Big Tech's Debt Binge: What It Means for Risk, Valuation, and Your Models
AI capex isn't slowing, but the funding mix is changing. The largest platforms are leaning harder on debt to build data centers, secure chips, and lock in supply - a clear shift from the cash-funded build that defined the last cycle.
Equity investors are starting to price in the added risk. The use of leverage, vendor financing, and circular revenue relationships is introducing new points of failure. Volatility is picking up as the market weighs how long the payoff will take against the size of the checks being written.
The funding shift, by the numbers
- The five major AI spenders - Amazon, Alphabet, Microsoft, Meta, and Oracle - have raised a record $108 billion in debt in 2025, more than 3x the prior nine-year average.
- Oracle sold $18 billion in US investment-grade bonds and saw banks launch a $38 billion data center financing deal tied to its buildout.
- Additional issuance: Meta $30 billion, Alphabet $38 billion, Amazon $15 billion.
- Valuation context: the forward 12-month P/E of the Bloomberg Magnificent 7 Index has slipped to its lowest in 2+ months, now roughly in line with its five-year average.
Why this raises the risk profile
"I view this as the AI story maturing and entering a new phase, one that is likely to be marked by more volatility and additional risk," said Lisa Shalett, CIO of Morgan Stanley's wealth management unit.
The concern isn't capex itself - it's the mechanics. More balance sheets are getting involved (including players like Oracle and CoreWeave), leverage is climbing, and revenue links between vendors, hyperscalers, and financiers are getting tighter. That interconnectivity can transmit stress.
Oracle: the flash point
- Capex guide: $35 billion this fiscal year, largely for cloud infrastructure.
- Free cash flow: expected at negative $9.7 billion this year, with deficits projected through fiscal 2028 (down to negative $24.3 billion).
- Credit signal: five-year CDS widened to a three-year high; S&P Global Ratings shifted its outlook to negative amid anticipated capex and debt issuance.
"These companies are investing massive amounts and committing to massive amounts of capex, some of which will be financed with debt," said Arnim Holzer, global macro strategist at Easterly EAB. "This doesn't mean Oracle's stock is trash, but it should be more volatile."
For a quick refresher on how rating outlooks differ from rating levels, see S&P Global Ratings: What a credit rating means.
Market signals to track
- CDS spreads: widening often leads equity drawdowns for overextended issuers.
- Rate windows vs. issuance: heavy supply into thin liquidity can pressure spreads and equity multiples.
- Index-level tells: the Nasdaq-100's swings around AI headlines and export-policy chatter (e.g., potential approvals for Nvidia's H200 chips to China) are amplifying factor volatility. Reference: Nasdaq-100 (NDX).
- Valuation mean-reversion: with the Magnificent 7 back near five-year average multiples, the market is demanding cleaner ROI visibility.
We're in the "show me the money" phase
"When companies that don't need to borrow are borrowing to make investments, that sets a bar for the returns on those investments," said Bob Savage, head of markets macro strategy at BNY. The hurdle rate just moved higher.
At the same time, investors still like the megacaps' earnings durability and moats. UBS estimates suggest 80-90% of planned capex is still covered by internal cash flows. Debt complicates the story; it doesn't end it.
Practical checklist for equity and credit analysts
- Capex coverage: map capex to operating cash flow and FCF after interest; flag years where FCF turns negative and the duration of deficits.
- Leverage glidepath: pro forma net leverage under base, slow-demand, and higher-rate scenarios; include off-balance sheet data center financing and vendor prepayments.
- Payback clarity: tie spend to unit economics - utilization assumptions, revenue per compute hour, margin impact, and depreciation schedules.
- Revenue circularity: identify interlocking deals (customer-supplier-financier loops) that could propagate stress.
- Covenant and liquidity: undrawn capacity, maturity walls, refi risk, CP reliance, and interest coverage under stressed spreads.
- Supply chain concentration: exposure to a single GPU vendor or foundry; sensitivity to export restrictions and allocation shifts.
- Valuation discipline: adjust multiples for balance-sheet risk and capex intensity; use sum-of-the-parts where AI infrastructure is separable.
- Cross-asset tells: monitor basis between CDS and equity; widening basis without fundamental deterioration can be a timing signal.
What could change the narrative
- Faster revenue ramp: clearer monetization of AI services that closes the cash burn gap earlier than modeled.
- Supply normalization: easing GPU constraints cuts prepay needs and reduces the impulse to lever up.
- Policy outcomes: export approvals or denials shifting deployment timelines and capex cadence.
Bottom line
AI capex was once a clean story funded by cash. Now it's a capital structure story. The winners will prove ROI on a tighter clock, manage leverage proactively, and keep optionality on supply and policy.
The market will give them time - but it won't give them a free pass.
Resource
- For a curated view of AI tools relevant to finance workflows, see AI tools for Finance.
Your membership also unlocks:
