LOOKING BACK | Around We Go: Circular Finance and AI

"Yes, of course! Hence the name: movement. It moves a certain distance, then it stops, you see? A revolution gets its name by always coming back around in your face." - William Strannix, Under Siege. 

We'll go ahead and acknowledge that we've lead with a quote from a Steven Segal movie that was probably a restatement of a line from a Terry Pratchett book, so we're already off to an unusual start, but in this Looking Back column we wanted to have ChatGPT discuss the notion of circular finance, not only to provide an explainer of what circular finance is and how it plays out in the financial industry, but also of the risks and benefits it carries.  

When things go wrong, circular finance can lead to problems very much like Strannix's (or Pratchett's) revolution—they just keep coming around again, and again, and again.  

As Microsoft, Nvidia, Google, Amazon and other technology giants pour billions into artificial intelligence companies that subsequently purchase their products and services, investors and regulators are questioning whether the industry's extraordinary growth reflects genuine demand or a self-reinforcing financial ecosystem. Circular financing may be accelerating the AI revolution, but could it also be inflating an investment bubble? 

The artificial intelligence industry has developed a remarkable financial arrangement: Some of its largest companies are investing billions of dollars in their customers, which then spend enormous sums purchasing products and services from those same investors. Microsoft, Nvidia, Google and Amazon are simultaneously financiers, suppliers and commercial partners to AI developers, cloud infrastructure providers and data center operators. This arrangement, known as circular financing, has become one of the most controversial features of the AI investment boom. Supporters argue that it is a legitimate and necessary mechanism for financing an extraordinarily capital-intensive technological transformation. Critics worry that it artificially amplifies demand, obscures financial risks and inflates valuations across an increasingly interconnected industry. With trillions of dollars committed to AI infrastructure, the distinction could have profound consequences for investors, financial institutions and the broader economy. 

What Is Circular Finance? 

Circular financing describes an arrangement in which money provided by one company to another eventually returns to the original financier through commercial transactions. 

Imagine a semiconductor manufacturer investing $10 billion in an AI startup. The startup uses that investment to purchase $10 billion worth of computing hardware from the manufacturer. The semiconductor company receives revenue, the startup acquires equipment, and both companies may benefit from rising valuations. 

The manufacturer has effectively financed demand for its own products. 

These arrangements can involve equity investments, loans, convertible securities, cloud credits, purchase commitments, lease guarantees and other financial instruments. The money does not necessarily return directly or immediately to its original source, and the commercial commitments may exceed the initial investment. 

Circular financing is closely related to vendor financing, in which suppliers extend credit or otherwise help customers finance purchases. Vendor financing has long existed in industries such as telecommunications, manufacturing, transportation and construction. 

However, an important distinction is necessary. Circular financing is not the same as circular finance associated with the circular economy. 

Organizations including the United Nations Environment Programme Finance Initiative (UNEP FI) use circular finance to describe financing that supports resource efficiency, recycling, reuse, refurbishment and waste reduction. That definition concerns the circulation of physical resources rather than the recycling of investment capital. UNEP FI has encouraged financial institutions to support circular business models through lending, investment and insurance.  

The two concepts share terminology but represent fundamentally different economic activities. AI-related circular financing is primarily about capital flows, commercial relationships and financial risk. 

How Circular Financing Works in AI 

AI development requires enormous upfront investment in semiconductor chips, data centers, electricity, networking equipment, software and engineering talent. 

The companies developing frontier AI models frequently lack the financial resources to independently construct the infrastructure required to train and operate their systems. 

Meanwhile, the largest technology companies possess considerable capital, infrastructure and strategic incentives to support the emerging AI ecosystem. 

This creates an opportunity for mutually beneficial financing arrangements. 

Microsoft's relationship with OpenAI provides one of the clearest examples. Microsoft invested more than $13 billion in OpenAI through its longstanding partnership, while OpenAI became a major customer of Microsoft's Azure cloud platform. 

Microsoft benefits from OpenAI's technological development, commercial success and demand for computing resources. OpenAI receives financial support and access to infrastructure that would otherwise be extraordinarily expensive to develop. 

Similarly, Amazon and Google have invested in Anthropic, while also providing cloud infrastructure and computing capacity to the AI developer. 

Nvidia has invested in AI infrastructure companies, including CoreWeave, that purchase and operate Nvidia's graphics processing units. 

These relationships create interconnected financial and commercial networks rather than simple bilateral transactions. 

A chipmaker may invest in a cloud operator, which borrows money to acquire chips, leases computing capacity to an AI developer, and relies on that developer's future revenues to service its debts. 

If everything works as expected, the participants benefit from a growing market. If demand disappoints, financial difficulties can propagate through multiple companies simultaneously. 

The BIS findings are particularly revealing. Its researchers also determined that 55.2% of incoming investments in AI firms between 2021 and 2025 originated with other AI companies. 

