CoreWeave has used its Nvidia H100 GPUs to secure $2.3 billion in debt financing, turning the equipment at the center of its cloud business into collateral for expansion. The deal reflects how the AI industry's growing need for computing power is prompting companies to seek new ways to finance costly infrastructure.
Chips become part of the financing
The lenders include Blackstone, BlackRock, and Carlyle. By pledging GPUs, CoreWeave offered lenders an asset connected directly to its operations. The arrangement also required the company and its lenders to agree on how much collateral would be included, how quickly the GPUs would depreciate, and how that schedule would compare with loan repayment.
CoreWeave CEO and co-founder Michael Intrator described borrowing against the company's assets as a cost-effective way to reach debt markets. For lenders, the collateral provides a basis for the loan beyond the company's promise to repay. The structure ties financing terms to both the value of the equipment and the schedule for paying down the debt.
Why CoreWeave has sought-after hardware
CoreWeave is backed by Nvidia and has access to Nvidia chips that are in short supply. That access gives the cloud computing company an advantage as AI companies build the next generation of AI supercomputers and seek computing capacity.
The GPUs are both a resource CoreWeave can use to serve customers and an asset it can pledge to raise money. That dual role illustrates how central specialized hardware has become to the economics of AI cloud services. Access to chips can support expansion, while the equipment itself can help a company obtain the capital needed to grow.
What the funding will support
CoreWeave plans to use the new financing to acquire additional GPUs, build data centers, and hire staff. Those investments are intended to increase its capacity to meet rising demand for AI computing power.
The debt deal follows an earlier equity financing. Earlier this year, CoreWeave raised $421 million at a $2 billion valuation. The company also aims to have 14 data centers in the U.S. by the end of 2023. Together, the funding and expansion plans show the scale of infrastructure the company expects to build.
A wider shift in AI finance
Jack Clark, co-founder of AI startup Anthropic, sees the deal as part of a broader change in how capital is flowing into AI. As the sector becomes more capital-intensive, he expects companies and cloud providers to use more complex financing approaches.
Clark also points to the possibility that assets such as chips, and signs of future demand such as pre-committed customer contracts for cloud allocations, could become financial instruments. In that picture, AI infrastructure becomes more closely connected to the wider capital economy.
CoreWeave's loan offers a concrete example of that shift: the company has secured financing by using the hardware it needs to expand. The approach links the growth of AI computing to decisions about collateral, depreciation, repayment, and investment in physical infrastructure.