AI Circular Financing: Is the Same Money Going Around?

AI companies keep announcing enormous deals, and every one is presented as more proof that demand is unstoppable.

Look closer, though, and the same companies keep appearing on both sides of the money.

Nvidia sells the chips. It also invests in the AI companies buying access to those chips. It owns part of CoreWeave, which buys Nvidia hardware, and has agreed to purchase CoreWeave capacity if other customers do not. Now it is reportedly considering guaranteeing roughly $250 billion in financing connected to a huge OpenAI data centre project.

None of this automatically makes the deals fake. The chips exist, the data centres are being built and the computing power is being used.

But it raises a fair question. How much of the AI boom comes from independent customer demand, and how much comes from the biggest companies financing one another so the spending can continue?

That is the circular-financing problem, and investors have finally started paying attention.

Quick Answer

Yes, circular financing is happening in AI.

Nvidia sells the hardware, invests in some of the companies buying it and, in certain deals, helps protect the financing or unused capacity behind those purchases. The revenue can be real while the demand is still less independent than the headline numbers suggest.

That does not prove the AI boom is fake. It means several supposedly separate success stories may depend on the same companies, the same customers and the same borrowed money.

What Is Circular Financing in AI?

Circular financing happens when companies inside the same business chain fund one another and then buy from one another.

The basic AI version looks like this:

1. A chip company invests in an AI lab or cloud provider.
2. That company uses its new capital to buy computing power.
3. The data centre providing the computing power fills its servers with chips from the original investor.
4. The chip company reports stronger demand and invests more money across the same market.

The products and revenue can be real. The concern is that the sales are not as independent as they first appear. If a supplier invests in customers, guarantees their financing or promises to buy unused capacity, it becomes harder to tell where outside demand ends and supplier-supported demand begins.

Vendor financing is not automatically suspicious. The problem is treating every connected deal as separate confirmation that AI demand is booming.

Why Is Nvidia at the Centre of It?

Because Nvidia is no longer only selling chips into the AI boom.

It is also investing in the labs, owning shares in the cloud providers and discussing financial guarantees for the data centres that will be filled with its hardware.

In February 2026, OpenAI announced $110 billion in new investment, including $30 billion from Nvidia. The announcement also said OpenAI had secured next-generation inference computing capacity using Nvidia systems.

That does not mean OpenAI sent the exact same money back. It means Nvidia invested in a major customer while securing a role in that customer’s next round of infrastructure spending.

Nvidia has made similar moves elsewhere. In January, it invested another $2 billion in CoreWeave, making it the cloud company’s second-largest shareholder at the time.

CoreWeave uses large numbers of Nvidia GPUs.

Nvidia benefits from CoreWeave’s growth as a shareholder, hardware supplier and now something else too: a fallback customer.

The CoreWeave Deal Is the Clearest Example

CoreWeave disclosed a $6.3 billion agreement with Nvidia in September 2025.

Under that agreement, Nvidia is required to buy residual CoreWeave cloud capacity that other customers do not use, subject to the contract’s conditions, through April 2032.

Read that again.

Nvidia supplies CoreWeave with GPUs. Nvidia owns shares in CoreWeave. If CoreWeave cannot sell all of its available computing capacity to other customers, Nvidia has agreed to buy the unused part covered by the deal.

CoreWeave gets more certainty before spending billions on infrastructure, while Nvidia secures capacity and helps a major customer expand. Fair enough.

But this is not the same as CoreWeave building entirely against independent demand. If other customers do not want all the capacity, Nvidia may still become the customer.

Microsoft, Nvidia and Anthropic Made Another Loop

The Anthropic deal is less direct, but the pattern is familiar.

In November 2025, Microsoft and Nvidia announced plans to invest up to $5 billion and $10 billion respectively in Anthropic.

As part of the same partnership, Anthropic committed to purchase $30 billion of Microsoft Azure computing capacity and contract up to one additional gigawatt. That Azure capacity runs on Nvidia systems.

So Microsoft and Nvidia agreed to put up to $15 billion into Anthropic. Anthropic committed to spend $30 billion with Microsoft, using infrastructure powered by Nvidia.

It may be good business for all three. Anthropic needs computing power, Microsoft wants Claude on Azure and Nvidia wants more workloads on its hardware.

But the investment and spending commitment are not two unrelated votes of confidence. They are parts of the same deal.

The Reported $250 Billion OpenAI Guarantee

The latest OpenAI report pushed this concern from an interesting accounting debate into the centre of the market.

Nvidia is reportedly in talks to provide roughly $250 billion in financing guarantees for a 10-gigawatt data centre project in southern Ohio.

