantified share. | Direct-customer concentration is measurable for the disclosed periods, but indirect exposure can remain uncertain when end customers are not identified or quantified. |
| NVIDIA, Form 10-K for the fiscal year ended January 25, 2026 | Direct customers accounted for 22% and 14% of annual revenue, respectively; the filing also noted estimated indirect exposure through cloud services. | Annual concentration gives a different period view from the later quarterly and first-half figures; the periods should not be pooled. |
| Cerebras Systems, 2026 prospectus (figures for 2025) | G42 accounted for 24.0% and MBZUAI for 62.0% of revenue. The prospectus identifies G42 and MBZUAI as related parties with respect to each other under ASC 850. One customer accounted for 77.9% of accounts receivable as of December 31, 2025. | Revenue concentration and receivables concentration are separate risks. The receivables percentage is a balance at a particular date, not a revenue share. |
The NVIDIA customer figures are issuer disclosures in its Form 10-Q for the quarter ended July 26, 2026 and Form 10-K for the fiscal year ended January 25, 2026. Cerebras’s figures are in its 2026 prospectus. These examples illustrate different risk channels; they do not establish an industry-wide concentration rate or threshold.
Examine financing tied to a customer or supplier
A financing arrangement becomes particularly important to diligence when repayment, delivery, or the customer’s purchasing ability is tied to the same commercial relationship. Cerebras’s 2026 prospectus describes an OpenAI collaboration that included a secured working-capital loan of approximately $1.0 billion funded by OpenAI in January 2026 to support infrastructure and capabilities needed to provide compute services OpenAI had contracted to purchase. The prospectus also describes a warrant. That is an example of connected lender-and-customer exposure to analyze; the disclosure alone does not establish that the arrangement is improper.
For any similar structure, determine whether the loan is recourse or secured, what triggers repayment, how the warrant affects economics, what happens if the customer reduces or delays purchases, and whether the company could still meet obligations if the customer’s own funding weakened. The key is to identify the actual loop: which capital is advanced, what purchases it supports, and where repayment or losses ultimately land.
Assess capacity commitments and who carries utilization risk
Capacity arrangements can put a supplier or investor on both sides of demand risk. NVIDIA’s Form 10-Q for the quarter ended July 26, 2026 reported $36 billion in AI-cloud commitments as of that date, typically with six-year terms. The filing describes multi-year agreements with AI clouds, possible revenue sharing on third-party sales, and commitments to purchase capacity those clouds do not sell to third parties. It says commitments reduce as third-party customers or NVIDIA’s research-and-development use of capacity increases. These are NVIDIA-specific disclosed commitments, not an industry benchmark, and the filing does not label them circular financing.
For a capacity contract, identify who owns the equipment, who must pay for reserved capacity, whether there is a minimum purchase or take-or-pay obligation, and whether the supplier can repurpose unused capacity. Also establish who bears residual-value, pricing, execution, and customer-performance risk if utilization disappoints. NVIDIA’s filing says agreements with AI clouds are intended to enable broader access to its data-center infrastructure products; that is the company’s description of the arrangements, not an independent assessment of their risk.
See the relevant NVIDIA Form 10-Q disclosure for the commitment and contract context.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Stress-test the largest exposure
Model more than a simple customer loss. Test the consequences of delayed payment, a contract dispute, renewal failure, and lower utilization by the largest customer or end user. Consider secondary effects on debt covenants, lease payments, GPU depreciation or obsolescence, minimum-purchase obligations, cloud take-or-pay terms, capacity buybacks or guarantees, and future capital needs.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchAsk who bears the shortfall and whether the AI company has a practical alternative use for the capacity. Include a correlated downside case: the same investor or lender that funds the company may also support the customer’s ability to buy. If that funding weakens, both demand and collection could deteriorate together.
Compare arrangements on the same risk dimensions
When comparing two or more customer or financing arrangements, use the same period and examine the following dimensions. A company with one large but creditworthy customer paying cash under short, cancellable commitments may have a different risk profile from a company with more nominal customers whose purchasing power depends on one connected funding ecosystem.
| Dimension | Question to ask |
|---|---|
| End-demand independence | Is the end user identified, and is its demand independent of funding from the seller, investor, or supplier? |
| Credit quality | Can the customer and the party funding it pay through a downturn? |
| Revenue and receivables exposure | What share of revenue and outstanding receivables depends on the customer or connected group? |
| Cash versus accounting revenue | How much recognized revenue has been collected in cash, and on what timetable? |
| Term and flexibility | How long does the commitment run, and what are the termination and renewal rights? |
| Recourse and residual risk | Are there guarantees, purchase-back obligations, or other terms that return demand or asset risk to the company? |
| Utilization | Who absorbs the cost if reserved compute capacity is unused, and can it be redeployed? |
| Related-party links | Do customers, lenders, suppliers, or investors share ownership, control, financing, or strategic direction? |
| Disclosure quality | Are counterparties, end users, amounts, conditions, and cash flows disclosed clearly enough to test the exposure? |
State the conclusion precisely
Describe what the evidence establishes and what remains unknown. Useful bounded descriptions include “concentrated revenue,” “high receivables dependence,” “linked financing and demand exposure,” or “limited end-customer transparency.” If using “circular financing,” explain the actual flow of capital and how it returns through purchases, revenue, or repayment rather than using the phrase as a synonym for a supplier investment or strategic customer relationship.
There is no universal customer-concentration percentage in the cited disclosures that defines a company as safe or unsafe. Apply any covenant or internal limit that is relevant to the specific company, and do not substitute an invented industry threshold. Public filings are issuer disclosures: they show what companies report, but do not independently verify every commercial assertion or settle every question about private terms, side letters, ownership links, end-user payment paths, or utilization.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
