The AI ‘credit card’ is getting costlier as CoreWeave bond yield tops 16%

Credit stress is rising around AI infrastructure: CoreWeave’s 2032 bond yield topped 16%, and its five-year CDS spread exceeded 1,289 bps before easing to about 1,155 bps Friday. CDS spreads also rose over the past month for Nvidia (76%), Broadcom (49%) and Amazon (44%), though their absolute spreads remain much lower than CoreWeave’s. Potential data-center moratoriums could affect 45% of planned U.S. capacity, while Goldman expects $440 billion of hyperscaler debt issuance next year and $300 billion of additional project-level issuance; delayed AI monetization and political constraints are key downside risks.

Analysis

The key transmission is not a uniform rise in AI default risk; it is a higher marginal cost of capital that makes long-dated, power-constrained projects easier to defer. That raises the hurdle rate for neocloud builds first: shorter customer commitments or slower prepayments can weaken project economics before any hyperscaler cancels headline capex. CRWV is therefore the clearest equity/credit vulnerability, while NVDA, AVGO and ORCL face second-order risk through slower orders or utilization—not equivalent balance-sheet stress. AMZN and other large platforms may gain relative advantage if customers and capital migrate toward operators able to fund infrastructure internally, though their own returns on incremental capex still matter.

Near term, political signals are a volatility catalyst, not proof of a national build freeze: federal support and state-level permitting/legislation pull in opposite directions. Over 1–3 months, watch state moratorium language, project permits, hyperscaler capex guidance, prepayment/commitment terms, and CRWV refinancing or liquidity disclosures. Over 6–18 months, the decisive test is monetization and free cash flow against committed investment. A slowdown would also hit power, cooling and equipment suppliers through delayed orders; conversely, constrained grid capacity could shift spending rather than eliminate it.

Contrarian point: CDS widening from low levels at investment-grade issuers is a warning about risk appetite and funding capacity, not evidence that their solvency is impaired. The more actionable signal is dispersion between project-financed builders and cash-generative platforms. Thesis weakens if state measures remain narrow, projects continue securing financing, and cloud monetization/cash generation keeps pace with commitments.

Key Decisions for Investors

• Prefer a relative-value short in CRWV versus long AMZN, sized modestly: it targets financing/refinancing asymmetry while retaining exposure to continued AI demand. Enter on rallies or after confirming weaker contract/prepayment terms rather than chasing a distressed move. Reassess if CRWV demonstrates durable contracted cash flows and financing access, or if AMZN materially cuts infrastructure plans.
• For accounts able to trade credit, consider CRWV protection only as a defined-risk, catalyst-driven hedge; the elevated spread makes carry and sharp spread tightening meaningful risks. Avoid extrapolating CRWV pricing to NVDA, AVGO or AMZN credit without issuer-specific deterioration.
• Keep NVDA/AVGO exposure tied to order and guidance evidence, not CDS percentage moves alone. Reduce or hedge if hyperscalers defer capex, commitments rise faster than monetization, or state moratoria materially delay projects; unwind the hedge if permits and financing proceed without customer or cash-flow deterioration.
• Alert: verify the scope and timing of proposed state restrictions, CRWV liquidity/maturity profile and customer concentration, and hyperscaler free-cash-flow/commitment disclosures. Without confirmation, treat a broad AI-capex short as premature.