CoreWeave's $8.5B GPU debt
CoreWeave raised up to $8.5 billion in GPU‑backed debt from banks to expand cloud capacity under a deal tied to Meta, creating one of the largest GPU-financing transactions this cycle. The move signals heavy institutional appetite for GPU-backed infrastructure financing as AI demand grows. (x.com)
CoreWeave structured the financing as a delayed‑draw term loan (DDTL 4.0) that lets the company draw roughly $7.5 billion initially with an expansion feature tied to asset stabilization, and the facility is non‑recourse and secured by substantially all assets of CoreWeave Compute Acquisition Co. VIII, LLC. (investors.coreweave.com) The deal received investment‑grade ratings — A3 from Moody’s and A (low) from DBRS — marking the first time a high‑performance computing/GPU‑backed financing achieved investment‑grade status. (investors.coreweave.com) Lenders split the financing into a floating‑rate tranche priced at SOFR + 2.25% and a fixed‑rate tranche at about 5.9%, a notable step down from the roughly 9% coupons on some of CoreWeave’s earlier fixed‑rate debt. (datacenterdynamics.com) Mitsubishi UFJ Financial Group and Morgan Stanley served as co‑structuring agents and joint bookrunners, with Goldman Sachs and JPMorgan as coordinating lead arrangers; the facility was oversubscribed and had anchor participation from Blackstone Credit & Insurance. (uk.finance.yahoo.com) The loan is secured in part by a long‑term customer backlog with Meta: CoreWeave signed an initial agreement to supply up to $14.2 billion of compute capacity to Meta in September 2025. (bloomberg.com) Bloomberg and datacenter reporting say CoreWeave and Meta expanded that arrangement further this year — adding roughly $5 billion to the original deal and creating a multi‑year contract backlog north of $19 billion that underpins the facility. (datacenterdynamics.com) NVIDIA made a strategic equity commitment to CoreWeave in January 2026, buying Class A stock for $2 billion at $87.20 a share as part of a collaboration to build more than 5 GW of AI compute capacity by 2030. (investors.coreweave.com)
