CoreWeave Contract Repricing and Renewal Thesis

Bottom line: There is no near-term mass repricing of CoreWeave’s installed base. Only a “very limited” portion is nearing renewal. Near-term margin improvement comes mainly from new contracts, while legacy repricing is longer-dated upside.

Key point	Timing / implication
Only a very limited portion of the fleet is nearing renewal	No disclosed renewal wave or maturity schedule
Typical contracts last ~5 years	Legacy pricing remains until individual contracts expire
Prices increased ~25% across SKUs in July	Benefits new pricing; unclear whether existing contracts reset
Q2 new contracts have 5–10 points higher contribution margins	Improvement is happening now, before major renewals
Older GPUs remain sold out; an A100 was contracted through 2029 at an attractive price	Strong evidence legacy assets can be repriced profitably
Expired capacity can be recontracted, sold on shorter 2–3 year terms, or moved into inference	Shorter terms and inference may generate higher margins
Asset-level debt should be repaid by initial contract expiry	Renewal revenue should be highly incremental and cash-generative

Thesis takeaway: Don’t underwrite a large 2026–27 renewal uplift. Underwrite improving new-contract economics now, with legacy-fleet repricing as an unquantified longer-term option.