Circular Deals: How AI Companies Are Funding Their Own Demand

Series 2: Circular Capital — The AI Industry’s Breton Pulley Problem

1 The New Loop of AI Capital

In traditional finance, a company builds products, sells them to customers, and records revenue. In 2025’s AI economy, many of the biggest players are selling to themselves — or to their own investees.

Nvidia invests in CoreWeave, a cloud startup that buys Nvidia’s GPUs in bulk.Microsoft invests in OpenAI, which spends billions on Microsoft Azure compute.Amazon and Anthropic mirror the same pattern through AWS and Bedrock.

On paper, revenues soar and valuations expand. But much of the money is circulating inside a loop of interdependent contracts and cross-shareholdings — a modern-day “Breton Pulley,” where control and capital run in circles.

“In the AI economy, the rope isn’t made of shares — it’s made of contracts.”

2 From Ownership Loops to Contract Loops

In Part 1 of Bridge Connect’s earlier Breton Pulley series, we showed how industrial groups like Bolloré created ownership circles to magnify control.The AI sector has invented its own version — the contract pulley — a mechanism that recirculates cash and demand through mutual investment and procurement.

[Nvidia] ─ invests in ─▶ [CoreWeave]
[CoreWeave] ─ buys GPUs from ─▶ [Nvidia]
[Microsoft] ─ invests in ─▶ [OpenAI]
[OpenAI] ─ runs on ─▶ [Azure]
[Anthropic] ─ backed by ─▶ [Amazon & Google Cloud]
[Cloud Providers] ─ resell compute from ─▶ [Nvidia Hardware]

Each arrow generates headline revenue for one party and paper gains for another. Viewed together, the same dollar passes through multiple balance sheets, creating the illusion of explosive external growth.

3 Case Study: Nvidia ↔ CoreWeave ↔ OpenAI ↔ Microsoft

Nvidia supplies GPUs to CoreWeave, a cloud startup specialising in AI compute. Nvidia also holds an equity stake in CoreWeave and committed billions in financing to expand its data-centre capacity.CoreWeave leases this capacity to clients including OpenAI, which builds models for Microsoft, hosted on Microsoft Azure — which runs on Nvidia GPUs.

Capital, hardware and contracts form a circle:

Nvidia → CoreWeave → OpenAI → Microsoft → Azure → Nvidia

To investors, each node reports revenue growth. To an auditor, the aggregate system resembles a single machine counting the same transaction several times.

4 Why the Loop Exists

1. Securing Supply: AI chips and compute capacity are scarce; vertical loops guarantee access.

2. De-risking Scale: Investing in customers locks in demand for expensive infrastructure.

3. Boosting Valuation: Reinvested capital keeps the ecosystem’s headline growth metrics rising.

4. Mutual Leverage: Each partner’s valuation supports the others’ balance sheet.

It is the same financial alchemy as a Breton Pulley — leverage without debt, control without clear ownership.

5 Synthetic Demand and Mirror Revenue

In telecom history, vendor financing once fueled the 3G bubble: equipment suppliers lent to operators who bought their equipment. The AI sector’s contract loops are similar.

When Nvidia or Microsoft invest in their customers, a portion of recorded sales comes from capital they supplied themselves. Revenue becomes self-referential — mirror revenue.

This distorts metrics such as:

• Free cash flow: inflated by vendor-financed sales.

• Return on capital: boosted by circular gains on investments.

• Market share: amplified by closed-loop contracts.

6 Regulatory and Governance Concerns

• Disclosure: Most filings note the investments but not the contractual interdependencies.

• Competition Law: When supply and demand merge, pricing transparency erodes.

• Systemic Risk: A single failure (e.g., if OpenAI cannot pay Azure fees) propagates through every investor-supplier link.

• Audit Challenge: GAAP/IFRS consolidation does not capture inter-company contract loops outside ownership control.

“Circular capital creates the illusion of diversity — until one node fails and the whole system repeats the same mistake.”

7 Parallels to Breton Pulleys

	Breton Pulley (Industrial Era)	AI Contract Pulley (Digital Era)
Mechanism	Cross-shareholdings	Cross-contracts + cross-investments
Control driver	Equity loops	Dependency loops
Leverage type	Synthetic equity	Synthetic demand
Hidden risk	Ownership opacity	Revenue opacity
Governance gap	Voting rights vs capital	Contract rights vs economic exposure

The physics is identical: the same capital does multiple laps around the wheel before meeting reality.

8 Investor Blind Spots

Boards and investors often celebrate AI growth without asking where the demand originates.Questions to ask:

1. What percentage of sales comes from investee customers?

2. How many revenue streams depend on partner capacity commitments?

3. Is there a clear exit path if the loop needs to unwind?

9 Bridge Connect Perspective

Bridge Connect advises boards in telecoms and infrastructure that the same structural risks apply to AI and data-centre investments:

• If a vendor owns a piece of its customer, governance must treat that relationship as related-party for all decision approvals.

• Cash-flow transparency must be maintained even when funding and revenue flow through different entities.

• Capital allocation should be stress-tested for loop collapse — what happens if one partner defaults or regulators intervene?

Bridge Connect calls this approach the “Circular Capital Audit.”

10 The Bigger Picture

Circular AI deals are not fraudulent — many are strategically rational. But when entire industries become self-referential, they risk mistaking internal momentum for market traction.The Breton Pulley analogy reminds us that loops magnify forces in both directions. When growth stalls, the same feedback mechanisms amplify the fall.

11 Conclusion — The Rope Tightens

AI’s contract loops reflect a classic human pattern: engineer certainty through mutual dependence. But true resilience requires independent demand signals, not circular ones.

“When every player is both supplier and customer, nobody knows where the market ends and the mirror begins.”