Please see relevant disclosures here.
Fitch assigned CoreWeave’s proposed $2.6 billion senior secured delayed-draw term loan, DDTL 5.5, a ‘BB+’ rating with a Recovery Rating of ‘RR2’ this morning. It affired CoreWeave’s Long-Term Issuer Default Rating (IDR) at ‘BB-’ with a Positive Outlook, and affirmed DDTL 5.0 at ‘BB+’/’RR2’ in the same action. The facility will be borrowed by a subsidiary SPV and used to buy GPUs and related equipment for newly signed take-or-pay customer contracts.
A delayed-drawn term loan commits the money up front and lets the borrower take it in pieces. CoreWeave pays an undrawn fee, of 0.5% a year on the two prior facilities, on whatever it hasn’t taken yet. The loans are structured this way because GPUs don’t arrive all at once. A contract is signed, servers are ordered, racks are built, and the hardware comes online over quarters. A term loan funded at close would leave CoreWeave paying full interest on capital sitting idle. The delayed draw lines the funding up with the installation schedule.
Each facility sits in its own SPV, with a first-priority security interest in that SPV’s equipment, the customer contract behind it, and the associated data center leases. The customer pays the SPV, the SPV services the debt. The template is deliberately repeatable: CoreWeave has run it six times in under two years.
DDTLs 1.0, 2.0, 2.1, and 3.0 are unrated and privately held.
DDTL 4.0 closed March 31, 2026 at $8.5 billion, borrowed by CoreWeave Compute Acquisition Co. VIII. It’s baced by roughly $19 billion of Meta take-or-pay contracts, priced at SOFR + 225 on the floating tranche with a fixed tranche around 5.9%, matures March 2032, and carries a 1.15x DSCR covenant. Moody’s rated it A3, DBRS A (low), Fitch ‘A-sf’. It’s the first investment-grade (IG) GPU-backed financing.
DDTL 5.0 closed May 18 at $3.1 billion through CoreWeave Financing DDTL V. There are two large, unnamed, non-investment-grade customers. The interest rate is SOFR + 450, floating, and the debt matures November 15, 2031. Ratings are Ba2 from Moody’s, ‘BB+’/’RR2’ from Fitch. It was the first of these to be publicly syndicated, which means it can trade in secondary rather than sit with the original lenders to maturity.
Finally, there’s DDTL 5.5, the newest debt issuance. It has the same rating as DDTL 5.0.
Every prior facility (except for DDTL 2.1) got a whole number, and each one marked a change in structure. 4.0 was investment grade off a single IG counterparty. 5.0 was the first public syndication, with non-IG customers and a wider spread. A point release suggests that DDTL 5.5 is the same financing family, with a larger deployment, no new architecture, using the 5.0 deal documents as the template. Fitch reinforced that by handing 5.5 the same rating and the same recovery notching as 5.0.
Where BB+ comes from
Two different rating frameworks are running in this stack, and the suffixes tell you which is which.
DDTL 4.0 is rated ‘A-sf’ — a structured finance rating. The instrument gets assessed on the isolated Meta cash flows and the structure around them. CoreWeave’s own corporate credit is largely beside the point.
DDTL 5.0 and 5.5 are rated ‘BB+’/’RR2’ — a corporate rating. Fitch starts at CoreWeave’s ‘BB-’ IDR and notches up for recovery. RR2 implies 71–90% recovery in a default, and it’s worth two notches. BB- plus two is BB+.
That’s the arithmetic, and it means the collateral package determines where DDTL 5.5 sits above CoreWeave, not whether it’s CoreWeave risk at all. Until the IDR moves, BB+ is where this template lands.
The rest of the stack, for scale: about $10 billion of unsecured high-yield notes at ‘BB-’/’RR4’, roughly $6.5 billion of unsecured converts (unrated), $5 billion of OEM financing (unrated), and a $281 million Magnetar deposit classified as debt.
Fitch’s headline numbers: revenue backlog of $99.4 billion as of March 31, up 284% year over year, with about 36% recognized inside 24 months. Gross EBITDA leverage excluding leases at 7.0x in 2025, forecast to 6.6x in 2026 and 4.5x in 2027. Capex around $41 billion this year and $37 billion next. Free cash flow negative through 2027. Top two customers were 65% of 1Q26 revenue.
The assumption doing the most work is the first one Fitch lists: that CoreWeave renews or replaces customer contracts expiring before DDTL 5.5 matures, at favorable GPU lease rates. Customer contracts run three to five years. The 5.0 template runs about five and a half. There’s a tail where the loan is live and the contract isn’t, and nothing observable to price it against.
There’s no 8-K for DDTL 5.5. Fitch calls it proposed, so the terms aren’t public.
Three things worth waiting for. Who the customers are — Fitch says newly signed take-or-pay, and CoreWeave’s recent announcements include OpenAI, Meta, Anthropic and Jane Street, but the facility hasn’t been tied to any of them. What it prices at, against 5.0’s SOFR + 450. And the guarantee package: DDTL 5.0 is unconditionally guaranteed by the parent, while DDTL 4.0’s parent guarantee is limited-recourse for bad acts only. The identical rating implies 5.5 follows 5.0, but that’s an inference until the filing lands.
That last one is the tell worth reading first.
If you enjoy this newsletter, consider sharing it with a colleague.
Share Buy the Rumor; Sell the News
I’m always happy to receive comments, questions, and pushback. If you want to connect with me directly, you can:
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.