Today, I am returning to a topic I have been writing about for a long time at MacroXX: circular financing in the AI industry.
In my November 2025 article, “Circular financing in AI: Funding loops and stock implications,” I explained how companies developing AI, supplying computer chips, and providing computing services can also help finance one another. They are not always just buyers and sellers. Sometimes, they are investors in their own customers.
Circular financing in AI
Circular financing has become a defining feature of the AI ecosystem, weaving together chipmakers, cloud providers, platform operators, and AI developers in a self-reinforcing funding loop. This dyna…
These relationships can help the industry grow faster. But they raise an important question: Is that growth supported by customers who can pay for AI services—or does it depend on companies continuing to raise money and pass it around?
The MacroXX approach
At MacroXX, my goal is simple: to make economics and financial markets easier to understand. Too often, market commentary feels written only for professionals—filled with unfamiliar terms, technical charts, and reports that seem disconnected from everyday life.
I take a different approach at MacroXX: clear explanations, practical examples, and a focus on what economic developments mean for workers, households, businesses, and investors.
The AI financing story is a good example. It is not just about technology stocks. It is also about retirement savings, electricity bills, local construction, and who could face losses if the boom falls short of expectations.
What is circular financing?
Imagine that a company selling computer chips invests money in an AI startup.
The startup then uses some of that money to buy chips from the company that invested in it.
The chipmaker has helped fund its own customer. Some of its investment has come back as sales.
A similar arrangement can happen when a company providing computing services invests in an AI developer, which then pays to use the investor’s services.
That is the basic idea behind circular financing: money flows into a business as funding, then flows back to the funding provider through purchases.
This does not automatically mean anything improper is happening. A promising new business may need financial support before it can stand on its own.
The important question is what happens next. Does the startup eventually earn enough from its own customers to pay its bills? Or must it keep raising more money to keep buying?
A simple example: the bakery and the oven supplier
Suppose an oven manufacturer invests in a new bakery. The bakery uses part of that investment to buy the manufacturer’s ovens.
Both businesses benefit. The bakery gets equipment, and the manufacturer makes a sale.
But selling ovens to the bakery does not prove that the bakery will succeed. That depends on whether enough people buy its bread.
If the bakery builds a profitable customer base, the arrangement may work well for everyone. If it keeps losing money and needs repeated investments to survive, the oven manufacturer’s sales depend partly on continued financial support.
AI raises a similar question on a much larger scale.
Buying chips and renting computing power are steps toward building a business. They are not, by themselves, proof that the business can support itself.
Not every AI deal is a financing loop
In my earlier MacroXX article, I discussed relationships involving Nvidia, OpenAI, Oracle, CoreWeave, and major cloud providers.
But these relationships should not all be placed in the same category.
Buying a stake in a company is different from lending it money. A promise to purchase computing services is different from a completed payment. Borrowing to build a data center does not, by itself, make the financing circular.
To understand a deal, I focus on three questions:
Who is providing the money?
Does the recipient use that money to buy from the provider?
Where will the money to repay lenders or reward investors ultimately come from?
The concern is not that every AI transaction is circular. It is that several companies may appear to be growing independently while relying on the same funding sources and the same hopes for future AI sales.
Why data centers make this a bigger story
AI depends on physical infrastructure.
Behind the software are computer chips, buildings, cooling systems, power lines, and electricity supplies. Building and operating all of this costs money—often before anyone knows how much profit it will eventually produce.
The spending chain can look like this:
Investors provide money → AI companies buy computing services → computing providers build more capacity → more financing is raised to support further expansion.
This can be a healthy growth process if paying customers eventually generate enough revenue to cover the costs.
But if customer revenues remain too low, the chain may depend on a continuing supply of new investment and loans.
At MacroXX, I am not questioning whether the spending is real. The buildings, equipment, and electricity bills are real.
I am asking whether the businesses using them will earn enough to justify the expense.
The building may last longer than the equipment
A data-center building can remain useful for many years. The computer chips inside it face a different challenge: newer and more efficient models can change what customers are willing to pay to use older equipment.
That matters when the equipment is financed with borrowed money.
Imagine taking out a loan to buy machinery. Before the loan is paid off, a better machine becomes available. Your machine still works, but customers may prefer the newer version—or expect a discount to use yours.
You still owe the money, even if your equipment earns less than expected.
Data-center owners also face questions about their customers. A major customer may sign a contract to rent computing capacity for several years. But if the owner’s debt lasts longer than the contract, someone must consider what happens afterward.