Those statistics do not establish that AI demand is artificial. They do, however, demonstrate how deeply the industry's investors and suppliers are interconnected. 

LSEG's October analysis illustrates the complexity through Anthropic's relationships with Google, Broadcom and infrastructure financing entities. 

One arrangement involves a special-purpose vehicle purchasing approximately $35 billion in Google-designed tensor processing units, with financing involving Apollo and Blackstone. Broadcom provides credit support, while other companies participate in data center development and infrastructure leasing. 

Separately, Google's financial backstops for infrastructure operator Fluidstack's obligations to TeraWulf demonstrate how credit exposure can extend beyond the original customer-supplier relationship. 

LSEG also highlighted Nvidia's announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish AI infrastructure financing platforms targeting more than $500 billion in third-party capital. These are proposed financing arrangements, not evidence that the entire amount has already been raised or deployed.  

The emerging system resembles an elaborate financial network in which ownership, revenue, borrowing, collateral and guarantees can be distributed across numerous entities. 

Real Problems: Artificial Demand, Hidden Debt and Inflated Valuations 

The central criticism of circular financing is that it can make an industry appear healthier than its underlying economics justify. 

When an investor finances a customer's purchases, the supplier may recognize legitimate revenue even though the customer's ability to continue purchasing depends on additional investment. 

That distinction is critical. Circular financing does not automatically constitute fraudulent accounting or fabricated sales. A cloud provider delivering computing services has provided something of economic value, regardless of whether its customer obtained funding from the provider itself. 

However, investors must distinguish revenue generated by independently profitable customers from revenue supported by repeated financing rounds. 

The distinction becomes particularly important when investors value AI companies using revenue multiples, projected growth rates or long-term cash flow assumptions. 

Research firm IDC addressed this problem in a June report titled Circular Financing Has Muddied the AI Story: Watch the Application Layer Instead. 

IDC executive Eric Newmark argued that the industry's fundamental test will be whether businesses continue paying for AI applications because they generate measurable economic benefits. 

Enterprise customers renewing AI software subscriptions from their operating budgets provide stronger evidence of sustainable demand than infrastructure purchases financed by suppliers. 

IDC's research also identified a significant gap between individual productivity improvements and measurable organization-wide financial returns from AI adoption.  

That gap raises an uncomfortable question: Are AI companies building infrastructure for a market that already exists, or for a market they hope will eventually materialize? 

The financial risks extend beyond equity valuations. 

AI infrastructure increasingly depends on debt, leasing arrangements, special-purpose vehicles and private credit. These structures can transfer financial exposure from technology companies to banks, insurers, pension funds, investment managers and private lenders. 

On October 19, Reuters reported that S&P Global warned major hyperscalers may have more AI-related financial liabilities than current disclosures reveal. The rating agency estimated that six major U.S. hyperscalers could spend more than $7 trillion on AI-related capital expenditures between 2025 and 2030. 

The concern is not simply how much money these companies are spending. It is whether investors can identify who ultimately bears the losses if projected revenues fail to materialize. 

An Economic Threat? 

The broader economic risk comes from the possibility that a financing structure supporting rapid expansion could amplify a downturn. 

Consider a hypothetical scenario in which enterprise AI adoption grows more slowly than expected. 

AI developers might reduce spending on computing capacity. Cloud providers could subsequently cancel equipment orders or struggle to meet lease obligations. Data center operators might experience declining utilization and falling rental rates. 

Chipmakers could suffer reduced revenue, while lenders holding debt backed by computing infrastructure might face deteriorating collateral values. 

If these developments occurred simultaneously, losses could spread across several interconnected industries. 

Rapid technological obsolescence compounds the problem. A data center's physical infrastructure may remain useful for decades, but its specialized computing equipment can depreciate rapidly as newer processors become more efficient. 

Unlike conventional real estate, AI hardware may have uncertain resale values and relatively short economically useful lives. 

In his September research published through Brookings, Columbia Business School economist Stijn Van Nieuwerburgh warned that off-balance-sheet financing could obscure correlated exposures. 

Nevertheless, he cautioned against prematurely concluding that AI infrastructure poses systemic risks comparable to previous credit crises.  

That distinction is important. 

The late-1990s telecommunications boom offers a potentially closer historical comparison than the 2008 financial crisis. Telecommunications equipment manufacturers helped finance customers that purchased their equipment, contributing to an infrastructure expansion that ultimately exceeded near-term demand. 

The resulting collapse damaged equipment suppliers, network operators, lenders and shareholders. Yet the underlying telecommunications infrastructure subsequently supported enormous economic activity. 

AI could follow a similar trajectory: The technology might become indispensable even while investors in particular companies or infrastructure projects experience substantial losses. 