The site is being developed by SB Energy, a SoftBank subsidiary, and OpenAI is considering leasing it. The total project could cost more than $500 billion once the chips are included.

The reported $250 billion guarantee would cover the lease and debt financing, not the Nvidia hardware. Separate discussions could involve financing up to another $350 billion of OpenAI chip purchases.

These are reported talks, not a completed agreement. Reuters said it could not independently verify the original Wall Street Journal report, and Nvidia, OpenAI and the US Commerce Department did not immediately comment.

So this is not an actual $250 billion liability yet. Even so, the structure being discussed is extraordinary.

Nvidia would sell much of the hardware inside the project. Its financial backing could help OpenAI secure the site and reassure lenders. If the separate chip-financing plan also happened, Nvidia could help finance the purchase of its own products as well.

Nvidia would not simply be waiting for a customer with enough money to buy its chips. It would be helping that customer obtain the money needed to keep buying.

Does Circular Financing Inflate AI Revenue?

It can make demand look stronger and more independent than it really is. That is not the same as saying the revenue is fake.

If CoreWeave buys Nvidia GPUs, those chips are real. If Anthropic uses $30 billion of Azure capacity, Microsoft is providing a real service. If OpenAI leases a data centre, somebody has still built the site and supplied the computing power.

The question is what happens without the financial support surrounding those transactions.

Would the customer buy the same amount at the same speed if the supplier had not invested in it?

Would the data centre build as much capacity if the chip company had not promised to buy what remained unused?

Would lenders finance the project on the same terms without Nvidia standing behind it?

Those questions do not erase the revenue. They make it harder to assume the same demand will continue without support.

Why Can the AI Money Loop Become a Problem?

The main danger is concentration.

Nvidia, OpenAI, Microsoft, CoreWeave, Anthropic and a small number of hyperscalers appear across an enormous share of the industry’s biggest deals. One company can be a supplier in one agreement, an investor in another and a customer or guarantor in the next.

If AI revenue grows quickly enough, the data centres fill and the companies financing the buildout make a fortune.

If it disappoints, an AI lab cuts orders, the cloud provider loses revenue, the data centre has unused capacity and the chip supplier gets hit through weaker sales plus its investments, commitments or guarantees.

Several positions that looked diversified turn out to depend on the same customer continuing to spend. The broader concentration risk became painfully visible when Leopold Aschenbrenner’s leveraged AI fund lost 67% in one month and sold most of its public-equity portfolio to Citadel. That was a hedge-fund collapse rather than the corporate money loop itself breaking, but it showed how quickly concentrated AI exposure could unravel once leverage forced the selling.

The debt market is already becoming less relaxed about the wider AI buildout. Amazon, Alphabet, Meta and Oracle issued about $194 billion in bonds through July 7, 2026, up 79% from their combined issuance in all of 2025.

Investor demand remains substantial, but it has weakened as supply has grown. Reuters reported that cover ratios for hyperscaler bond sales fell from nearly five times in February to below two times in July, while borrowing spreads widened.

This is not proof of a crash. It shows that the money is becoming more expensive and investors are asking harder questions.

Is Circular Financing Why Nvidia Stock Fell?

It was one reason, but not the only one.

Nvidia shares fell 5% on July 27 after the reported OpenAI guarantee brought circular-financing concerns back into focus.

The wider chip sector was under pressure too. Investors were also worried about growing Chinese competition, the scale of AI infrastructure spending and major technology earnings due later that week.

So it would be lazy to blame the entire fall on one OpenAI report.

The market reaction still matters. Investors had largely celebrated Nvidia putting money into its customers because those customers then ordered more computing power. The possibility of a $250 billion backstop made the other side of that strategy harder to ignore.

Financing your customers can protect future sales. It also means taking on some of the risk that used to belong to the customer or the lender.

Should Nvidia Investors Be Worried?

Yes, but one 5% fall is not a serious reason by itself to sell anything.

The useful questions are much more boring:

• How much Nvidia demand comes from customers that can finance themselves?
• How much exposure does Nvidia have through investments, guarantees and capacity commitments?
• What happens if OpenAI, CoreWeave or another major buyer slows its spending?
• Can AI companies generate enough revenue to support the infrastructure already being ordered?

Nvidia still sells the leading hardware behind most advanced AI systems, and it has enormous financial strength. It can use that balance sheet to create more customers, infrastructure and chip sales. It can also end up carrying several layers of exposure to the same spending boom.

Investors do not need to call it fraud or sell every technology stock.

They do need to stop treating every Nvidia-backed investment, cloud contract and data centre announcement as independent proof that demand is unstoppable.

Sometimes the new money entering the AI boom is the same money coming back around.