Will the customer renew? Will another customer take its place? At what price?
A reliable customer today does not remove every risk tomorrow.
Different stocks can carry the same risk
An investor might own shares in a chipmaker, a cloud provider, and a data-center company and feel well diversified.
Yet all three investments could depend heavily on the same outcome: continued growth in AI spending.
If that spending slows, the effects could move through the chain:
AI developers reduce their computing purchases.
Computing providers delay expansion.
Chipmakers receive fewer orders.
Highly indebted businesses find it harder to meet payments or replace old loans with new ones.
This does not mean every company would suffer equally. Some have stronger finances, more customers, or other profitable businesses.
But owning several company names is not necessarily the same as owning several independent sources of return.
That is an important distinction for MacroXX readers evaluating their exposure to AI.
Why this matters even if you do not own AI stocks directly
Households can have indirect exposure through retirement accounts, pension funds, or investment funds that hold shares and debt issued by companies involved in the buildout.
Communities face a different kind of exposure.
A proposed data center may bring construction jobs and tax revenue. It may also require expanded electricity infrastructure and additional public services.
The question is what happens if the project is smaller than promised, opens late, or shuts down.
If a utility builds new infrastructure expecting a large customer to use it, who pays if that demand disappears?
The answer depends partly on the contracts and rules in place. Requirements for minimum payments or fees for leaving early can help keep costs with the company that requested the expansion.
Without adequate protections, some costs could fall on other customers.
This is not circular financing itself. It is a related issue: how the risks of the AI buildout are shared beyond the companies making the investment decisions.
Popular does not necessarily mean profitable
A service can attract millions of users without earning enough to cover its costs.
Free users show that people are interested. Paying users show that some are willing to spend money. But a sustainable business must collect enough revenue to pay for computing, electricity, employees, equipment replacement, and financing.
That is why “AI usage is growing” and “AI investment will produce strong returns” are not the same statement.
Both could turn out to be true. But the first does not guarantee the second.
The same applies to funding loops. A purchase financed by an investor is a real sale. What remains uncertain is whether the buyer can keep making those purchases once it must rely more heavily on its own earnings.
What I watch at MacroXX
Rather than focusing only on announcements of bigger investments and larger contracts, I ask a few practical questions.
1. Where is the customer getting the money?
Is the customer paying from money earned by its business? Or does it depend on new investments and loans?
Outside funding is normal for growing companies. The issue is how long that dependence lasts and whether there is a credible route to covering costs.
2. Is the announcement a promise or a payment?
A proposed investment, a conditional agreement, and money already spent are different things.
Large headline numbers deserve a closer look at the terms.
3. Does the business depend on only a few customers?
A company can report strong sales while relying heavily on one or two buyers.
If those buyers cut spending or face financial trouble, the supplier may feel the impact quickly.
4. Will the equipment earn enough before it needs replacing?
The purchase price is only part of the calculation. The business must also consider operating costs, loan payments, and the possibility that newer equipment will reduce what customers pay for older machines.
5. What happens if borrowing becomes harder?
A company that regularly needs new loans may face trouble if lenders demand higher interest rates or become less willing to lend.
6. Is money coming in from outside the financing network?
Ultimately, households and businesses must be willing to pay enough for AI products and services to support the system.
That is the clearest test of whether the industry is becoming financially self-sustaining.
AI can succeed while some investments disappoint
There is one final distinction I keep coming back to at MacroXX: a useful technology is not automatically a good investment at any price.
Infrastructure can benefit society while some of the investors who financed it lose money. Companies can build too much, borrow too heavily, or pay more than future profits justify.
AI could become deeply embedded in everyday life while some businesses involved in today’s expansion still disappoint their shareholders and lenders.
My concern is not that AI has no value. It is that excitement about the technology can overshadow questions about how the bills will be paid.
Circular financing can help businesses grow. The real test comes when those businesses must rely less on one another’s funding and more on money earned from customers.
When the money stops circulating within the AI ecosystem, how much is coming in from outside it?
This newsletter is for informational and educational purposes only—not financial advice. Use of this information is at your own risk, and the authors are not responsible for any resulting losses or damages.
The key question you raise, whether AI growth is backed by customers who can pay or by suppliers investing in their own buyers, is the right lens for reading chipmaker revenue right now. Vendor financing tends to look fine until funding conditions tighten, so the share of revenue tied to these loops seems worth tracking.