The Pulse
BIS: AI-to-AI Deals Accounted for 55.2% of Disclosed Incoming Investment Value
A Bank for International Settlements bulletin says 55.2% of incoming investment in AI firms from 2021 through 2025 came from other AI firms. It examines how investment ties overlap with commercial relationships and the risks those connectio
AI.info Team ·
Deals involving AI investors represented 55.2% of the disclosed incoming investment value received by AI firms from 2021 through 2025, according to a Bank for International Settlements bulletin published October 1. The figure counts full disclosed transaction values, not the amounts contributed by individual investors. The authors also find that commercial ties often overlap with those investments, linking firms’ financing and sales relationships.
What the BIS counts as a circular tie
The bulletin examines investment among a universe of 1,246 AI firms classified across five supply-chain layers: compute, infrastructure, data tools, models and applications. Using investment information from PitchBook and commercial-relationship data from FactSet, the authors identify 972 investment relationships between AI firms from 2021 to 2025. They define a relationship as circular when the investor and target also share a supplier-customer relationship at some point in that period.
That definition is broader than one that counts only a specific investment deal tied to a purchase commitment, or only reciprocal financial investments. Among AI firms’ investment deals, 28.7% of disclosed value went to targets that were also AI firms. Of the AI-to-AI deals, 16.1% by count and 46.4% by disclosed value involved firms with a commercial supply-chain relationship.
“Circular investment relationships reflect key economic features such as the need to secure critical inputs and the presence of information asymmetries, yet they entail macroeconomic risks and increase opacity.”
Jon Frost, Rudraksh Kansal, Kumar Rishabh, Vatsala Shreeti and Leanne Si Ying Zhang, the bulletin’s authors, write.
Most circular relationships involved an AI firm that invested in a company it also supplied; the bulletin puts that share at 64%.
Why suppliers and customers finance each other
The paper describes several business reasons for the overlap. A supplier may know more about a customer’s operations than outside lenders do, and financing that customer can help sustain demand for the supplier’s own products. In AI, the authors point to model developers’ high capital needs and the compute services they buy from chip makers and cloud providers.
Funding can also flow from customer to supplier when a company wants to secure access to scarce inputs. The bulletin cites memory chips as a possible bottleneck and notes that financing may give a customer influence over a supplier’s strategic choices. Firms can also use investment to reduce the risk that a trading partner changes terms after one side has made specialised, hard-to-repurpose investments.
How financing can blur demand
Overlapping ties may solve problems for individual companies, but the authors warn they can make sector-wide demand harder to read. If a supplier finances a customer that then buys the supplier’s products, part of the supplier’s sales growth rests on its own financing decision rather than independent end-user demand. The bulletin compares that mechanism with the late-1990s telecommunications boom, when equipment vendors financed network operators that bought their equipment.
A downturn could hit both sides of a relationship: an investor-supplier might lose the value of its stake and future sales to the same customer. Among circular investment relationships, 73% came from firms supplying compute or infrastructure, the paper says. That concentration could transmit stress through both commercial and financial channels.
Disclosed deal values have limits
The figures are estimates, not a complete accounting of cash exchanged. The authors say PitchBook analysts have not reviewed their calculations and that the results may differ from PitchBook’s own methodology and reported figures. Deal values represent full disclosed transaction values, not the portion contributed by each investor, and disclosed figures may not match amounts ultimately disbursed.
The authors also point to limited disclosure by private firms and complex arrangements that may combine cash investments with purchase commitments or guarantees. Such terms can create obligations that surface only during a downturn, while deals spanning sectors and jurisdictions complicate oversight. The bulletin’s data establish the scale of overlapping investment and commercial relationships; they do not determine how much of the AI sector’s current demand depends on those arrangements.