CoreWeave Closes $2.6 Billion Loan Facility (DDTL 5.5)
$2.6bn DDTL at SOFR+550, ~5yr, Ba2/BB+, JPMorgan/MUFG; >$30bn debt+equity raised YTD. ~5yr facility against ~3yr average customer contracts — lenders now financing shorter-dated contracts, taking re-lease/residual risk.
Lead
CoreWeave closed a $2.6 billion delayed draw term loan ("DDTL 5.5 Facility") priced at SOFR + 5.50%, tenor approximately five years, rated Ba2 (Moody's) / BB+ (Fitch). JPMorgan and Mitsubishi UFJ Financial Group acted as joint lead arrangers and bookrunners. Proceeds fund purchase and deployment of HPC-backed infrastructure dedicated to customer contracts. The company states it has now raised more than $30 billion in debt and equity capital year to date. The transaction was meaningfully oversubscribed.
Key figures
| Item | Value |
|---|---|
| Size | $2.6bn |
| Pricing | SOFR + 5.50% |
| Tenor | ~5 years |
| Underlying customer contracts | ~3 years average |
| Ratings | Ba2 / BB+ |
| Arrangers | JPMorgan, MUFG |
| Raised YTD (debt + equity) | >$30bn |
The structural change, in the company's own words
"Lenders are now comfortable financing shorter-dated contracts, which allows us to target a wider variety of customers, including global enterprises that typically favor shorter-term agreements" — Brannin McBee, co-founder and CDO
That is the substantive novelty and it is a term-mismatch statement: a ~5-year facility against ~3-year average customer contracts. The facility explicitly contemplates renewal of existing contracts or re-leasing capacity to other customers at contract end, subject to credit-agreement criteria. The residual-value risk on the GPU fleet has, in other words, been partly moved onto the lender — priced, presumably, in the spread.
Why it is on this rung — read as a series, not a row
This rung's value here is comparative. Across three CoreWeave facilities on the same borrower and broadly the same collateral, inside about four and a half months (corrected 2026-09-24: DDTL 4.0 closed 2026-03-31, not in 2025 — see DDTL 4.0):
| Facility | Date | Size | Pricing | Rating |
|---|---|---|---|---|
| DDTL 4.0 | 2026-03-31 | $8.5bn | SOFR + 225 / ~5.9% fixed | A3 (Moody's) / A (low) (DBRS) |
| DDTL 5.0 | 2026-05-18 | $3.1bn | SOFR + 450 | Ba2 / BB+ |
| DDTL 5.5 | 2026-08-10 | $2.6bn | SOFR + 550 | Ba2 / BB+ |
SOFR+225 → +450 → +550, with ratings falling from A3/A(low) to Ba2/BB+, and facility size shrinking $8.5bn → $3.1bn → $2.6bn. That is a cost-of-capital curve on GPU-backed debt, and it exists only as a comparison — no single release contains it.
Correction, 2026-09-24: with the DDTL 4.0 release now read, the first step (+225 to +450) lines up with a change of customer: DDTL 4.0 finances a contract with a "leading AI enterprise" and is rated investment grade, while DDTL 5.0 finances two customers stated to be below investment grade. Much of that first step is probably customer credit rather than the passage of time, and DDTL 4.0's second rating is from DBRS, not Fitch.
Limitations — and the honest complication
- Issuer press release, not a filed credit agreement. No covenants, advance rates, amortisation schedule, draw conditions or collateral haircuts are disclosed. Those are where the real risk sits.
- The three facilities are not strictly like-for-like. DDTL 4.0 was rated investment grade and structured differently (fixed-rate component, different customer concentration); comparing headline spreads across differently-structured facilities overstates precision. The direction is solid; the magnitude is not a clean spread series.
- Widening spread ≠ closing market. Each of these was oversubscribed, and DDTL 5.0 priced tighter than launch. A close read that reported only the widening would be wrong — see the DDTL 5.0 file.
- ">$30bn raised YTD" mixes debt and equity and is unaudited company framing; it is a fundraising total, not leverage.
- Customer names and contract counterparty credit are not disclosed here.
Source: CoreWeave, "CoreWeave Closes $2.6 Billion Loan Facility", 2026-08-10.