Nvidia chips will be used for training and operating OpenAI AI
Oracle plans to purchase Nvidia chips for its new data center. The corporation will allocate US$40 billion toward this purchase, which will be one of the largest acquisitions in the world.
The chips will serve a data center in Texas that OpenAI intends to use. According to the Financial Times, this will be the first facility of its kind in the U.S. It is part of OpenAI and SoftBank’s long-term plan to develop large-scale AI projects. The company expects to complete construction of the Texas data center in 2026. The center should provide 1.2 GW of computing power, which would make it the largest in the world. This indicates an increase in demand for AI training and support infrastructure.
Project features
Oracle plans to purchase 400,000 new GB200 series chips. Nvidia designed these advanced microchips—considered its most sophisticated—to work with AI. The company is expected to lease the new computing power and chips to OpenAI.
1. Crusoe and Blue Owl Capital own the site where the data center is under construction.
2. The companies raised a total of US$15 billion through debt and equity to finance the project.
3. Construction of the facility began in summer 2024 and is scheduled for completion in 2026.
4. The data center will consist of eight buildings.
5. Oracle has signed a 15-year lease for the facility.
Importance for OpenAI
JPMorgan was the primary lender for the project, providing US$9.6 billion in financing. This amount includes a US$7.1 billion loan. Crusoe and Blue Owl each invested US$5 billion. The funds will be provided in cash.
The project is strategically important for OpenAI. With its help, the startup intends to reduce its dependence on Microsoft. Microsoft was a key provider of cloud technologies. Microsoft invested about US$14 billion in OpenAI, most of which went toward financing cloud capabilities. However, Microsoft’s computing power is insufficient for OpenAI’s active development. Consequently, the companies have decided to end their exclusive partnership.
Projects of other companies
Demand for Nvidia chips is also growing among other companies. In June 2025, Elon Musk announced plans to expand his Colossus data center in Memphis. The facility will require approximately one million processors to operate. Unlike Oracle, Musk’s data center uses previous-generation H100 and H200 chips. Musk intends to implement an ambitious project: the first gigawatt AI training supercluster.
Amazon is also building its data center. The Virginia facility will have a capacity of over 1 GW.
