What it is
A credit default swap (CDS) is a financial contract that works much like an insurance policy on debt. The buyer of a CDS makes regular payments to the seller; in exchange, the seller agrees to compensate the buyer if a specified borrower — the “reference entity” — suffers a defined “credit event,” typically a default, bankruptcy, or debt restructuring. The International Swaps and Derivatives Association (ISDA), the trade body that standardizes these contracts, defines a CDS as an arrangement in which the seller “agrees to confer a financial benefit on” the buyer if a credit event affecting a third party occurs.
Crucially, a CDS buyer does not need to own the underlying bond. This means CDS contracts can be — and routinely are — used for speculation as well as hedging: an investor with no exposure to a company’s debt can still buy protection against its default, effectively betting that the company will struggle. The price of that protection is quoted as a spread, often expressed as the annual cost to insure $1 million of a company’s debt for five years. A rising spread signals that the market sees default risk increasing; a falling spread signals the opposite.
CDS contracts trade over the counter between banks, hedge funds, and institutional investors rather than on public exchanges, though ISDA’s standardized documentation and post-2008 clearing reforms have made pricing more transparent than it once was.
Why it matters for AI governance and narratives
CDS spreads have become one of the few hard, market-generated signals in the AI buildout story — a counterweight to the uniform optimism of equity markets and corporate announcements. According to Semafor’s reporting, investors are now paying sharply more to insure against default by the hyperscalers and “neoclouds” financing the AI infrastructure boom with debt: the cost to insure $1 million of CoreWeave’s bonds for five years has risen from under $40,000 a year ago to roughly $100,000 today, while Oracle’s equivalent cost has risen 12% in a single week to about $28,000. CoreWeave’s debt load has reportedly tripled over the past year to more than $50 billion, and Oracle’s roughly $160 billion in debt sits just above junk-rated territory, with investor anxiety compounded by setbacks at an Oracle-financed New Mexico data center.
This matters for how AI is narrated because it exposes a split the equity story obscures: credit markets are differentiating sharply between AI infrastructure players by balance-sheet quality, while the broader AI growth narrative — told through earnings calls, product launches, and capex announcements — has largely stayed undifferentiated. A rising CDS spread is a claim made with money, not rhetoric, about whether a specific company’s AI bet will be repaid. That makes it a useful corrective lens for an observatory tracking framing contests: when builder-ecosystem communications emphasize unlimited AI demand, the debt-insurance market is one of the few venues pricing the alternative scenario.
Key facts and dates
The CDS market grew explosively in the years before the 2008 financial crisis, from roughly $6 trillion in notional value in 2004 to a peak near $57–62 trillion by mid-to-late 2007, according to Bank for International Settlements data cited in NBER research. That crisis is inseparable from the instrument’s reputation: AIG Financial Products had written CDS protection on more than $500 billion of assets, including $78 billion tied to mortgage-related collateralized debt obligations. When ratings downgrades in September 2008 triggered massive collateral calls AIG could not meet, the U.S. government intervened with what became one of the largest bailouts of the crisis — a episode documented in SEC congressional testimony and subsequent NBER and Federal Reserve research on AIG.
The market contracted sharply after 2008, falling to about $41 trillion in notional value by year-end, and has continued to shrink and consolidate since, with current estimates of notional CDS outstanding in the low single-digit trillions of dollars globally — far below the pre-crisis peak, though industry market-research estimates of the broader “CDS market” (which sometimes include revenue and services, not just notional exposure) vary widely and should be read with caution. ISDA has published successive standardized definitions governing CDS contracts, most recently a comprehensive 2014 update that remains the market standard. In the current AI financing cycle, CDS spreads on CoreWeave, Oracle, Meta, Amazon, and Alphabet have been cited by Reuters/LSEG and S&P Global Market Intelligence data, reported via Semafor, as trading at or near record highs for several of these names as of mid-October 2026.
Where to learn more
- ISDA — Credit Default Swap definitions and standard documentation — the industry body’s official contractual framework for CDS.
- Semafor — “The cost of insuring against AI company defaults spikes” (Oct. 15, 2026) — the reporting that prompted this explainer, with specific CoreWeave and Oracle spread data.
- U.S. SEC — Testimony Concerning Credit Default Swaps, Erik Sirri (Oct. 15, 2008) — primary-source regulatory account of the CDS market’s role in the 2008 crisis.
- NBER Working Paper — “AIG in Hindsight” (McDonald & Paulson) — academic analysis of AIG’s CDS exposure and collapse.