OpenAI has signed a five-year, $11.9 billion agreement to use CoreWeave’s cloud infrastructure to train and deploy cutting-edge AI models. OpenAI will continue to host workloads on the Azure cloud platform, maintaining its partnership with Microsoft until at least 2030.
As part of the agreement, CoreWeave will provide OpenAI with high-performance AI infrastructure. CoreWeave’s cloud uses a variety of NVIDIA cards, including NVIDIA GB200 NVL72 super accelerators. In total, CoreWeave has purchased more than 250,000 NVIDIA AI accelerators, including the H100 and H200.
Image source: CoreWeave
Under the terms of the signed agreement, OpenAI will receive a stake in CoreWeave in the form of securities worth $350 million. CoreWeave itself recently filed for an initial public offering (IPO) on the Nasdaq exchange. The procedure will be carried out with the participation of Morgan Stanley, JPMorgan Chase and Goldman Sachs, and CoreWeave is expected to receive a market valuation of more than $35 billion.
It should be noted that Microsoft also leases computing power from CoreWeave: the Redmond giant will spend almost $10 billion for these purposes between 2023 and 2030. In 2024, Microsoft accounted for 62% of CoreWeave’s revenue, which grew to $1.9 billion. In addition, Microsoft leases accelerators from Oracle for OpenAI.
Thus, as TechCrunch notes, the competition between OpenAI and its investor Microsoft is intensifying. Tensions between the parties have been growing for several years, as OpenAI is actively strengthening its position in the global AI market. In particular, OpenAI competes with Microsoft for corporate clients.