Oracle is planning to spend around $40 billion on Nvidia chips to support a massive new data centre being developed by OpenAI in Texas, according to reporting by the Financial Times. The move marks one of the largest chip purchases to date and signals the growing demand for AI computing power.
The site is located in Abilene, Texas, and is described as the first US “Stargate” data centre. It’s part of a long-term plan backed by OpenAI and SoftBank to create large-scale AI facilities. When finished next year, the Texas centre is expected to provide 1.2 gigawatts of computing power, placing it among the largest in the world.
Oracle has indicated it intends to purchase around 400,000 of Nvidia’s new GB200 chips, some of Nvidia’s most advanced processors designed for training and running AI models. Oracle plans to lease the computing power from these chips to OpenAI, according to people familiar with the deal.
The site is owned by Crusoe and Blue Owl Capital, which have raised $15 billion through a mix of debt and equity to fund the project. Construction began in June 2024, and the campus will include eight buildings. Oracle has signed a 15-year lease for the site, although Stargate itself hasn’t put any money into this specific facility.
JPMorgan is providing most of the debt financing, providing two loans totalling $9.6 billion, including a $7.1 billion loan announced this week. Crusoe and Blue Owl have each added roughly $5 billion in cash to the project.
Elon Musk is expanding his own project in Memphis, Tennessee, with plans to house about 1 million Nvidia chips. That centre, called “Colossus,” is mostly using Nvidia’s earlier H100 and H200 chips. Musk said this week the next phase will become the “first gigawatt AI training supercluster.” Meanwhile, Amazon is working on a separate data centre in Virginia that will also surpass 1GW.
The Abilene project is part of a larger shift for OpenAI as it looks to lessen its dependence on Microsoft. Until now, OpenAI has relied heavily on Microsoft for access to cloud computing. Much of Microsoft’s nearly $14 billion investment in the company came in the form of cloud credits. But OpenAI has grown frustrated that Microsoft can’t keep up with its increased demand for computing power.
Earlier this year, OpenAI and Microsoft agreed to end their exclusive partnership, and are still working to resolve how long Microsoft will continue to hold licensing rights to OpenAI’s models.
Looking ahead, Stargate is expected to supply a large share of OpenAI’s computing needs. The group, which launched in early 2025, aims to raise as much as $500 billion over four years to fund more data centre projects in the US.
So far, OpenAI and SoftBank have each pledged $18 billion to Stargate. Oracle and MGX, a sovereign wealth fund based in Abu Dhabi, have committed another $7 billion each, according to people familiar with the details. The four companies will hold equity in the project, with OpenAI and SoftBank taking the lead.
Stargate hasn’t yet invested in any specific project, but its plans are expanding beyond the US. Just last week, during President Donald Trump’s visit to the Gulf region, OpenAI announced a new Stargate facility in Abu Dhabi. The site will cover 10 square miles and offer 5GW of data centre power – enough to run more than 2 million Nvidia GB200 chips. Local AI firm G42 is managing the construction.
