A look at CoreWeave, whose financing strategy involves using GPUs as collateral for large loans, which enabled rapid expansion but resulted in $11.2B in debt
Rashi Shrivastava / Forbes : Bluesky: @rashis Bluesky: Rashi Shrivastava / @rashis : A series of bold bets and calculated risks helped Michael Intrator turn @CoreWeave from a no-name crypto miner into a $50 billion company, serving compute to the likes of OpenAI and Microsoft. Elated to share my latest @forbes.com cover story with @pheebini.bsky.social www.forbes.com/sites/rashis...
Context & Ripple Effects
CoreWeave’s GPU-backed borrowing was established early: it raised $2.3B against Nvidia-chip collateral in 2023, then continued adding debt as it built out its specialized cloud offering.
The company’s funding cadence accelerated in 2024, including a $7.5B debt raise following $1.1B in equity funding. This profile puts that expansion strategy in focus as CoreWeave serves major compute customers including OpenAI and Microsoft.
First-order effects
- CoreWeave has financed rapid capacity expansion with GPU-secured loans, but now carries $11.2B of debt; its operating model is therefore closely tied to keeping those GPU assets productive.
- The arrangement gives CoreWeave’s lenders direct exposure to the value and utility of the GPU fleet used to support its cloud-compute customers.
Second-order effects
- CoreWeave’s financing structure raises the bar for other specialized GPU clouds: competing for capacity may require either comparable access to asset-backed credit or more equity capital.
- Large customers such as Microsoft and OpenAI gain another scaled compute supplier, while CoreWeave faces greater pressure to convert contracted demand into cash flow sufficient to support its debt load.
Third-order effects
- If GPU-collateralized lending remains viable through expansion cycles, compute infrastructure can be built increasingly through structured debt rather than solely through equity, concentrating advantage among operators with financeable fleets and large customers.
- That model also makes AI-cloud growth more sensitive to asset utilization and collateral values: a sustained gap between financed capacity and customer demand would test lenders’ and operators’ assumptions.
The trend: AI infrastructure is becoming a finance-led buildout in which GPUs function both as scarce computing equipment and as collateral for scaling cloud capacity.