The Bank for International Settlements has found that nearly half of the disclosed value of investment deals between artificial intelligence companies overlaps with a supplier-customer relationship, exposing how funding, chip access, cloud contracts and future computing capacity are becoming tightly connected across the sector.
In an eight-page brief published on Oct. 1, the BIS said 46.4% of the disclosed value of AI-to-AI investments between 2021 and 2025 involved companies that were also commercially linked as suppliers and customers. The pattern was far more pronounced in the largest deals than in the overall transaction count: 16.1% of investment transactions overlapped with a supply relationship.
The findings do not show that investment proceeds were directly used to purchase products from the backer. The BIS instead identifies a growing structure in which the same companies can simultaneously finance each other, supply critical technology and commit to long-term purchases of computing capacity. That arrangement gives AI developers access to scarce infrastructure while tying major hardware and cloud providers more closely to the financial health of their customers.
Capital and computing capacity are increasingly linked
The BIS examined 1,246 AI companies spanning five layers of the industry: compute, infrastructure, data tools, models and applications. It combined PitchBook financing data with FactSet supply-chain records, then manually reviewed relationships to identify 972 internal AI investment links between 2021 and 2025.
AI companies are increasingly important sources of financing for their peers. According to the BIS, 28.7% of the investment value deployed by AI firms went to other AI companies over the period. From the recipient side, 55.2% of funding raised by AI firms came from other companies in the same sector.
That funding concentration reflects the structure of the AI buildout. Model developers require enormous volumes of advanced chips, data-center capacity and cloud services, while the suppliers of those resources benefit from securing committed customers for years rather than months. Equity investments can help customers finance their expansion, and long-term commercial agreements can make that expansion more predictable for suppliers.
The BIS defines a “circular relationship” as one in which two AI companies have both an investment connection and a supplier-customer relationship during the same period. The definition does not require a specific financing round to be legally connected to a particular procurement agreement. It covers relationships that exist within a five-year window.
Among the circular relationships identified by the BIS, 64% involved the investing company also supplying products or services to the funding recipient. Compute and infrastructure firms accounted for 73% of circular investment relationships, placing chipmakers, cloud platforms and data-center providers at the center of the network.
Microsoft, Amazon and Nvidia illustrate the structures
The BIS outlined three recurring forms of circular ties. A supplier may invest in a customer that needs capital to buy infrastructure. A customer may invest in a supplier to secure access to a constrained technology or service. Two companies may also buy from each other while holding equity stakes or participating in each other’s financing.
Microsoft’s relationship with OpenAI is among the examples cited. Microsoft has provided financing to OpenAI, while Azure has served as OpenAI’s exclusive cloud provider. The commercial arrangement gives OpenAI access to computing infrastructure needed to train and deploy models, while Microsoft gains a major cloud customer and close access to OpenAI’s technology.
Amazon’s arrangement with Anthropic follows a comparable model. Amazon committed up to $4 billion to Anthropic, which uses Amazon Web Services as its primary cloud provider and deploys its models on AWS Trainium and Inferentia chips. The relationship connects Amazon’s capital commitment to demand for its cloud and custom silicon products.
The BIS also pointed to Nvidia and CoreWeave as an example of how chip supply, financial participation and data-center expansion can operate in parallel. Such structures can support rapid construction of computing capacity, particularly where developers and cloud providers face limited availability of high-performance hardware.
The report’s comparison with the wider U.S. corporate market suggests these arrangements are unusually common in AI infrastructure. Across more than 10,000 U.S. customer-supplier relationships reviewed by the BIS, around 3.3% involved one trading partner holding an equity stake in the other. In AI compute, cloud and related infrastructure markets, the equivalent figure was 15.2%.
Larger deals create more concentrated exposure
The concentration of circular ties in high-value transactions creates a potential fault line if demand for AI services fails to match the capacity now being built. Suppliers could face weaker orders from customers in which they also hold equity stakes, have extended financing or have signed contracts containing commitments related to future capacity.
The BIS drew a parallel with the late-1990s telecommunications equipment cycle. Equipment makers including Lucent and Nortel provided financing to network operators that purchased their products, allowing suppliers to report rising equipment sales while accumulating credit exposure to the same companies. When telecom demand weakened, vendors faced pressure from both reduced sales and losses tied to their financing arrangements.
AI contracts can take several forms beyond straightforward equity purchases. The BIS said supervisory reviews may need to consider minimum purchase obligations, reserved computing capacity, exclusivity provisions, prepayments, discounts and residual-value guarantees. A company’s risk can therefore extend beyond its reported equity holding if it has committed to buy capacity, guaranteed the value of equipment or provided financing connected to a customer’s expansion plans.
The report also cautioned that its deal-value figures have limits. It uses the total disclosed amount of each investment round and cannot determine how much any individual participant contributed when several backers joined a financing. The 46.4% figure therefore measures the overlap between large funding rounds and commercial relationships, rather than the amount of investment capital that was specifically tied to procurement.
Disclosure becomes more difficult across contract chains
The emerging AI financing model places equity, credit and supply-chain risk in overlapping parts of the market. A chip supplier may have exposure to a customer through direct shareholdings, future product orders, financing arrangements and commitments from cloud or data-center partners. Each agreement may be understandable on its own, but their combined effect can be harder to assess from conventional financial disclosures.
The BIS said information sharing among competition, securities and banking supervisors may become more relevant where commercial procurement, equity stakes and credit arrangements involve the same counterparties. Competition authorities may examine whether access to scarce computing infrastructure is being locked up through exclusivity or capacity reservations, while financial regulators may focus on credit and concentration risks that sit alongside those contracts.
For cryptocurrency markets, the report offers a more useful warning about correlations than a direct prediction for token prices. Digital assets linked to AI computing, decentralized infrastructure or technology-sector risk appetite can react sharply when expectations for hardware sales, cloud spending or corporate capital expenditure change. The BIS findings show that headline demand figures in AI should be read alongside the financing and contractual ties supporting them.
The immediate issue is not whether circular AI relationships are inherently improper. They can help developers secure the computing resources required to build and operate models, while giving suppliers more certainty before committing billions of dollars to chips and data centers. The financial strain would emerge if the same contracted capacity, equity stakes and financing commitments remain in place after revenue growth or outside customer demand slows.
Explore how AI reshapes finance and trading in our guide on Web3, AI, and crypto and what’s coming next.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
