The Yield Curve, the Debt Market and CoreWeave

A short-rate market pricing another Fed hike, a long end that rallied as risk appetite cracked, and the first real credit test of the AI infrastructure borrowing boom. Every chart in this deck is built from the FinzData database: point-in-time macro vintages, CFTC positioning, Fed liquidity and foreign flows, survivorship-free prices, and CoreWeave's own filings, insider trades, 13F owners and short interest.

Nine FinzData datasets feed this deck, each available through the same REST API, Python client (pip install finzdata) and MCP server for AI agents that our customers use. The endpoint for each dataset is printed on every slide, so any number here can be pulled and checked in seconds.

The 2-year Treasury at 4.77% sits 89bp above the 3.88% effective fed funds rate: the market is pricing tighter policy. The 10-year fell 12bp in a week to 5.10%. CoreWeave shares fell 25% in the week to $61.39, and its five-year CDS reached a record near 1,195bp on October 15, according to The Motley Fool.

From FinzData macro vintages: over twelve months the 2-year is up 127bp and the 5-year 128bp, while the 30-year is up 79bp. In the latest week bills and the 2-year rose while the 10-year fell 12bp and the 20- and 30-year fell 16-17bp, a flattening led by a falling long end.

The 10-year minus 2-year spread has narrowed from 74bp on January 30 to 35bp; the 10-year minus 3-month fell 16bp in a week to 83bp. Both are still positive. A flattening curve with hikes priced at the front is the pattern to watch, not yet a recession signal.

Effective fed funds fell to 3.63% by July and has risen to 3.88%. The 3-month bill (4.28%) and the 2-year (4.77%) both trade well above it. For leveraged borrowers this gap sets the cost of the next loan.

The 10-year TIPS real yield is 2.78%, up from 2.62% a month ago; the 10-year breakeven is 2.35%, slightly lower. Nearly all of the long yield is real return: the market is charging for tighter policy and term risk, not for runaway inflation.

FinzData COT data (CFTC Traders in Financial Futures) show leveraged funds net short 2.14m 10-year note contracts on October 6, a position more short than in 92% of weeks since 2019, and 0.73m Ultra Bond contracts (9th percentile). Asset managers are long 2.83m 10-year contracts. That mirror image is the cash-futures basis trade: hedge funds own cash Treasuries financed in repo and short futures. It is a leveraged position, which is why funding conditions on the next slide matter.

FinzData liquidity data: Fed net liquidity (Fed assets minus the Treasury General Account minus reverse repo) is $5.87tn, the reverse repo facility holds just $0.1bn, bank reserves are $3.02tn and the TGA $883bn. The St. Louis Fed stress index (−0.72) and the Chicago Fed NFCI (−0.50) are both below zero, meaning conditions are looser than average. This is a credit event in one corner of the market, not a system-wide funding squeeze.

From FinzData's TIC dataset: foreign investors bought a net $108.9bn of Treasuries in August 2026, of which official holders $18.1bn, and Treasuries in custody at the New York Fed for foreign official accounts total $2.55tn. The long-end rally has a buyer base behind it.

In the week to October 16 long Treasuries (TLT) gained 2.7% and investment-grade corporates (LQD) 1.0%, carried by falling yields, while high-yield ETFs fell (HYG −0.5%, JNK −0.6%). IG up and HY down in the same week means high-yield spreads widened. Returns are price-only.

CoreWeave and Nebius each fell 25.2% in the week; Oracle fell 9.3%, Nvidia 6.6% and SPY 2.0%. The selling concentrated on the companies most dependent on debt to fund data-centre buildouts.

FinzData point-in-time fundamentals, taken from CoreWeave's XBRL filings and stamped with each filing date: debt (carrying amount) rose from $11.2bn at June 30, 2025 to $35.6bn at June 30, 2026, and operating lease liabilities from $3.5bn to $16.3bn, about $51.9bn together. Against that, the contracted backlog (remaining performance obligations) is $103.7bn. Second-quarter revenue was $2.58bn, interest expense on debt $0.56bn (22% of revenue) and the net loss $0.63bn; cash was $5.52bn. Because each value carries the date it was filed, a backtest sees only what was public at the time.

FinzData insider data (Form 4, open-market sales, de-duplicated across joint filers): insiders and 10% holders sold $4.94bn of CoreWeave stock in 2026 through September 22 at an average of $106.77. Magnetar sold $2.40bn, Brian Venturo $0.89bn, CEO Michael Intrator $0.83bn and Brannin McBee $0.55bn; 47% of the total was under pre-arranged 10b5-1 plans. Institutions moved the other way: FinzData 13F data show holders rising from 782 to 966 in Q2 2026, with BlackRock adding 14.7m shares and Invesco 11.4m, while Magnetar cut 16.5m. Nvidia holds 47.2m shares. Short interest was 60.0m shares on September 30.

The bond and CDS figures here are not FinzData data; they are as reported. CNBC put the 9.625% notes due 2032 at about 14.8% (briefly 15%) and the 9.75% 2031s at about 14.9% on October 14; The Motley Fool reported the 2031s above 16% and a record five-year CDS near 1,195bp on October 15, with some bonds near 79 cents. S&P rates CoreWeave B+. The $51.9bn of debt and leases is from FinzData's filing data. Results are due in November.

Short rates pricing a hike raise the cost of new borrowing. Issuance is enormous: Alphabet, Amazon, Meta and Oracle had sold more than $228bn of bonds in 2026 by October 13 (LSEG via CNBC). And investors are sorting by balance sheet, buying Treasuries and IG while selling high yield and the most leveraged AI names, against a backdrop of ample liquidity and low stress. What to watch: the December Fed decision and the 2-year, the 2s10s, leveraged-fund shorts in Treasury futures, and CoreWeave's November 10-Q.

Datasets and endpoints: macro vintages (/v1/macro), COT (/v1/cot), liquidity and TIC (/v1/liquidity, /v1/liquidity/tic), daily prices (/v1/prices/daily), point-in-time fundamentals (/v1/fundamentals), insiders (/v1/insiders), 13F (/v1/13f/holders) and short interest (/v1/short_interest). Bond yields, CDS and ratings are as reported by the press and marked as such. Nothing here is a recommendation to buy or sell any security.