CoreWeave Secures $2 Billion Nvidia Investment, Boosts Stake to 11.5%
On January 26 Nvidia made a $2 billion equity investment in CoreWeave at $87.20 per share, raising its stake to 11.5%. CoreWeave reported over $1 billion in quarterly revenue, holds a $55.6 billion contract backlog and carries $841.4 million in debt through Q3 2025 while targeting 5 GW of AI data centers by 2030.
1. Nvidia’s $2 Billion Equity Infusion
On January 26, Nvidia committed $2 billion in a strategic equity investment in CoreWeave, acquiring new shares at $87.20 each and increasing its ownership stake to 11.5%. This transaction underscores Nvidia’s confidence in CoreWeave as a critical partner for AI infrastructure, and it represents one of the largest single equity investments by Nvidia in a cloud computing provider to date.
2. Massive Backlog and Revenue Momentum
CoreWeave’s order backlog has surged to $55.6 billion, reflecting multi-year commitments from hyperscale AI and enterprise clients. The company reported more than $1 billion in quarterly revenue for the first time, marking a year-over-year increase of over 100%. This rapid top-line expansion highlights strong market demand for CoreWeave’s GPU-accelerated services, particularly in generative AI model training and inference workloads.
3. Leverage and Interest Expense Challenges
Despite robust growth, CoreWeave carries significant leverage, with total interest expense reaching $841.4 million through the first three quarters of fiscal 2025. Rising benchmark rates have elevated financing costs, pressuring net margins even as gross margins benefit from scale. Investors are monitoring the company’s ability to deleverage and manage cash flow as it ramps capital expenditure for data center expansion.
4. Aggressive Data Center Build-Out
CoreWeave is on track to deploy 5 gigawatts of AI-optimized data center capacity by 2030, up from approximately 1.2 gigawatts in service today. This build-out includes partnerships for land and power procurement in key U.S. markets and leverages Nvidia’s hardware road map. Management forecasts that utilization rates will exceed 70% by 2027, driven by sustained demand for large-scale model training and latency-sensitive inference workloads.