Nvidia commits $6.3 billion to secure GPU cloud capacity from CoreWeave extending through 2032.

U.S. AI chipmaker Nvidia has agreed to purchase unsold compute capacity from AI cloud provider CoreWeave in a deal valued at $6.3 billion, according to filings with the U.S. Securities and Exchange Commission (SEC). The agreement ensures that Nvidia will acquire any residual CoreWeave data center capacity not taken up by other customers through April 13, 2032. However, Nvidia's obligation to purchase is contingent on the availability of services and the fulfillment of delivery requirements.

CoreWeave disclosed the arrangement in its SEC filing, stating: "Under the terms of agreement, in instances where the company's data center capacity is not fully utilized by its own customers, Nvidia is obligated to purchase the residual unsold capacity through April 13, 2032, subject to any termination described below and satisfaction of delivery and availability of service requirements." The volume of compute that Nvidia may ultimately purchase has not been disclosed.

This arrangement follows Nvidia's recent $1.5 billion agreement with AI cloud provider Lambda, under which the chipmaker will lease back as many as 18,000 of its own GPUs over five years. Both deals underscore Nvidia's broader strategy of securing reliable access to GPU-intensive cloud resources amid strong demand for AI workloads. The agreements reflect the company's efforts to maintain control over GPU supply chains while ensuring access to infrastructure critical for training and deploying large-scale AI models.

CoreWeave has become one of Nvidia's key cloud partners, offering GPU-based compute to enterprises and AI developers, making these capacity agreements essential to Nvidia's infrastructure security strategy.