CoreWeave (CRWV) J.P. Morgan 54th Annual Global Technology, Media and Communications Conference summary

Business performance and demand environment

• Achieved $5.1 billion in FY2025 revenue, with nearly $100 billion in backlog and $40 billion in new bookings in a single quarter.

• Demand for AI infrastructure is described as insatiable, with customers being turned away due to capacity constraints.

• Client base spans hyperscalers, AI labs (including nine of the top 10 global non-China labs), and rapidly growing enterprise customers.

• Enterprise segment is expanding, with a record number of new enterprise logos added in Q4.

• Financial services represent over $10 billion of current backlog, heavily utilizing infrastructure for inference workloads.

Technology and product strategy

• Focused on building and operating large-scale GPU infrastructure, primarily using Nvidia technology.

• Infrastructure is designed for seamless transition between AI training and inference, supporting both with the same hardware.

• Software stack (Mission Control, SUNK) enables efficient provisioning and operation of clusters at scale.

• Clients value the ability to use infrastructure flexibly for both training and inference, with no significant demand for edge or inference-only facilities.

• Product differentiation and operational excellence are key drivers for clients moving workloads from hyperscalers.

Financial structure and contract dynamics

• Long-term contracts (four to six years) are standard across client segments, targeting a 25% contribution margin at the SPV level.

• Contracts are structured to de-risk depreciation and capital expenditures, with high utilization rates leading to renewals and expansion contracts.

• Margin compression in Q1 is expected to reverse as new deployments ramp in Q2 and Q3, supporting guidance for low double-digit operating margins by year-end.

• Execution on infrastructure delivery has enabled a significant reduction in cost of capital, with recent debt facilities priced much lower than earlier deals.