CoreWeave's revenue jumps 112% while debt climbs to $35.6B amid AI compute shift

CoreWeave says AI workloads are no longer isolated training jobs but a continuous cycle of training, inference, evaluation and redeployment. This shift turns compute from a large upfront purchase into a recurring expense that grows with each application in production. The company reports that its revenue rose 112% year-on-year to $2.575 billion, reflecting demand for that ongoing GPU access. At the same time, total indebtedness climbed to $35.6 billion as it funds the infrastructure needed for the steady compute stream.

To sustain the continuous compute model, CoreWeave is planning $35 billion to $39 billion of capital expenditure in 2026 and retains $10 billion of undrawn credit for additional borrowing. Quarterly interest expense jumped 140% to $640 million, contributing to a net loss of $626 million despite the revenue gain. Operating expenses exceeded revenue by $49 million, yielding an operating loss. The firm notes that most of the revenue increase came from existing customers, with about 93% of the growth attributable to expansion within the current base. This reliance on a small set of large clients means that any reduction in their spend would directly affect CoreWeave’s top line.

CoreWeave also warns that it cannot guarantee it will maintain its current growth rate or achieve positive net income in the future. Competition from established cloud providers such as AWS, Microsoft Azure and Google, some of which are also current customers, could limit its ability to win new workloads. The concentration of revenue, with three customers accounting for 36%, 26% and 10% of quarterly sales, heightens exposure to shifts in those accounts’ strategies. Together, high debt, heavy capex and customer concentration create points where the continuous compute model may become unsustainable if market conditions change.