CoreWeave, a leading AI cloud provider backed by Nvidia, has signed a new agreement with OpenAI, the creator of ChatGPT, committing to provide up to $4 billion in AI infrastructure services through April 2029. The deal, disclosed in a regulatory filing on Thursday, builds on an existing $11.9 billion, five-year contract signed in March 2025, further strengthening the partnership between the two companies.
Under the new agreement, CoreWeave will supply OpenAI with access to its high-performance data centers and Nvidia GPUs, enhancing OpenAI’s ability to train and deploy advanced AI models for its millions of users worldwide. The deal follows CoreWeave’s recent initial public offering (IPO) and comes as OpenAI seeks to diversify its compute resources beyond its long-standing partner, Microsoft.
CoreWeave’s CEO, Michael Intrator, emphasized the company’s role in supporting cutting-edge AI development, stating, “This expansion with OpenAI highlights our ability to deliver reliable infrastructure for the world’s leading AI labs.” OpenAI’s CEO, Sam Altman, added, “CoreWeave’s compute power is critical for scaling our models and serving more users effectively.”
The agreement reduces OpenAI’s reliance on Microsoft, which previously accounted for 62% of CoreWeave’s $1.9 billion revenue in 2024. It also aligns with OpenAI’s broader strategy, including its Stargate project with SoftBank and Oracle, to secure independent compute capacity. Posts on X reflected market enthusiasm, noting a 7% rise in CoreWeave’s stock price after the announcement, with investors viewing the deal as a sign of strong customer confidence.
Critics, however, point to CoreWeave’s significant debt of $7.9 billion and its dependence on a few major clients, including OpenAI and Microsoft, as potential risks. The company’s rapid growth, driven by the AI boom, has also raised questions about the sustainability of its data center expansion.
This latest deal solidifies CoreWeave’s position as a key player in the AI infrastructure market, while OpenAI continues to push the boundaries of artificial intelligence with enhanced computational resources.