CoreWeave Credit Fears Echo SVB Panic, Not 2008 Crisis, Wall Street Says
TREE NEWS reports: Yields on CoreWeave’s unsecured bonds have surged to about 15%, and its credit default swap spread has blown past 1,000 basis points, triggering fears that the AI capital-spending cycle is cracking. But top Wall Street strategists argue the market is misreading the signal, drawing parallels to the 2023 Silicon Valley Bank episode rather than the 2007 subprime collapse.
What Happened
CoreWeave, a fast-growing AI cloud provider, has roughly $35 billion in total debt. Its unsecured notes — about $1.5 billion — sit at the bottom of a three-tier capital structure, behind secured delayed-draw term loans (DDTL) held in special purpose vehicles and roughly $9 billion of convertible bonds. The violent repricing of this small, structurally subordinated slice has become a proxy for the entire AI infrastructure buildout.
Barclays global macro research head Ajay Rajadhyaksha compared the panic to 2023, when a roughly $4 billion regional-bank ETF (KRE) became the tail wagging the dog. “The mistake was not recognizing that SVB had failed. It had,” he wrote. “The mistake was extrapolating from a small, volatile, structurally fragile instrument to the entire U.S. economy.” He noted that markets priced 75-100 basis points of rate cuts in 2023; the Fed instead kept hiking, the economy outperformed, and the S&P 500 rose about 40% over the following two years.
Market Implications
Goldman Sachs analysts Gabriela Borges and Maura Hager argue the unsecured bond market is the wrong benchmark for CoreWeave’s true funding cost. The company’s main financing tool is the DDTL, an asset-backed facility underwritten against specific customer contracts — such as Meta’s multi-year take-or-pay commitments — and GPU infrastructure. Goldman estimates CoreWeave’s weighted average debt cost at 800-900 basis points, with new financing likely at 750-1,050 basis points, implying only a 100-200 basis point increase.
The bull case hinges on CoreWeave raising revenue per megawatt from about $10 million to $13 million annually. At $13 million, a 5-year, 200MW contract with 80% debt financing yields a 15% unlevered internal rate of return, a 3-point spread over an 11.8% WACC. At current $10 million levels, the IRR falls to 6.3%, a negative 6-point spread. The entire thesis depends on a 30% pricing increase.
Supply-chain data undercuts the systemic-collapse narrative: TSMC beat guidance by $1 billion, ASML reported record sales, and hyperscalers are set to spend $725-800 billion on AI infrastructure in 2026. CoreWeave’s 2026 capex guidance of $35-39 billion represents just 4-5% of that total. “Even if CoreWeave stopped spending entirely, we’d lose only 4-5% of AI capex,” Rajadhyaksha said.
Still, both Barclays and Goldman flag a key difference from 2023: there is no central-bank backstop like the Bank Term Funding Program. The real risk trigger is whether compute rental rates can cover capital costs. If pricing falls as supply catches up, CoreWeave remains locked into 7- to 16-year data-center leases with costly exit terms.
Key Takeaways for Investors
- Don’t over-extrapolate: A small, subordinated bond tranche is not a reliable barometer for a trillion-dollar capex cycle.
- Watch compute pricing: The sustainability of AI infrastructure financing depends on revenue per megawatt holding above roughly $13 million.
- Hyperscaler spending is the real backstop: Funded largely by free cash flow, it is far less sensitive to credit-market stress than a single emerging cloud provider.
- No policy safety net: Unlike 2023, there is no Fed facility to break a panic loop in AI infrastructure financing.
