U.S. GPU cloud provider CoreWeave announced on Monday that it has signed an initial $6.3 billion agreement with Nvidia, securing the AI chipmaker guaranteed access to any unsold data center capacity through April 2032.

The deal expands on an April 2023 agreement and further strengthens CoreWeave's position as a key Nvidia cloud partner. In addition to being a major supplier of GPUs to CoreWeave, Nvidia also holds about 7% of the company's Class A shares.

CoreWeave operates AI data centers in the U.S. and Europe, renting out computing power built on hundreds of thousands of Nvidia GPUs. The agreement provides CoreWeave with a backstop against potential demand fluctuations, ensuring utilization of its infrastructure regardless of the end customer. Barclays analysts said Nvidia's incremental spending represents a “healthy diversification” for CoreWeave beyond its two largest customers, Microsoft and OpenAI.

Following the announcement, CoreWeave's shares rose 7.6%, extending a rally of more than 200% since its March IPO, which has pushed its market capitalization above $58 billion.

The company has secured other major contracts this year, including a five-year, $11.9 billion deal with OpenAI signed before its IPO, followed by an additional $4 billion agreement in May. These commitments, alongside the latest Nvidia order, highlight surging demand for AI infrastructure.

In its second-quarter earnings, CoreWeave reported revenue of $1.21 billion, up 207% year-on-year, though operating expenses nearly quadrupled to $1.19 billion, resulting in a net loss of $291 million. Despite financial pressures, CoreWeave has emerged as one of the biggest beneficiaries of the generative AI boom.

By securing Nvidia's long-term backing, CoreWeave not only gains stability in a competitive cloud services market but also reinforces the semiconductor industry's broader trend: AI demand continues to drive unprecedented investment in computing capacity.