Nvidia Credit Default Swaps

NVDA · 5-year USD senior corporate CDS · reported trades & model scenarios · basis points
Drag = pan · shift+drag = box zoom · scroll = X zoom · drag/scroll Y axis = scale · double-click = resetOne finger = pan · two fingers = zoom · drag Y axis = scale
Nvidia · 5-year USD CDS—
Latest reported spread——
Model · buyer pays upfront——
Model · buyer receives upfront——
Public trade coverage——
Coloured dotted lines are model scenarios, not market quotes or confidence bounds. Faint white dotted connectors bridge gaps between reported observations. The public tape omits the direction of the upfront payment. White points are explicit reported spreads. Dates without eligible trades stay empty.
What this study measures, model assumptions & updates

A five-year credit default swap transfers the reference company’s credit risk for that tenor. Higher spreads mean more expensive protection. 100 basis points = 1% annual spread. A CDS spread is not a direct default probability.

The source is DTCC’s public SEC security-based swap transaction reports, not a continuous dealer quote feed. The archive begins 2 September 2024; each company’s line begins when eligible records are available. We use execution dates, deduplicate reports and apply later corrections/retractions. The solid white series is the unweighted daily median of explicitly reported running spreads. These are trade observations, not end-of-day closes. Gaps longer than seven calendar days in reported spreads are joined by faint dotted white connectors between observed endpoints. These connectors add no observations or hover dots inside gaps. Model lines still break across long gaps; missing records are never forward-filled.

Records must identify Nvidia, USD notional, senior single-name corporate CDS, a trade event, a remaining tenor from 4.4 to 5.6 years, and no package or nonstandard-term flag. Bond identifiers and public underlier names identify the company. Coupon-only records are not interpreted as market spreads: a 100 bp or 500 bp fixed coupon can coexist with a very different market spread.

Upfront model: for trades with an exact uncapped notional, a USD upfront payment and a 100/500 bp coupon, a flat-hazard model estimates an equivalent running spread. It uses 40% recovery, quarterly IMM dates, ACT/360 coupon accrual and accrued premium on default. The flat discount rate is the latest FRED five-year Treasury yield (DGS5) available on or before the execution date, at most seven days old. This Treasury proxy, flat curve, simplified payment timing and treatment of the fee as net cash settlement differ from full ISDA-standard pricing. These estimates are approximate.

The tape reports the upfront amount without its payment direction. Orange assumes the protection buyer pays that amount; teal assumes the buyer receives it. Each plotted value is the daily median of that scenario’s feasible trades. The scenarios may use different trade counts and are not bounds on a uniquely observed spread. An infeasible scenario stays blank. Capped notionals such as “5,000,000+” cannot recover an accurate upfront percentage and are excluded from the model; their explicit reported spreads can still be used.

Flagged model estimates: an isolated estimate based on one trade is flagged when it exceeds four times the median of at least six neighboring estimates within 31 calendar days and has no reported spread at least half as high. These values remain in the downloadable history and readouts; tick “Show flagged estimates” to plot them. A flag is a statistical review signal, not proof that the source is wrong. Reported spreads are never hidden by this filter. Flags can change as nearby observations arrive.

Automatic checks run every four hours. The previous 15 calendar reports are downloaded again to catch late corrections, and missed dates are recovered after an outage. Reports can lag trades; sparse activity can leave the most recent plotted date unchanged even when checks succeed. Failed downloads, stale public reports or a suspicious historical regression retain the saved chart and show a warning. The web page checks for new saved data every 15 minutes and when revisited. This public tape cannot guarantee every market trade or reproduce a proprietary Bloomberg CDS quote.

Download observations CSV · Data & source metadata · Latest update status