The Financial Stability Board has also emphasized the need for vigilance regarding AI-related financial risks. In an October 12 letter, FSB Chair Andrew Bailey highlighted vulnerabilities involving AI and private credit. Separately, the FSB published 12 sound practices for responsible AI adoption on October 15, emphasizing governance, risk management and third-party dependencies. Those practices concern financial institutions' AI use rather than constituting specific rules for circular financing.  

Is Circular Finance All That Bad? 

Circular financing is not inherently harmful, nor is it unique to speculative technology markets. 

Automobile manufacturers have financed dealerships and customers for decades. Aircraft manufacturers, industrial equipment suppliers and telecommunications companies have employed similar arrangements. 

Such financing can make expensive products accessible to customers that otherwise could not afford them. 

In September, Ameriprise Financial Chief Market Strategist Anthony Saglimbene argued that AI circular financing should not automatically be classified as either beneficial or dangerous. 

He identified several advantages, including faster infrastructure construction, improved access to computing capacity, opportunities for smaller providers to expand and closer alignment between suppliers and customers. 

However, Saglimbene also noted that AI financing lacks the maturity, transparency and established collateral markets associated with conventional automobile financing.  

The benefits are substantial. 

Without strategic investments from major technology companies, smaller AI developers might struggle to compete against established cloud providers. 

Vendor financing can also accelerate construction of infrastructure that eventually benefits unrelated businesses and consumers. 

Furthermore, infrastructure investments may generate positive economic spillovers, including improvements to electrical grids, fiber-optic networks and computing capabilities. 

IDC offered a more optimistic interpretation in an October 22 analysis, AI Is Short of Capacity, Not Demand. The research firm argued that insufficient computing capacity remains a major constraint on the AI industry and that additional infrastructure could support future demand. Nevertheless, IDC identified a withdrawal of circular financing as a potential warning sign for the broader market.  

The central issue, therefore, is not whether circular financing exists. It is whether financing arrangements support productive investment at sustainable prices, with risks appropriately disclosed and allocated. 

What They Need to Know 

For financial advisors, investment managers and institutional investors, circular financing introduces an additional layer of complexity into evaluating AI-related investments. 

An investor purchasing shares in a major technology company may already have exposure to the same underlying AI ecosystem through semiconductor manufacturers, cloud providers, utilities, data center operators and private credit funds. 

Diversification across different companies does not necessarily provide diversification across economic risks. 

Financial professionals should therefore examine how much of a client's portfolio depends on continued AI infrastructure spending, rather than relying exclusively on conventional sector classifications. 

They should also distinguish between the technology's commercial prospects and the valuations attached to individual investments. 

AI can generate substantial productivity improvements without necessarily delivering attractive returns to every company financing its development. 

The most useful due diligence questions concern the origins of revenue, the durability of customer demand, the structure of financial obligations and the economic value of the underlying assets. 

Advisors evaluating AI-related securities should consider whether customers can continue purchasing products without additional financing, whether infrastructure utilization supports projected earnings, and whether companies have disclosed meaningful lease obligations, guarantees and contingent liabilities. 

Private credit and infrastructure investments warrant particular attention because investors may be exposed to AI-related risks through less obvious channels. 

Stress testing portfolios against declining AI capital expenditures, falling semiconductor prices, weaker cloud demand and tighter credit conditions could help reveal hidden concentrations. 

Advisors should also communicate that the presence of circular financing does not justify automatically abandoning AI investments. 

An indiscriminate retreat could sacrifice exposure to businesses generating genuine revenues, durable cash flows and valuable technological innovations. 

Instead, clients should understand that an industry's long-term success does not guarantee favorable investment returns at every valuation. 

Follow the money, then follow the customer. 

Circular financing has become a defining feature of the AI investment boom because the technology's development requires extraordinary amounts of capital before its ultimate commercial potential can be fully demonstrated. 

The arrangements linking Microsoft, Nvidia, Google, Amazon, OpenAI, Anthropic and numerous infrastructure providers are neither inherently fraudulent nor inherently sustainable. 

They represent a powerful financing mechanism whose benefits depend on whether the underlying economic activity eventually justifies the investment. 

For investors, the greatest danger may not be circular financing itself, but the possibility of mistaking supplier-financed growth for independent, enduring demand. 

The AI industry could ultimately deliver transformative economic benefits while simultaneously producing substantial investment losses, overcapacity and financial failures. 

Financial professionals should therefore focus less on whether circular financing proves that AI is a bubble and more on whether the companies and assets in their clients' portfolios can survive if the financing cycle slows. 

Ultimately, the industry's most important financial indicator will not be how much money AI companies invest in one another. It will be how much customers outside that interconnected financing network are willing to pay for artificial intelligence—and whether those payments can support the infrastructure already being built. 

That is where the money-go-round must eventually meet the real economy.  

Researched by DWN Staff