CoreWeave’s $15 B Debt Load and $311 M Interest Expense Raise Profitability Concerns
CoreWeave carries $15 billion of debt—including leases—almost four times its trailing-12-month revenue, and incurred $311 million of quarterly interest expense, over 20% of revenue and six times its gross profit. It forecasts $12–14 billion of 2025 capex (carryover may push total near $30 billion), and GPU lifecycle uncertainty plus hyperscaler concentration heighten profitability risks.
1. Rapid Revenue Growth and Massive Infrastructure Buildout
CoreWeave has tripled its revenue over the past year, reaching $1.36 billion in Q3 2025, driven by surging demand for AI data-center services. Management projects full-year 2025 sales of $5.1 billion, up 165% from 2024’s $1.92 billion, and forecasts roughly $12 billion in revenue for 2026. To support this growth, the company plans capital expenditures of $12 billion to $14 billion in 2025, which could swell to $30 billion when carry-forward capex is included. At the end of Q3, CoreWeave operated 41 dedicated AI data centers with 590 MW of active power capacity and has contracted an additional 2.9 GW of power, positioning it to add at least 1 GW in the next 12–24 months.
2. High Leverage and Rising Interest Costs
CoreWeave’s aggressive expansion is financed primarily through debt and lease obligations totaling approximately $15 billion—nearly four times its trailing 12-month revenue. In the most recent quarter, the company incurred $311 million of interest expense, up 200% year-over-year, representing over 20% of revenue and roughly six times gross profit. With average borrowing costs elevated, continued reliance on external financing increases refinancing risk and could strain cash flow if AI compute demand growth decelerates.
3. Customer Concentration and Strategic Dependencies
Nearly all of CoreWeave’s revenue comes from a handful of hyperscale customers, including Microsoft and other cloud giants that are simultaneously expanding their own AI infrastructure. While a $6.3 billion backstop agreement with Nvidia provides some balance, the hyperscalers could shift workloads in-house if they build sufficient capacity. A meaningful slowdown in external AI spending or a pivot by key customers toward proprietary data centers would leave CoreWeave with underutilized assets and heighten existential risk.