Morgan Stanley Flags Circular AI Financing As Risks Build Behind Surging Infra Stocks
Morgan Stanley is warning that AI's hyperscale buildout is being propped up by circular financing - suppliers funding customers, revenue sharing across the stack, and cross-ownership. Analyst Todd Castagno says the market needs clearer disclosure to map who is financing whom, and on what terms.
The concern lands as infrastructure names surge. Nvidia and AMD set new highs, while Dell Technologies and Super Micro Computer ride the server boom. The question: will current cash commitments be supported by durable AI cash flows?
What's changing under the hood
Nvidia has invested in xAI and will supply it with processors. It also has supplier-investor ties with CoreWeave and OpenAI. AMD disclosed a supplier-investment arrangement with OpenAI as well.
Microsoft and Oracle are expanding AI data center partnerships, but disclosures remain thin. When non-AI businesses are massive, related AI deals often miss "materiality" thresholds - yet AI expectations dominate valuations.
Why finance teams should care
- Demand signals can be distorted when suppliers fund customers, share revenue, or hold equity stakes in each other.
- Accounting and cash flow quality can be obscured by vendor financing, prepayments, and minimum-commitment contracts.
- Valuation risk rises if AI workloads don't scale fast enough to service the capital being deployed.
Red flags to test in filings and calls
- Customer and supplier concentration: Top customers by revenue and receivables; top vendors by payables and prepayments.
- Vendor financing terms: Interest rate, tenor, collateral, recourse vs. non-recourse, convertibility, revenue kickers, warrants.
- Related-party exposure: Identify cross-ownership, board overlaps, and any transactions under related-party definitions.
- Revenue economics: Gross vs. net presentation, rebates, revenue-sharing, and take-or-pay or minimum commitments.
- Capex and asset life: Useful life and depreciation of GPUs and racks, residual value assumptions, impairment testing cadence.
- Cash flow quality: Adjust FCF for financing extended to customers; track contract assets, prepayments, and working capital swings.
- Off-balance-sheet items: Guarantees, supplier commitments, and capacity reservations that don't show up in leverage ratios.
- Utilization and pricing: GPU hours, data center utilization, unit pricing trends, backlog, and cancellation terms.
Questions for management
- What portion of current AI revenue is tied to customers that receive financing or equity from the company?
- How are revenue-sharing agreements structured, and what triggers changes in economics?
- What are the top three contractual risks in current AI data center commitments?
- What percentage of capex is backed by minimum-use contracts vs. speculative capacity?
- How sensitive is guidance to GPU pricing, utilization, and refresh cycles over the next 12-24 months?
Valuation and risk framing
- Scenario analysis: Stress test AI workload growth, pricing, and utilization; haircut revenue tied to financed customers.
- Capital intensity: Model replacement cycles for accelerators and networking; pressure-test depreciation vs. economic life.
- Disclosure gap: Where AI is driving the multiple, request segment-level detail even if transactions are below quantitative thresholds.
- Credit view: Map covenants to AI-related commitments; watch for recourse financing that can amplify downside.
Where to look for clarity
- 10-Q/10-K: MD&A, liquidity, capital resources, and commitments footnotes.
- Related parties and concentration disclosures under Reg S-K Item 404 and, where applicable, IAS 24.
- Revenue recognition and principal/agent assessments; customer financing and contract asset footnotes.
Stocks in focus
Infrastructure leaders Nvidia and AMD are hitting records, while Dell Technologies and Super Micro Computer extend gains on server demand. The bull case rests on workloads and pricing staying strong through the upgrade cycles.
The bear case centers on financing-fueled demand, thin disclosures, and the risk that expected cash flows don't materialize fast enough to cover capital outlays.
Bottom line
Cross-ownership, vendor financing, and revenue-sharing are tightening the AI loop. That can accelerate buildouts - and blur true demand.
Push for granular disclosure, haircut financed revenue in your models, and anchor valuation on cash flows that don't depend on supplier money to survive the next cycle.
Helpful resource: If you're budgeting AI inside a finance org, here's a curated list of practical tools to evaluate and cost out: AI tools for finance.
