Wall Street feels danger as AI firms fund their own demand engines
The AI supply chain is starting to finance its own revenue. Chip and infrastructure providers are investing in their biggest customers, who then spend more on the same suppliers. It's a clean loop on paper-and a messy one for price discovery.
Recent moves highlight the pattern: Nvidia's planned multibillion investment in OpenAI, billions more into CoreWeave and xAI, while OpenAI strikes capacity deals with Oracle, CoreWeave, AMD, and taps Broadcom for in-house chips. Analysts warn this is transforming buyers into portfolio companies, blurring incentives and inflating sector valuations.
Blackstone's Jonathan Gray called the circularity a top risk. His message to deal teams: treat AI exposure as a first-page issue. The takeaway for finance pros is simple-assess cash realism, not headline growth.
How the loop works
- Supplier → Customer: Hardware vendors invest equity or extend favorable financing to compute buyers.
- Customer → Spend: Those buyers lock in capacity, often via prepayments and long-term commitments.
- Supplier → Revenue: Suppliers book sales to companies they fund, supported by capital rather than end-user demand.
- Feedback: Higher valuations enable more fundraising, which enables more capacity purchases, which props up supplier growth.
Why this matters for your P&L
- Valuation drift: Multiples expand on financed demand, not proven unit economics.
- Quality of revenue: Related-party exposure and prepay-driven sales can mask weak pull-through.
- Credit risk: If funding tightens, customers cut capacity first; receivables and utilization take the hit.
- Concentration: A few hyperscalers and model labs dominate orders; single-name shocks can cascade.
- Regulatory overhang: Expect scrutiny of related-party disclosures and competition issues.
What to scrutinize in filings and data rooms
- Related-party transactions: Equity stakes, board seats, and any supplier financing of customers.
- Customer and supplier concentration: Top-5 revenue share; non-cancelable purchase obligations; take-or-pay terms.
- Capacity commitments: Long-dated compute agreements, prepayments, and minimums versus realized utilization.
- Revenue recognition: Timing of revenue tied to financed deals, side letters, and price protection.
- Cash conversion: Delta between reported growth and operating cash flow; inventory days and WIP swell.
- Unit economics: Cost per token/compute hour trends, power pricing assumptions, and gross margin durability.
Bubble signals vs. durable demand
- Signals: Loss-making buyers scaling capacity faster than user adoption; rising DSOs; repeat equity bridges to fund opex.
- Durable: Sticky workloads with clear ROI, diversified customers paying with cash from operations, and improving utilization without subsidies.
Portfolio moves to consider
- Stress test: Cut funded demand by 30-50% and raise power costs; re-rate supplier margins and working capital needs.
- Hedges: Use put spreads on crowded AI indices; collars into events; skew tends to cheapen after relief rallies.
- Pairs: Long cash-generative enablers with power or networking bottleneck leverage; short capacity takers dependent on vendor financing.
- Credit: Watch HY issuance to AI infra buyers; use CDX HY or single-name CDS as funding-tightening hedges.
- Position sizing: Cap single-theme exposure; assume higher gap risk around regulatory headlines.
Market posture right now
Equities bounced last week but remain range-bound as positioning resets. Investors still price multiple Fed cuts and resilient growth, which keeps AI-linked leadership in play but fragile.
Bank of America reports high-net-worth equity allocations near cycle highs, limiting incremental demand. Hedge funds re-levered into strength, then registered the largest global de-risking since April, with a notable rise in macro shorts, per Goldman Sachs. The tape can grind up on lighter positioning-but liquidity is thin and factor reversals bite.
For policy odds, track the CME FedWatch tool for implied path shifts: CME FedWatch.
3-6 month scenarios to model
- Managed glide: Capex growth slows, utilization improves, multiples cool without a break.
- Air pocket: Funding window narrows; financed buyers pause; suppliers guide down; spillover to semis and infra credit.
- Regulatory jolt: Tighter related-party or competition oversight challenges co-investment structures.
- Supply constraint: Power and networking bottlenecks cap shipments; pricing holds; mix shifts to higher-margin services.
Due-diligence checklist for AI-exposed deals
- How much revenue depends on customers that the company also funds or guarantees?
- What percent of backlog is prepaid, and at what utilization threshold do penalties trigger?
- If compute prices fall 20% or power rises 30%, what happens to gross margin and payback periods?
- Does the buyer base generate cash from operations, or rely on constant external capital?
- Are there change-of-control or exclusivity terms that trap capacity or limit optionality?
Skills and tools for finance teams
If you price AI exposure daily, level up your toolkit. Curated options for finance professionals: AI tools for finance.
Bottom line: Treat financed demand as provisional. Separate cash-backed adoption from loop-driven growth, and price the spread.
Your membership also unlocks:
