CoreWeave's $3.1B GPU Loan Gets Its Official Legal Close, What the Kirkland and Latham Disclosures Add

The May 19 brief established the headline: CoreWeave closed a $3.1B delayed draw term loan facility, the first publicly syndicated GPU-backed financing in the AI infrastructure market. What law firm press releases add isn’t the story, it’s the paper trail. Kirkland & Ellis, advising CoreWeave, and Latham & Watkins, advising the lender syndicate, have both published transaction announcements confirming the legal close as of May 26.

Law firm disclosures are the official record. They’re not journalism, they’re transaction counsel confirming deal completion, which is a structurally different kind of evidence than a press release or news report. When adverse counsel on both sides of a deal confirm the same transaction facts, that’s the closest thing to primary transactional documentation available in the public record for a private financing.

The new details, per transaction counsel disclosure: the facility reportedly received non-investment-grade ratings of Ba2 from Moody’s and BB+ from Fitch. It’s reportedly priced at SOFR plus 4.50%. Morgan Stanley and Mitsubishi UFJ Financial Group reportedly served as lead arrangers. The facility is reportedly backed by contracts with two large enterprise customers whose identities haven’t been disclosed.

The ratings matter. Ba2/BB+ is speculative grade, the market’s formal acknowledgment that GPU-backed debt carries meaningful credit risk. That’s not a surprise given CoreWeave’s debt load and the novelty of compute infrastructure as collateral. It’s a useful signal for institutional investors evaluating GPU-backed debt as an emerging asset class. High-bandwidth memory now accounts for an estimated 52% to 63% of total AI chip component spending, according to Epoch AI’s May 21, 2026 data insight, the cost concentration that structured financing vehicles like DDTL 5.0 are built to address.

TJS Synthesis: The Ba2/BB+ ratings and SOFR + 4.50% spread are the market’s pricing of GPU infrastructure credit risk. That spread is wide, 450 basis points over the risk-free rate for a company with two undisclosed enterprise customers as primary collateral. The institutional appetite for that risk, at that price, tells you more about where investors think AI infrastructure demand is going than any analyst price target does. Watch CoreWeave’s next debt issuance for whether the spread tightens as the asset class matures.