The Role of Circular Finance in Funding AI Investment
November 16, 2026 (IF13349)

Over the past five years or so, rapid artificial intelligence (AI) advancements have triggered massive capital demands. AI technologies comprise various systems centered around data centers that require specialized computer chips, networking and cloud infrastructure, and electric and cooling systems. Analysts estimate that investments in AI-related infrastructure in the United States will be around $500 billion in 2026 and around $10.3 trillion for the period from 2025 to 2032. AI companies view this funding as critical to refine their models and secure market share. Assessing the financing need, one AI lab executive implied that innovations to financing models to support technological development will be as important as the technology itself.

AI firms, infrastructure providers, and financial firms are using various methods to fund the ongoing build-out. This In Focus details one specific approach: circular financing, where AI industry players establish reciprocal financial and commercial relationships to drive growth and achieve strategic objectives.

Circular Financing Arrangements

In circular financing arrangements, one company provides financial support to another based on a commitment or an expectation that the beneficiary will buy goods and/or services from the other. Various media reports have documented such arrangements among companies in the AI industry. The Bank for International Settlements (BIS) broadened this concept to circular "relationships," defining them as arrangements where an investor and a target maintain a mutual commercial relationship alongside their financial one. The BIS found that between 2021 and 2025, 28.7% of AI firms' investments (by deal value) were in another AI company. Additionally, 55.2% of investments into AI companies came from another AI company.

The BIS identifies three primary market relationships, detailed below with examples.

  • 1. Investor-supplier relationship: A company both invests financial resources in and sells materials and services to the target company. For example, the computer chip developer Nvidia reportedly pledged to invest $100 billion in OpenAI in October 2025. OpenAI stated that it will use the funds to "deploy" 10 gigawatts of Nvidia graphics processing units (GPUs) at a data center and described its relationship with Nvidia as a "strategic partnership." In another example, Amazon invested $5 billion (with the potential to invest $20 billion more) in April 2026 in Anthropic. Anthropic committed to spending $100 billion on Amazon chips and cloud services over 10 years. Increasingly, suppliers are guaranteeing purchases, promising to lease back surplus equipment if needed.
  • 2. Investor-purchaser relationship: A company invests in its supplier. In one example, OpenAI agreed to use 6 gigawatts of AMD (another chip developer) GPUs. AMD offered OpenAI the ability to buy AMD's stock as part of the agreement, as discussed further in the following section. OpenAI also reportedly agreed to buy computing power from Cerebras, another chip developer, in a similar way.
  • 3. Reciprocal arrangements: Goods and services flow in both directions. For example, OpenAI uses chips produced by Amazon, while Amazon and OpenAI create customized models for use with Amazon's consumer-facing applications. (Amazon invested $50 billion in OpenAI.)

The industry's circular relationships may also include three or more party arrangements, including financial institutions such as private equity firms, only some of which may be explicit. In other words, because of the infrastructure and financing needs and the relatively limited number of participants, agreements that are technically between two parties may also establish implicit relationships between companies.

On an aggregate scale, some companies may be connected through multiple vectors, while some are keystones to many agreements. For example, OpenAI committed $317 billion to Oracle for computing power. However, Oracle does not manufacture chips and will reportedly buy them from Nvidia and AMD. Thus, OpenAI is directly and indirectly linked to Nvidia for infrastructure and financing.

Figure 1. Multiparty Circular Finance

Source: CRS analysis based on media reports and corporate press releases.

Note: For illustrative purposes; not all details are represented.

Benefits and Risk of Circular Financing

These circular arrangements are essentially amplified versions of vendor financing arrangements, where a vendor provides funds that a corporate customer uses to buy the vendor's goods or services, which are subsequently used to produce the customer's product. Suppliers, such as chip companies, provide capital to a customer, such as an AI lab requiring GPUs, that the AI lab otherwise may not have been able to purchase. The investment allows the lab to expand its business, finance research and development (e.g., testing), and create new products (models). Investors stimulate demand for their chips, increasing sales to (and generating revenue from) the AI labs.

There are various types of financing AI labs may rely on, such as investments of capital for equity (or an ownership stake) in firms, which may include warrants or the option to take an equity stake, and venture debt or loans that must be repaid. In various recent cases, investors have taken ownership stakes in AI labs. The leading AI labs are start-ups that spend more than they earn and rely on external funding for large investments. Investors are motivated by the belief that financing will support both the company's ability to operate independently eventually and increase the value of the investor's stake as compensation.

AI firms may face other nonfinancial constraints. The current phase of the AI development and investment cycle is characterized by significant demand for various inputs, the delay of any of which could impede future model development and broader adoption of a firm's technology. Thus, the need for specialized material or inputs may, according to the BIS, drive customer-financed relationships. Agreements between OpenAI and Oracle, and between OpenAI and Amazon, were seen to stem from concern that a prior exclusive agreement with Microsoft would not provide sufficient computing resources to achieve OpenAI's goals.

AI vendors may be more knowledgeable about the industry and lending risk and therefore better positioned to provide financing than banks or other traditional lenders in an industry that exhibits significant overlap across vendor and customer expertise. A vendor may also be able to obtain other relevant market information, such as how fast a customer or an industry is scaling, the rate of technological adoption, or both, which may be useful for its own business model.

Benefits may also arise from traits that are idiosyncratic to the AI industry. In two of the examples above, OpenAI effectively "subsidized" its purchase of AMD and Cerebras chips. Furthermore, its option to invest in the companies is based on the expected increase of the investees' market valuation generated by the affiliation with OpenAI. The relationship also diversified OpenAI's chip sourcing. OpenAI's option to take a stake in the chip companies may allow it to direct the supplier's future operations, including in relation to its own needs. In addition to the valuation, AMD reportedly benefited from securing ongoing business from one of the most popular AI labs.

Opacity may obscure market dynamics. The self-sustaining nature of the relationships—chip demand is fueled by chip company financial support—may make it hard for investors to determine what portion of the demand is "organic," driven by actual end users of the technology. Circular financing may also extend the amount of time before an AI company must achieve self-sufficiency.

Circular financing raises corporate governance concerns. While some of the companies involved are not required to disclose information publicly, even when agreement details are provided, agreements may be so novel or complex as to obscure or understate the true magnitude of interconnectedness or misrepresent performance and financial exposures. This can limit the public's and policymakers' ability to accurately identify and assess risks. For example, months after the $100 billion OpenAI-Nvidia agreement was announced, media reported that it was being significantly downsized. Circular finance may also motivate the parties to disregard the nominal requirement that the recipient achieve certain milestones prior to additional investment disbursements to avoid introducing weakness to either or both parties and breaking the cycle.

Studies and commentators have pointed to the circular relationships prevalent among telecom suppliers and dot-com merchant customers during previous boom-bust cycles that coincided with the bubble bursting as a cautionary tale of how the current scenario can unfold. When funding from industry firms disappeared, newer firms could not find funding elsewhere and shrank or failed.

Whether the AI industry can avoid a similar outcome is unclear and may hinge on broader market assumptions about the viability of the AI. For example, there is a question of whether speculation about the promoted vision of AI's utility will materialize and be monetizable. While few doubt AI's potential, developers likely must create practical uses for enterprise and retail customers (organic demand) to generate enough revenue to wean the labs from supplier finance. Whether there is a bubble and, if so, whether circular finance is a symptom of a bubble and/or a factor that could amplify its consequences should it burst are open questions.

Policy Issues

Circular financing in the AI sector raises different policy issues for Congress, depending on (1) whether the industry introduces economic and/or financial shocks and (2) whether such shocks can be attributed specifically to circular financing.

Interconnectedness from circular financing may create the channel through which a potential shock at one firm could extend to a counterparty or other firms, creating systemic vulnerabilities across the industry. Moreover, because many AI and chip firms are engaged in other forms of financing and market activity involving major financial institutions—such as issuing debt and creating joint special purpose vehicles—in part to finance the commitments from the circular relationships, challenges in the AI industry may spill over to the financial sector. Alternatively, to the extent that these practices are financed from investor companies' own operations, they may keep risks out of financial institutions. In its Global Financial Stability Report, however, the International Monetary Fund noted that market participants consider a correction in AI investments to be a major risk to financial markets. And, by some accounts, much of the financial industry has exposure to AI investments. This suggests that circular financing may create the conditions for financial systemic risk, meaning market risk that poses a risk to the financial system.